Learning To Spell Research, Theory, and Practice Across PDF

cover next page >
title: Learning to Spell : Research, Theory, and Practice Across

Languages
author: Perfetti, Charles A.
publisher: Lawrence Erlbaum Associates, Inc.
isbn10 | asin: 0805821600
print isbn13: 9780805821604
ebook isbn13: 9780585115085
language: English
subject Language and languages--Orthography and spelling--Study and
teaching.
publication date: 1997
lcc: P240.2.L43 1997eb
ddc: 411
subject: Language and languages--Orthography and spelling--Study and
teaching.
cover next page >

< previous page page_ii next page >
intentionally left blank

< previous page page_iii next page >
Page iii
Learning to Spell
Research, Theory, and Practice Across Languages
Edited by
Charles A. Perfetti
University of Pittsburgh
Laurence Rieben
University of Geneva
Michel Fayol
University of Bourgogne, Dijon
LAWRENCE ERLBAUM ASSOCIATES, PUBLISHERS

1997 Mahwah, New Jersey London
< previous page page_iii next page >

< previous page page_iv next page >
Page iv
Disclaimer:
This book contains characters with diacritics. When the characters can be represented using the ISO 8859-1 character set
(http://www.w3.org/TR/images/latin1.gif), netLibrary will represent them as they appear in the original text, and most computers
will be able to show the full characters correctly. In order to keep the text searchable and readable on most computers, characters
with diacritics that are not part of the ISO 8859-1 list will be represented without their diacritical marks.
Copyright 1997 by Lawrence Erlbaum Associates, Inc.
All rights reserved. No part of this book may be reproduced in any form, by photostat, microfilm, retrieval system, or any other
means, without the prior written permission of the publisher.
Lawrence Erlbaum Associates, Inc., Publishers
10 Industrial Avenue
Mahwah, New Jersey 07430
Library of Congress Cataloging-in-Publication Data
Learning to spell/edited by Charles A. Perfetti, Laurence Rieben,
Michel Fayol.
p. cm.
Includes bibliographical references and index.
ISBN 0-8058-2160-0 (alk. paper). ISBN 0-8058-2161-9 (pbk.:
alk paper).
1. Language and languages Orthography and spelling Study and
teaching. I. Perfetti, Charles A. II. Rieben, Laurence.
III. Fayol, Michel, 1947-
P240.2.L43 1997
411-dc21 96-51911
CIP
Books published by Lawrence Erlbaum Associates are printed on acid-free paper, and their bindings are chosen for strength and
durability.
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
< previous page page_iv next page >

< previous page page_v next page >
Page v
CONTENTS
Contributors ix
Introduction
Perfetti, Rieben, Fayol xi
I
Theoretical Foundations: Writing Systems, Psycholinguistics, and
Neuropsychology 1
1
From Writing to Orthography: The Functions and Limits of the Notion
of System
Jaffr 3
2
The Psycholinguistics of Spelling and Reading
Perfetti 21
3
The Cognitive Neuropsychology of Spelling
Zesiger and de Partz 39
II
The Acquisition of Spelling 59
4
Spelling Acquisition in English
Treiman and Cassar 61
< previous page page_v next page >

< previous page page_vi next page >
Page vi
5
How Learning to Spell German Differs From Learning to Spell
English
Wimmer and Landerl 81
6
The Development of the Understanding of Number Morphology in
Written French
Totereau, Thevenin, Fayol 97
7
Lexical Spelling Processes in Reading Disabled French-Speaking
Children
Alegria and Mousty 115
8
Learning to Spell in the Classroom
Allal 129
9
Spelling and Grammar The Necsed Move
Nunes, Bryant, Bindman 151
III
The Relationship Between Spelling and Reading 171
10
Why Spelling Is More Difficult Than Reading
Bosman and Van Orden 173
11
The Inhibition of Polygraphic Consonants in Spelling Hebrew:
Evidence for Recurrent Assembly of Spelling and Phonology in Visual
Word Recognition
Berent and Frost 195
12
Children's Use of Analogy in Learning to Read and to Spell
Gombert, Bryant, Warrick 221
13
Learning to Read and Learning to Spell Are One and the Same,
Almost
Ehri 237
14
Interactions in the Development of Reading and Spelling: Stages,
Strategies, and Exchange of Knowledge
Ellis 271
15
Relations Between Word-Search Strategies and Word-Copying
Strategies in Children Aged 5 to 6 Years Old
Rieben and Saada-Robert 295
< previous page page_vi next page >

< previous page page_vii next page >
Page vii
16
Foundations of Orthographic Development
Seymour 319
17
Beginning Reading and Spelling Acquisition in French: A
Longitudinal Study
Sprenger-Charolles, Siegel, Bchennec 339
Author Index 361
Subject Index 369
< previous page page_vii next page >


< previous page page_ix next page >
Page ix
CONTRIBUTORS
Jesus Alegria Free University of Brussels
Linda Allal University of Geneva
Danielle Bchennec National Institute for Pedagogical Research, Paris
Iris Berent Florida Atlantic University
Miriam Bindman University of London
Anna M.T. Bosman University of Nijmegen
Peter Bryant University of Oxford
Marie Cassar Wayne State University
Marie-Pierre de Partz University Hospital of St.-Luc, Brussels
Linnea C. Ehri City University of New York Graduate School
Nick Ellis University of Wales
Michel Fayol University of Bourgogne
Ram Frost Hebrew University
Jean Emile Gombert University of Bourgogne
Jean-Pierre Jaffr CNRS-HESO, Paris
Karin Landerl University of Salzburg
< previous page page_ix next page >

< previous page page_x next page >
Page x
Philippe Mousty Free University of Brussels
Terezinha Nunes University of London
Charles A. Perfetti University of Pittsburgh
Laurence Rieben University of Geneva
Madelon Saada-Robert University of Geneva
Philip H.K. Seymour University of Dundee, Scotland
Linda S. Siegel University of British Columbia
Liliane Sprenger-Charolles CNRS & Ren Descartes University, Paris
Marie-Genevive Thevenin University of Bourgogne
Corinne Totereau University of Bourgogne
Rebecca Treiman Wayne State University
Guy C. Van Orden Arizona State University
Nicola Warrick University of Oxford
Heinz Wimmer University of Salzburg
Pascal Zesiger University of Geneva
< previous page page_x next page >

< previous page page_xi next page >
Page xi
INTRODUCTION
The word spelling can evoke images of classroom drill and testing, children writing strings of letters as a teacher pronounces
words ever so clearly. In parts of the United States, it can also bring an image of specialized wizardry and schoolroom
competition, the spelling bee. For countless adults, who confess with self-deprecation to being terrible spellers, it is a reminder
of a mysterious but minor affliction that the fates have visited on them. Beneath these popular images, spelling is a human
literacy ability that reflects language and nonlanguage cognitive processes. This collection of chapters presents a sample of
contemporary research, across different languages, that addresses this ability.
To examine spelling as an interesting scientific problem, we think several important perspectives are required. For one, spelling
is the use of conventionalized writing systems that encode languages. Thus, part of the spelling problem is how an orthography
(the details of a writing system) reflects the structure of a spoken language. This issue entails an examination of writing systems
and the influences writing systems and their orthographies might place on spelling. This constitutes one perspective in this
volume.
A second perspective is to ask how children learn to spell. If the designs of writing systems and their orthographies are
important in spelling, the learning-or acquisition-question is best addressed across different languages that have adapted
different writing systems. In addition to the work done in the English-speaking world on spelling, research in other languages is
now available to permit a broader comparative examination. We can begin
< previous page page_xi next page >

< previous page page_xii next page >
Page xii
to ask about both the common principles and specific variations of spelling across different writing systems and different
languages.
Especially interesting from a literacy point of view is a third perspective, one that asks the extent to which spelling and reading
are related. Of course, many self-professed poor spellers appear to be perfectly fine readers. This observation has tended to
reinforce a separation between reading and spelling: The former is seen as a basic skill, the latter often as a special skill. One of
the major themes of this collection of research on spelling is that a sharp separation of spelling and reading is misleading.
Although there are distinct differences between spelling and reading, there are also important shared foundations in knowledge
of word forms and in knowledge of how speech is related to writing.
In collecting some of the interesting research on spelling, we have adopted each of these perspectives. Many of the chapters
themselves reflect more than one perspective, and the reader will find important observations about orthographies, the
relationship between spelling and reading, and issues of learning and teaching throughout the collection. To organize the
collection, however, we have arranged three broad sections that reflect these three perspectives to some degree.
Part I, Theoretical Foundations: Writing Systems, Psycholinguistics, and Neuropsychology, contains two chapters that deal with
writing systems and orthographies. Jean-Pierre Jaffr provides a historical linguistic approach that sharpens distinctions, often
blurred in the literature, among scripts, writing systems, and orthographies, while also arguing for the limitations of static
linguistic approaches. Charles Perfetti develops a psycholinguistic foundation for spelling, one that, like Jaffr's, requires
attention to the principles of writing systems and orthographies but further suggests a wider spoken language basis even for
nonalphabetic systems. A central feature of this chapter is the argument that spelling and reading are related through shared
mental representations that bond orthographic with phonological information. Complementing these chapters on linguistic and
cognitive foundations of spelling, Pascal Zesiger and Marie-Pierre de Partz examine the cognitive neuropsychological
foundations. The neuropsychological approach emphasizes the potential for highly differentiated processes, with presumptive
functional neuroanatomy, that can contribute to spelling. The authors summarize some of the models of spelling and the
research they have prompted, noting both the progress and the unanswered questions.
Part II, The Acquisition of Spelling, contains six chapters reporting research on how children learn to spell. Here, the issue of
language and writing systems is represented by chapters that refer to research in English, French, German, and Dutch. (Chapter
11 in Part III refers to Hebrew.) Rebecca Treiman and Marie Cassar review existing theories of spelling development and
present research suggesting that, although sound is important in children's
< previous page page_xii next page >

< previous page page_xiii next page >
Page xiii
spelling, morphological information is also important. The early use of different kinds of knowledge, they argue, reveals that
stage models of spelling development are too simple. Heinz Wimmer and Karin Landerl present research on Austrian children's
spelling in German, with some interesting comparisons with English. The differences between English and German spelling
problems, they suggest, reflect differences in the two orthographic systems and perhaps in the instructional programs, which in
Austria are typically more code-based than in English-speaking countries. Providing data in a different language, Corinne
Totereau, Marie-Genevive Thevenin, and Michel Fayol examine children's spelling of plurals in French. In this examination,
they exploit a feature of French that is not present in either English or German, namely, the failure for both noun plurals and
third person verb plurals to be phonetically distinguished in speech, despite being marked in spelling. Their study exposes
morphological processes in spelling and suggests that these particular morphological processes are acquired in phases.
French language research is continued in the chapter by Jesus Alegria and Philippe Mousty, who discuss their studies of reading
disabled children, exploiting inconsistencies in how certain phonemes are spelled. From their findings, they suggest that reading
disabled children can use partial cues for reading that, however, are insufficient for complete representations of words. Also
reporting French language research, Linda Allal focuses on learning to spell in the classroom. This chapter is especially
important in providing a clear focus on instruction within a conceptual framework. Allal distinguishes broad categories of
instructional approaches, those that treat spelling as a specific school subject and those that treat spelling in an integrated literacy
context. Reviewing research in both English and French, she concludes that some coordination of these two approaches, each
having different potential benefits, is useful. In the final chapter of Part II, Terezinha Nunes, Peter Bryant, and Miriam Bindman
present a study of the role of morphology in the spelling of English children, focusing on the -ed past tense marker. They argue
that there is a developmental sequence driven by sensitivity to spoken language, with phonological awareness driving the earliest
stages and grammatical awareness driving the final stage. In between, the child uses morphological patterns such as -ed as units
but without grammatical awareness.
Part III, The Relationship Between Spelling and Reading, contains eight chapters. Again, there is a theme of learning to spell,
but chapters in this section more explicitly tie the acquisition of spelling to the acquisition of reading. Anna Bosman and Guy
Van Orden present a theoretical account of why spelling is more difficult than reading, drawing on studies of Dutch children and
on the dynamic systems framework for word perception (Stone & Van Orden, 1994). Their theoretical account captures the
intimate connections
< previous page page_xiii next page >

< previous page page_xiv next page >
Page xiv
between orthographic and phonological forms. One of the interesting implications is instructional: Contrary to simple pattern
learning ideas, reading may not be the most effective way to learn to spell, a skill that benefits from specific spelling strategies.
This chapter is also important in providing a theoretical account congenial to other chapters (Perfetti, chapter 2; Berent and
Frost, chapter 11). Iris Berent and Ram Frost provide a theoretical account of spelling and reading in Hebrew, which omits
vowel marks in most written texts, creating multiple word candidates for a given written string of consonants. They propose a
model of spelling in which both the assembly of letters from sound-letter connections and retrieval of whole words play a role;
knowledge about specific word forms suppresses erroneous letters that, in other words, would be correct spellings.
Jean Emile Gombert, Peter Bryant, and Nicola Warrick examine the role of analogy in spelling and reading, arguing that in both
cases analogy permits the beginning learner to go beyond explicit knowledge. They review studies in both French and English in
making the case for a central role for analogies, first in reading and then slightly later in spelling. Linnea Ehri presents a
comprehensive discussion of reading and spelling. Reviewing research on children's spelling and reading, she argues that
spelling and reading rely on nearly identical representations but with some differences in the processes. Both depend on specific
word knowledge, but the information adequate for reading is not always adequate for spelling.
Nick Ellis develops an account that uses Frith's influential 1985 developmental model of spelling and reading as a point of
departure and reviews the longitudinal studies that might shed light on what paces developmental changes. Ellis points out that
development is a matter of how orthographic and phonological representations change over time and concludes that, although
broad logographic stages characterize early development, orthographic and alphabetic influences are also present from the
earliest time. Laurence Rieben and Madelon Saada-Robert address the relationship between spelling and reading in learners at
the beginning of instruction. Using a longitudinal classroom study of students' copying, the authors conclude that the notion of
stage is contradicted by the longitudinal data. Emphasizing variability among learners and flexibility within learners, the authors
suggest that development trends show phases of dominance among spelling and reading strategies rather than stages.
Philip Seymour's chapter proposes a theory of orthographic and morphological development, in which spelling and reading
develop around a framework that encodes abstract properties of written language. The framework changes with development,
drawing on logographic and alphabetic foundations in interaction with linguistic awareness. In the final chapter of Part III,
Liliane Sprenger-Charolles, Linda Siegel, and Danielle Bchennec summarize their longitudinal studies of French children,
which suggest interesting
< previous page page_xiv next page >

< previous page page_xv next page >
Page xv
differences between spelling and reading acquisition. The earliest phase of spelling appears to rely more on sublexical
phonological connections than the earliest phase of reading, for example. At the same time, however, their study observed that
reading and spelling developed together, both relying on phonological processes and both contributing to the establishment of an
orthographic lexicon.
There are some additional observations to make about the collection of chapters. One is that the chapters are brief. The editors
instructed the authors to write brief stories about their work on spelling, saving full research reports for journals. We hoped to
have a slim volume that would be inviting rather than intimidating to students and researchers and that would be of some interest
to instructional practice. (Although we think the volume contains some of the best work in spelling, keeping the volume thin
meant that some very good work on spelling, unfortunately, is not represented.) Another point to note is that a French language
collection of these chapters, tentatively titled Des Orthographes et leur acquisition, edited by Laurence Rieben, Michel Fayol,
and Charles Perfetti, is being published. We hope that, in addition to illustrating how issues of spelling and reading require the
perspective of languages other than English, the publication of the same chapters in two languages will increase the prospects
for international cooperation as well as cross-linguistic comparisons.
Acknowledgments
The editors wish to express their appreciation to their respective institutions in Pittsburgh (The Learning Research and
Development Center and the Department of Psychology), Geneva (the University of Geneva), and Dijon (the University of
Bourgogne) for their support in this project. For very important help in tracking and organizing the contributions during the
early phases, we are grateful to Kathy Rud of the Department of Psychology of the University of Pittsburgh. To Mara Georgi, of
the Learning Research and Development Center, we express our appreciation for contributions to all phases of the project,
especially in editing chapter drafts, proofing final chapters, and preparing the index. So critical were her efforts on behalf of this
volume, that we think of her as the fourth editor.
CHARLES A. PERFETTI
LAURENCE RIEBEN
MICHEL FAYOL
< previous page page_xv next page >

< previous page page_1 next page >
Page 1
PART I
THEORETICAL FOUNDATIONS: WRITING SYSTEMS, PSYCHOLINGUISTICS, AND NEUROPSYCHOLOGY


Page 3
Chapter 1
from Writing to Orthography: The Functions and Limits of the Notion of System
Jean-Pierre Jaffr
CNRS-HESO, Paris
Written language, even today, is regarded as a poor relation in linguistics. The importance of written language in other fields of
research, however, adds irony to this situation. The reasons for it are to be found in the history of linguistic ideas over the last
few decades.
In France, as in many other countries, the ''Golden Age'' of linguistics coincided with the reign of structuralism that, following
the Saussurian model, primarily concerned itself with spoken language, relegating writing to a secondary level. Saussure's
(1972) now famous axiom, "The linguistic object is not both the written and the spoken forms of words; the spoken forms alone
constitute the object" (p. 45)1 is echoed by that of Bloomfield (1970): "Writing is not language, but merely a way of recording
language by means of visible marks" (p. 25).
Linguistics and Written Language

The Saussurian tradition, in introducing a dichotomy between synchronic and diachronic linguistics, established a useful
separation between spoken and written language. This helped to clarify a situation that, with traditional philology, had long been
muddled. However, by placing phonological theory
1 "L'objet linguistique n'est pas dfini par la combinaison du mot crit et du mot parl; ce dernier constitue lui seul cet
objet" (Saussure, 1972).

Page 4
at the center of linguistics, the Saussurian tradition effectively excluded written language (without which this linguistic theory
would, paradoxically, never have existed at all). Various claims were made in the name of this theory, including the claim that
writing should ideally be phonological. This gave rise, in the 1960s, to a certain number of ideological offshoots what Harris
(1986) termed "the tyranny of the alphabet" (p. 29).
In some linguistic schools, however, written language continued to survive in the glossematics of Hjelmslev (1966, 1985) and
Uldall (1944) and in the functionalism of Vachek (1973). Present-day research into written language owes much to the work of
these precursors and to their efforts to raise written language to the same status as spoken language. For the members of the
Glossematic school, written and spoken language exist alongside one another as expressions of one and the same language.
Hjelmslev (1985) thus explained that both speech and written language as well as many other sign systems have a semiotic
function, and he analyzed graphemes in the same way as phonemes. Uldall qualified this position, considering written and
spoken language as coexistent but asymmetrical.
Vachek (1973) went a step farther, underlining the complementary nature of what he termed the spoken norm and the written
norm. The impact of Vachek's thinking has been such that, today, scholars of quite different tendencies claim to be his
successors. Some have taken his concept of separate spoken and written norms as the basis for an autonomous view of writing
(Anis, Chiss, & Puech, 1988); others take a less radical stance (Jaffr, 1993). Vachek (1987) pointed out that although the
existence of a written norm may imply a certain amount of functional autonomy, writing can never be completely autonomous
from speech. He even added that it is essential that there should be links between written and spoken language, for if the gap
between the two norms became too wide, orthographic reform is necessary.
Although this position is by no means unanimously accepted today, many linguists now see the need for a linguistics of writing,
and they acknowledge that written and spoken language are two different yet equal types of expression. The precise nature of
the relation between these two linguistic media remains a matter for debate, and the frequently used notion of asymmetry only
goes some of the way toward answering the question. We can even find differences in the terminology used by various linguists:
Vachek (1973), for example, referred to norms; others referred to systems (Robins, 1973) and still others to languages (Lyons,
1970).
To refer to written and spoken norms, as Vachek (1973) did, seems to imply the existence of a single language system that sits
at a higher level. Vachek did not, however, acknowledge a theoretical construction of this kind. He merely pointed out that there
are structural correspondences between written and spoken language and that these correspondences become orthography when
they go from speech to writing and pronunciation

Page 5
when they go from writing to speech. However, in a somewhat obscure way, he also assumed that the written norm functions as
a superstructure with regard to its oral counterpart. Robins, on the other hand, stated quite clearly that speech and writing both
belong to the same language (1973). Catach (1988, 1995) went further, explaining the asymmetry between speech and writing
by introducing the concept of a prime language, a kind of high-level language (langue suprieure) that has been modified and
enriched by writing.
The Mixed Nature of Writing
Although this introduction explains how writing has gradually come to be accepted in linguistics, it has not given any indication
of how writing functions. There are a number of studies on this subject that emphasize the mixed nature of writing. Champollion
was, in this respect, something of a forerunner when, in 1822, he wrote the following in his Letter to M. Dacier. "I therefore
believe, sir, that phonetic writing existed in Egypt at a very early date, and that it was at first a necessary component of
ideographic writing" (pp. 41-42).2 This awareness of the coexistence of different principles within the same writing system is
very modern, but it has taken some time for it to become widespread. Until quite recently, writing systems tended to be
classified according to one or another principle (Gelb, 1973; Sampson, 1985). Robins (1973), who noted that all writing systems
are far from being pure specimens and that mixed systems are to be found, nonetheless considered mixed systems to be the
exception rather than the rule.
This idea of mixed systems has, in fact, come to be accepted by linguists because of their inability to describe written language
in the same way as spoken language. As long as the classification of writing systems forces each system to be in one category,
there is no way to recognize the multiplicity of linguistic stuctures honored to various degrees across systems. One has to
consider writing as nothing more than a historical accident, a secondary system that got out of hand, or else one must restrict the
role of the phonographic principle by setting it within a broader conception of writing, one that encompasses other principles
and involves other linguistic units (morphological and lexical). I prefer the second option, but, before developing this idea
further, I first explain the terminology used here.
Definitions and Terminology
Any theoretical discussion of written language must give a definition of the object being discussed, especially because the notion
of writing can cover
2 "Je pense donc, monsieur, que l'criture phontique exista en Egypte une poque fort recule; qu'elle tait d'abord une
partie ncessaire de l'criture idographique."

Page 6
a multitude of different things. Here, I focus primarily on the concepts of writing and orthography.
Writing. Although linguists' definitions of writing refer to language or a particular language, they may differ substantially
according to the theoretical conceptions on which they are based. I do not intend to draw an exhaustive list here, but it is
interesting to show the main steps in the gradation through a few examples.
For historians, writing is a process that is entirely subservient to language, without it ever being clear whether language refers to
spoken language or to something more abstract. For Cohen (1958), writing is "a visual and durable representation of language,
which enables it to be transmitted and preserved" (p. 1),3 whereas for Fvrier (1959/1984), it is "a device used to immobilize
and to fix articulated language, which is inherently volatile" (p. 9).4
This concept of writing, defined in very general terms, contrasts with that of Vachek (1939), who distinguished between writing
and written language. Whereas writing is defined as "the inventory of graphical means which can be used in putting down
written utterances" (p. 104), written language is "a system of graphical means which are accepted as a norm by the members of a
given linguistic community" (p. 104). Here, the idea of writing in a general sense is distinguished from its use within a particular
society.
Another step forward was made when the notions of internal structure and function were brought in, and the object of study was
referred to, in the plural, as scripts. This latter point is of considerable importance. Catach (1988) pointed out that "Scripts are
sets of discrete, articulated and arbitrary signs, which enable any constructed message to be transmitted without necessarily
using natural means" (p. 243).5
Orthography. What is the difference between writing and orthography? Vachek (1939/1989) used the term orthography to refer
to the necessary relations between the spoken norm and the written norm. However, this definition is exceptional: Elsewhere, the
term orthography, or spelling, is used to refer to a particular state of writing, without it always being clear what this state is.
Cohen (1958) used both terms, without giving a clear definition of either. He wrote at one point of the "orthographies of
alphabetical writing" (p. 234), suggesting that an orthography is an actualized form of writing.
3 "une reprsentation visuelle et durable de la langue, qui le rend transportable et conservable."
4 "un procd dont on se sert pour immobiliser, pour fixer le langage articul, fugitif dans son essence mme."
5 "Les critures sont des ensembles de signes discrets, articuls et arbitraires, permettant de communiquer n'importe quel
message construit sans ncessairement passer par la voie naturelle."

Page 7
However, further on, he wrote of Arabic writing and Arabic orthography as though they were the same thing.
It is interesting to compare studies of the history of writing (Cohen, 1958; Fvrier, 1959/1984; Gaur, 1984) with those of the
history of orthography and, particularly, of French orthography (Baddeley, 1993; Catach, 1968; Pasques, 1992). One may
wonder why hieroglyphs from the time of the Pharaohs (around 1500 BC) should be referred to as writing, whereas written
forms of the French 17th century are labeled as spellings. Is this difference in terminology due to a difference in the time period?
Writing, when it is not a theoretical concept, seems to refer to the means of expression of cultures from a distant past, whereas
orthography refers to more contemporary practices. A comparison of societies with a written tradition, however, suggests the
historical timeline is relative. In Mesopotamia, as in Egypt, writing lasted for at least 3,000 years, whereas the foundations of so-
called French orthography were laid no more than 1,000 years ago, which makes it a relatively new construction.
This brief overview shows that dealing with writing and its relationship with orthography first requires a statement that clearly
explains what certain terms and concepts are meant to cover. The best study from this point of view is that of Coulmas (1989),
who distinguished among writing systems, scripts, and orthographies and used each concept to help define the previous one.
Thus, when Coulmas talked about a writing system, he referred to the linguistic units that are represented (word writing, syllable
writing, phoneme writing, etc.), whereas a script refers to an assimilation of a given writing system with the language using it
(Chinese, Greek script, etc.). Finally, orthography refers to a standardization that is characteristic of a specific language
(German, French, etc.).
The main advantage of this terminology is that it allows us to describe writing as a general semiotic object as well as a linguistic
object whenever it interacts with a particular language. From now on, our analysis is organized along the lines of this tripartite
conception, going from the general to the particular. Accordingly, as a conclusion to this first section, I propose a set of
definitions that should help avoid certain ambiguities and should lead to a better understanding of what writing and orthography
have in common and what is proper to each term (see Table 1.1). The three concepts, taken as a whole, make up what I term
writing, or written notation.
From Writing to Scripts

For the linguist, writing is defined primarily by its relation to language first, in its spoken form, and then, gradually, as a broader
concept, whether in the sense of a social entity (as in Vachek's, 1939, linguistic community) or as

Page 8
TABLE 1.1
Writing, Script, and Orthography: Levels and Definitions
Concepts Levels Definitions
Writing Principles
Potential system of notation for basic linguistic
structures
(phonemes, syllables, morphemes, words)
Script
Principles Actual system of notation for basic linguistic
in structures
operation by means of visible marks
Usage
Orthography Normalization of an actual system of notation
Spelling for basic
linguistic structures by means of visible marks
a more abstract notion such as Catach's (1988) prime language. Only glottographic traces, unlike semasiographic traces, can be
defined as writing. The former are visible representations of linguistic utterances, whereas the latter represent ideas that may be
oralized, but cannot be translated word for word (Sampson, 1985).
Writing, therefore, is defined by its ability to represent linguistic units (phonemes, syllables, words, etc.), and typologies of
writing (or writing systems) have been made along these lines. Hence, we have logographic, syllabic, or alphabetic writing
(Gelb, 1973; Pulgram, 1976; Sampson, 1985), or even morphemic or phonemic writing (Hill, 1967). A script is defined by the
special relationship that it has either with the basic spoken units or with the meaningful elements of a language and, in
particular, with its morphology.
The fact that writing represents different kinds of units from language has often led to a misunderstanding of mixed systems.
Thus, Scholfield (1994) defined mixed systems as "more than one writing system used at once" (p. 56) and gave Japanese,
which uses both kanji (the logographic system) and kana (the syllabic system), as examples. However, DeFrancis (1989)
distinguished two ways of transmitting meaning using a single writing system: by symbols that represent sounds (and, therefore,
function as oral substitutes) and by symbols that supply information of a nonphonemic type. He insisted, rightly, that all scripts
result from the combination of these two tendencies, in varying proportions, and he called this the "Duality Principle." Thus, the
poorer a system is, from a phonemic point of view, the more it must compensate on a nonphonemic level. However, DeFrancis
retained a distinction between pure and mixed systems and, therefore, remains very strongly attached to the correspondence of
writing with spoken language.
Catach (1988) made better use of the concept of mixed systems by introducing the idea of "Protean" scripts.6 She set out a
range of modes of coexistence between nonmeaningful and meaningful units, ranging from virtual transparency (e.g., Finnish) to
virtual bilingualism (e.g., Chinese),
6 Proteus was a mythological god who often changed his appearance. So, the word Protean points out that writing can
take different forms.

Page 9
with a number of intermediate solutions in between, according to the role of the respective linguistic units. The phonographic
relation may be relatively regular as in Italian and German, or irregular, leaving more scope to morphology as in French and
English.
Thus, to return to the theoretical model I develop here, writing, at the most abstract, conceptual level, combines two principles:
phonographic and semiographic. The phonographic principle corresponds, more or less, to Vachek's (1973) orthography and is
manifested by correspondences between meaningless units of spoken language (phonemes or syllables) and meaningless units of
written language (phonograms or syllabograms). The semiographic principle encompasses the units and their functions in the
linguistic elements of written language. These units are determined by the morphological structure of the languages in question,
whether there are links with the spoken language (morphonograms) or not (morphograms), and by the way in which the written
words are assembled.
The Phonographic Principle in Writing
The phonographic notation is inseparable from the notion of writing, at least insofar as we are analyzing it from a linguistic
viewpoint. One could discuss how it came into existence and imagine a genetic hypothesis of phonographic units appearing first
randomly, then gradually becoming more systematic. On this point, most scholars agree that phonographic notation appeared at
an early stage but that it remained of secondary importance, as though it had appeared by accident (Bottro, 1987; Coulmas,
1989; Gelb, 1973). This line of thought often underpins an evolutionary hypothesis, which sees the history of writing as a
progression towards increasingly phonographic scripts whose crowning glory was the Greek alphabet (Havelock, 1981).
This evolutionary hypothesis has been widely, and rightly, criticized. However, it does not follow from such criticism that
scripts cannot be compared or treated as interrelated parts of one and the same phenomenon. In recent years, the analysis of
scripts as completely closed systems has gained increasing acceptance, no doubt in reaction to the evolutionary hypothesis.
Michalowski (1994) claimed that, around the end of the 4th millenium BC, Sumerian writing was already a system whose main
principles were to change very little from then on. A similar remark is made by Catach (1992) who said, while commenting on
the Palette of Narmer (4000 BC), that "the whole system is already there" (p. 5).
If the system is already there, it is in a very particular form and its phonographic dimension functions in a very specific way.
Our skepticism on this point is not due to the confusion to which this particular form of writing can give rise (and which, in the
case of Egyptian hieroglyphs, had long misled its decryptors) but rather to the apparently incidental nature

Page 10
of the phonographic notation represented here. The same thing applies to the Mesopotamian scripts that, according to
Michalowski (1994), comprise phonographic elements in such a way that it is impossible to see them as a later development. We
feel that it is unwise to make sweeping statements of this kind; otherwise, phonographic tyranny could well replace the
alphabetical tyranny formerly denounced by Harris (1986).
If the system is already there, does this mean that writing is immanent and unchanging? As Michalowski (1994) pointed out, the
functions that writing are supposed to fulfill and, accordingly, the relation to language that these functions imply, change over
time. These functions can have quite different relations with speech. For example, a system that notated language in an
approximate way was quite adequate for the economic needs of the Bronze Age Greeks in Crete. However, the private and
literary needs of Greeks during the Classical Age required a much more precise notation of spoken language, particularly of
vowels (Powell, 1991).
The Phonographic Enigma. The quest for a phonographic notation that adheres rigidly to the spoken language is illusory. If
writing, from the outset, had recourse to phonographic notations (even indirect ones such as rebuses), it was because writing
needed a practical means of recording language, not sounds. The aim of writing is not simply to record phonemes, and to do so
is only useful insofar as it can transmit messages, regardless of how accessible the forms it produces may be. What it comes
down to is finding a conventional means of representing linguistic content. That is why any phonographic notation is, above all,
useful as a recurrent infrastructure, and it must therefore be relatively economical.
The notion of economy, in linguistics, may have a number of meanings. I use it here mainly in a structural sense: as few units as
possible.7 These units form a bridge with the spoken language, allowing a kind of cloning process to take place, as recent
studies on reading acquisition have shown.8 Mastering this process is, however, only one of the conditions necessary for
learning to read and write, which also comprises an infrastructural dimension in which phonograms, or syllabograms, merge into
a bigger picture to make up larger linguistic units with a lexical or grammatical function.
The phonographic dimension is, therefore, present in all scripts. However, scripts cannot be analyzed on this basis alone. As we
explained earlier,
7 Another sense of economy is "linguistic economy" that concerns the distribution of the constituent elements of a system.
Structural economy, the sense used here, is not necessarily a reliable indicator of the degree of efficiency of linguistic
economy because the functionality of a script cannot be measured by the number of nonsignificant units that make up its
structural base.
8 Studies on learning to read are numerous. For a representative set, see Rieben and Perfetti, 1991.

Page 11
with the principle of mixed systems, the units that belong to this dimension are conditioned by linguistic units at a higher level.
DeFrancis (1989) saw this as sufficient grounds to distinguish between what he called pure systems on the one hand and
meaning-plus-sound systems on the other. He included Finnish (an alphabetical script) and the Japanese kana (a syllabic script)
among the pure systems, whereas in the second category, he put English (an alphabetical script) and Chinese (a syllabic script).
A similar idea is expressed in the "Four Solutions" of Catach (1988).
To describe the accessibility or the degree of transparency of the phonographic option, language psychologists often use the
terms shallow orthography and deep orthography. According to the definitions used here, orthography refers to script because
the difference between the shallow and the deep hinges on the transparency of the phonographic units. The closer the number of
phonograms is to the number of phonemes, the closer the script is to being completely transparent. Conversely, when there is a
considerable difference between the number of phonemes and the number of phonograms (the latter always being greater), the
phonographic link becomes complex, and it is difficult to access the script by a phonographic approach alone. This is the case in
French, which has about 30 phonemes and more than 130 phonograms (Catach, 1980). This is also the case in Chinese, in which
the syllabic notation is far from reliable. However, studies on these points are often contradictory: DeFrancis (1984), for
example, analyzed 895 phonetic keys in Chinese and concluded that 66% of the cases give a reliable indication of pronunciation.
Sampson (1985), however, explained that these keys, whether phonetic or semantic, only provide clues, which are then subject
to interpretation, and he concluded that in Chinese, pronunciation is very difficult to predict.
The Semiographic Principle in Writing
The semiographic principle, when applied to writing, is probably even more heterogeneous than the phonographic principle.
Most typologies of writing systems, whether old or new, generally use the term logographic (Sampson, 1985), which itself
replaced such inaccurate terms as pictographic and ideographic (DeFrancis, 1984). However, the term logographic, or writing in
words, is inadequate to describe the linguistic complexity of written forms that very often contain several meaningful units, or
morphemes.
Mediation of Meaning. Each script is conditioned by its raison d'tre: the mediation of linguistic meaning. In performing this
semiologic achievement, the script puts on stage linguistic signs that, de facto, acquire some phonographic autonomy. This
graphic production of language depends not only on some internal factors such as the morphologic structure but also on external
factors. Thus, the needs of Aegean society were not the same

Page 12
as those of Sumerian society, and both are equally remote from the needs of our present-day society. In Minoan Crete, the tables
inscribed with the Linear B script had an economic function and were adapted to a relatively rudimentary type of written
mediation: that of making lists (Godart, 1990). On the other hand, Sumerian culture, which valued literary expression at a very
early date, had to invent more complex linguistic forms in order to transcribe mythological works such as the Epic of Gilgamesh
(Bottro, 1994).
The requirements of even the most ancient scripts go beyond the simple notation of concepts, giving rise to a more flexible
notation that eventually established itself as a written norm. These developments presuppose an analysis of the source language
and a sense of linguistic community by the speakers of the language. However, the result is not just a simple copy of the spoken
language but a form of linguistic mediation created for the needs of a society with particular cultural demands in accordance
with a general semiology of writing.
The Semiographic Principle and Spacing. Leaving spaces between words is so much a part of our everyday written practice that
it is hard to imagine that it could ever have been otherwise. In fact, these spaces can be seen as the negative trace of the
semiographic dimension inherent to writing. A quick review of the different marks used at different periods to separate words
shows, however, that the use of spaces is only one of the many semiographic means available.
Whereas spacing is a kind of negative mark, other types of marks can be termed positive. These include the median point that
the Romans used to place midway between the upper and lower parts of letters (Wingo, 1972), or the wedge widely used in Old
Persian (Doblhofer, 1959). We can also include here the subdivisions used in Sumerian tablets to segregate words or sentences
in lines or in boxes (Andr-Leickman, 1982). In the earliest Semitic scripts, a short stroke was used for the same purpose
(Naveh, 1973). These different examples are not an exhaustive list, but they confirm that a marking process is inherent to
writing, although it has not always been used in a systematic way. The meaningful segments isolated in this way range from
linguistic elements to larger units made up of several morphemes.
The division of units can be made in even more elaborate ways, according to the internal functioning of the script in question.
One of the most recent examples is found in Japanese, in which the general tendency is to intersperse the kana and kanji scripts
according to function, with kanji used for lexical functions and kana used for grammatical functions. The identification of
meaningful units is helped along by objective associations whose functional status is reflected, at least in part, in the script itself
(Coulmas, 1989).
In ancient scripts, this role is more usually played by determiners. In Mesopotamian cuneiform, the polyvalence of written signs
led to the ap-

Page 13
pearance of determiners that were placed at the beginning or at the end of words. Determiners were not pronounced but
indicated the category to which the concept belonged (e.g., a god, a man, a woman, a bird, and so on) (Andr-Leickman, 1982).
A similar device was used in Egyptian hieroglyphs, in which certain signs were used alongside others to indicate the category
(Ziegler, 1982). These determiners acted, although in an indirect and irregular way, as semiographic clues.
Thus, the semiographic principle can take on different forms from one script to another, according to the linguistic structures of
the language, and, more generally, to the gradual development of writing practices. There is little point in speculating on
whether one type of device is better than another because all scripts must be considered in their own right and in their own
context. The study of morphogenesis does not, however, reveal a regular and progressive emergence of the practice of spacing.
This was emphasized by Naveh (1973) on the subject of Semitic scripts. He observed that, by around the 7th century BC, each
script had developed its own method. Thus, the more conservative Hebrew script tended to separate words by using dots,
whereas Aramaic was more inclined to introduce word spaces, and Phoenician preferred the system of scriptio continua, which
it passed on to the Greeks.
Morphology and The Dynamics of The Semiographic Principle. In modern scripts, division into words is the most obvious mark
of the semiographic function. The written word seems to be, from this point of view, the most widespread notion among the
linguistic community and, at the same time, the hardest to define in linguistic terms (Gruaz, 1987). However, this is due to the
fact that semiography, as we understand it here, includes morphological features that are subject to quite different laws.
Linguistics has long classified languages according to a number of predominant typological principles. Thus, the morphological
typology introduced in the 19th century by the Schlegel brothers and perfected by von Humboldt has given rise to a
classification of languages: analytic/isolating, agglutinative, inflectional, and polysynthetic. This typology has been criticized for
various reasons, mainly because it categorizes languages on morphological grounds alone and that these categories are mutually
exclusive (i.e., a language must be analytic or agglutinative or inflectional).
More recent attempts by linguists to define language typologies take into account more diverse factors, including linguistic
functions (Croft, 1990). Generally speaking, they consider languages as a combination of structures and functions that can be
described in terms of predominant features. From a morphological point of view, languages may therefore be predominantly
analytic, agglutinative, and so on (Hagge, 1985). A linguistic typology based on a morphology is useful if we want to examine
the relation between

Page 14
languages and scripts. To give two examples: First, a predominantly analytic language such as Chinese favors isolated graphic
representations. This explains why Chinese writing has sometimes been described as morphosyllabic (DeFrancis, 1984). Second,
an agglutinative language such as Turkish increases the interdependence of meaningful units, and we may, therefore, expect the
semiographic principle to bear on more complex meaningful units (apan, 1989).
Words, sequences of letters between two spaces, are meaningful units with variable geometry, structured on a morphological
basis. If we want to describe these variations by using existent linguistic typologies, we can say that there are two semiographic
options, which we term the morphographic option and the logographic option (Jaffr, 1994). The morphographic option is
found wherever each graphic unit corresponds to one, and one only, linguistic sign. This major influence of the analytic type,
which is predominant in Chinese, is also found in most scripts throughout the world, most notably in cases where words are
morphologically invariable. This is the case for words such as ''pour,'' "sur," "dans," and so on in French; "at," "from," "for,"
and so forth in English; "donde," "para," "con," and so on in Spanish.
Inflectional and agglutinative languages, on the other hand, favor an option that tends to assemble morphemes within a single
written sequence. Thus, in Turkish, written words can be quite long, containing several meaningful units. To simplify, we can
say that these languages function by placing suffixal elements to the right of a root, in a predictable order, even if these elements
are not realized (Kornfilt, 1987). In written Japanese, as in Chinese, morphemes are juxtaposed, but the intermixing of kanji and
kana makes for easier identification of the logographic blocks produced (Shibatani, 1987).
The inflectional type of language generates a variant of the logographic option that seems to be more complex than the types
described previously because it modifies the form of the meaningful units. Although it is based, as is the agglutinative type, on
the effective association of a root and affixes, it differs from the relatively regular distribution found in Japanese and Turkish.
We find frequent and systematic examples of this in Semitic scripts (e.g., Hebrew and Arabic) where the lexical part of words
(the root) and the grammatical part (the inflection) are amalgamated (Berman, 1985). We can also find this in French, in the
conjugated forms of irregular verbs such as "faire," "vouloir," "venir," and so on, where the boundary between the root and the
inflection is so unclear that linguists, in order to account for it, have had to create the notion of thematic stem (e.g., for the "eu"
in "je veux").
Thus, we can see that the ways in which scripts work are relatively simple as long as we are concerned only with principles.
However, when these principles are set into operation, various factors are brought into play that complicate the situation
considerably. Although it is still possible and necessary to adopt a systematic approach to writing, we are well aware that

Page 15
scripts are never fully predictable, mathematical means of notation. They bear the marks of their eventful development, full of
accidents and exceptions. In addition to this, we must bear in mind that scripts are only very rarely original creations.
Michalowski (1994) listed only four societies (the "Gang of Four") that created their own scripts: Sumerian, Egyptian, Chinese,
and Central American. In most cases, scripts were borrowed from an influential neighboring society. This borrowing favored the
importation of foreign linguistic features.
However, the history of writing shows that creations, borrowings, adaptations, and the apparently anarchic conflicts that resulted
from them, did not give rise to chaotic artifacts. Linguistic analyses show, on the contrary, that there is a kind of structural
balance behind which we can see the dual principles of semiographics and phonographics at work. This brings us back to a
typological analysis that tends to show that there is not an infinite number of possibilities but that everything oscillates between
syllables and phonemes on the one hand and morphemes and lexemes on the other. In spite of the differences between them, it is
as though scripts were the reflection of an archetypal human activity.
From Scripts to Orthography

Writing is not an exact science, but only the visible record of human knowledge. It reflects, at least to a certain extent, the
human capacity to think in an abstract way about one's own language (Veltman, 1992). Accordingly, we may try, as we have
just done, to grasp the recurrent, systematic part of writing as represented in the linguistic mediation of writing. However, such
as attempt can be only partly successful. There is always a part of scripts that cannot be contained within the boundaries of the
system. Because they are also cultural, institutional, and social objects, scripts can become conditioned by nonlinguistic factors.
Convention is reinforced by usage. Of course, the linguistic norm is not the same as the orthographical norm, but it is, so to
speak, a prerequisite for it and can help it become established. For this reason, the orthographical norm, a kind of supernorm,
has all the more chance of becoming implanted if the script it corresponds to is old and bears the weight of history. This point,
no doubt, deserves a far more detailed analysis than we can give here. We need to enlist the help of palaeography and epigraphy
to give us more information about the practices of scribes in ancient societies. However limited our information may be on this
subject, we can assume that scribal practices were particularly ritualized (Kramer, 1986). Scribes, as "professional writers," had
the power to introduce certain innovations; however, the history of writing shows only too well that the greatest innovations
took place when scripts were being borrowed, that is, at their earliest, formative stages (Naveh, 1987).

Page 16
Naveh (1973) and Saenger (1982) made quite similar observations concerning the segmentation of writing at different times and
in different societies. They explained that as far as writing is concerned, the most conservative groups are those whose mother
tongue is mediated by the script in question. Conversely, when an external user uses the same script, and the language is foreign
to him, he does not hesitate to take liberties with it and to introduce changes. Accordingly, Assyrian scribes seem to have been
the first to introduce spacing in Aramaic script, which they used for international purposes, at around 1000 BC (Naveh, 1973).
Similarly, Irish monks were apparently the first to reintroduce spacing in works written in Latin around the 7th century AD
(Saenger, 1982).
The establishment of an orthographical supernorm also depends on the requirements that writing is to fulfill. A supernorm has
more chance of being established when writing has a political or religious function, as was the case with Egyptian hieroglyphs
(Davies, 1987) or with Hebrew in the kingdoms of Judah and Israel (Naveh, 1987). This supernorm allows power to be
established and preserved. Conversely, when writing serves an economic purpose, it is more likely to be functional, and the
more widespread its commercial use, the more flexible it is likely to be. This is the case with Phoenician, which was used
throughout the Mediterranean from 1000 BC onward (Gras, Rouillard, & Teixidor, 1989).
These situations are obviously quite different from those of modern societies: Our orthographies are the result of an
unprecedented expansion in the use of writing. We must not forget that in ancient societies the use of writing was restricted to a
very small number of people: Even in Ancient Greece, it is likely that not more than 10% of men were able to write (Harris,
1989). However, having said that, we may say that all societies with a written tradition manage, each in its own way, to produce
an orthographical supernorm that makes writing an instrument of power and of social preference.
When faced with orthographies, linguistics can be concerned only with their systematic features (and those of their scripts). The
remainder is a topic for sociolinguistics, provided that they agree on what is meant by orthography. My emphasis on the
possible confusions that can arise between writing and orthography has been a reponse to a history of variant uses of these
terms, with accompanying misunderstandings. Consider that many authors have referred to orthography when discussing French
(Baddeley, 1993; Beaulieux, 1967; Catach, 1968; Pasques, 1992), English (Scragg, 1974), Catalan (Segarra, 1985), and Spanish
(Esteve Serrano, 1982) writing between the 13th and 17th centuries. However, because an orthographic norm was not truly
established until the 19th century (Guion, 1973), the use of the term orthography in reference to this earlier writing is
misleading. Although these studies do describe the foundations of what was to become the orthographical norm and how it was
formed, they show that writing is a human science, patterned

Page 17
by conflicts, disputes, and even "civil wars" between the partisans of a predominantly phonographic option and the partisans of a
predominantly semiographic option. The problems that they study are not, after all, very different from those raised by scholars
who study ancient scripts such as Greek (Powell, 1991) or Latin (Desbordes, 1990).
Conclusions
Linguistics describes facts and constructs systems that can give rise to working hypotheses in other fields, particularly that of
language acquisition. It is, moreover, increasingly likely that present studies on acquisition will lead us to modify some of our
linguistic hypotheses. Although the linguist's analyses are necessarily different from those of the psychologist, the theoretical
options each chooses can help both of them to see eye to eye on certain matters. The linguistics of writing has tended to favor
the methods of structural linguistics (and in particular the concept of synchrony) in order to be recognized as a discipline in its
own right. Because of this, there is now an unfortunate division between the theories of writing and the studies of the history of
writing, whether scripts or orthographies. It is now time to build another kind of linguistics, and this chapter has shown how this
may be possible (see Jaffr, 1995). All orthographies and all scripts are part of a bigger picture by which they are conditioned
and which can help us to gain a better understanding of how they function. If we want to construct this bigger picture, we must
first create a general theory of writing that can encompass all research carried out until now on writing systems, old and new.
We must also, however, consider each script in its own right as a unique merger of semiological processes with a particular
language. We must also bear in mind that languages themselves have general characteristics that can be described by linguistic
typologies.
Today, the linguistics of writing is at a turning point, for two reasons. First, its theoretical foundations are not quite solid
enough, as the often difficult debate over the relations between spoken and written language shows only too well. Second, too
few studies have as yet been carried out in this field, and those that have are very divergent, if at all necessary. It would be
useful, for example, to update existing works on general models of writing systems, such as that of Gelb (1973), in order to
construct new typologies that are more representative of the complex and mixed nature of scripts. We must also carry out more
detailed studies on contemporary scripts, with a different epistemological approach from the works on orthographies and their
formation. We also need a wider range of studies, so that large-scale comparisons can be made. The linguistics of writing needs,
above all, more studies on scripts and orthographies such as Arabic, Hebrew,

Page 18
Japanese, Chinese, and so on, which differ from the Western logic of the Greek alphabet and its successors.
The development of archaeology is, finally, giving rise to a promising increase in epigraphic and palaeographic data. However,
in spite of the importance of the findings from Mesopotamia, Egypt, Crete, the Near East, and Central America, they have as yet
only brought fragmentary theoretical contributions to the linguistics of writing. These findings could help to modify
considerably the all-too-static linguistic point of view.
After a long period dominated by structure and synchrony, the notion of the subject is now invading the whole of linguistics in
very diverse forms (historical, sociological, psychological). That is why it is not only useful but also indispensable to carry on
the dialogue with language psychologists and acquisition specialists.
References
Andr-Leickman, B. (1982). Les critures cuniformes [Cuneiform writings]. In Naissance de l'criture. Cuniformes et
hiroglyphes (pp. 73-114). Paris: Editions des Muses nationaux.
Anis, J., Chiss, J.L., & Puech, C. (1988). L'criture, thories et descriptions [Writing: Theories and descriptions]. Bruxelles: De
Boeck Universit.
Baddeley, S. (1993). L'orthographe franaise au temps de la Rforme [The French orthography at the Reform time]. Genve:
Droz.
Beaulieux, C. (1967). Histoire de l'orthographe franaise: Formation de l'orthographe, des origines au milieu du XVIme
sicle, 2 [The history of French orthography from the origin to the middle of the sixteenth c.]. Paris: Champion.
Berman, R.A. (1985). The acquisition of Hebrew. In D.I. Slobin (Ed.), The crosslinguistic study of language acquisition, Vol. 2:
The Data (pp. 255-372). Hillsdale, NJ: Lawrence Erlbaum Associates.
Bloomfield, L. (1970). Le langage [Language]. Paris: Payot.
Bottro, J. (1987). Msopotamie: L'criture, la raison et les dieux [Mesopotamia: Writing, reason and gods]. Paris: Gallimard.
Bottro, J. (1994). Babylone et la Bible: Entretiens avec H. Monsacr [Babylone and the Bible: Talks with]. Paris: Les Belles
Lettres.
apan, S. (1989). A linguistic study of reading and writing disorders in Turkish, an agglutinative language. In P.G. Aaron &
R.M. Joshi (Eds.), Reading and writing disorders in different orthographic systems (pp. 191-202). Dordrecht, The Netherlands:
Kluwer.
Catach, N. (1968). L'orthographe Franaise l'poque de la Renaissance [French orthography at the Renaissance age]. Genve,
Switzerland: Droz.
Catach, N. (1980). L'orthographe Franaise: Trait thorique et pratique [French orthography: Theoretical and practical
treatise]. Paris: Nathan.
Catach, N. (1988). L'criture en tant que plurisystme, ou thorie de L prime [Writing as a plurisystem or the theory of L']. In N.
Catach (Dir.), Pour une thorie de la langue crite (pp. 243-259). Paris: Editions du Centre National de la Recherche
Scientifique.
Catach, N. (1992). Les systmes d'criture [Writing systems]. Paris: Centre National de Documentation Pdagogique.
Catach, N. (1995). L'criture et la double articulation du langage [Writing and double articulation of language]. Linx, 31, 37-48.

Page 19
Champollion, J.-F. (1989). Lettre M. Dacier relative l'alphabet des hiroglyphes phontiques. Fata Morgana Reprint.
(Original work published 1822)
Cohen, M. (1958). La grande invention de l'criture et son volution [The great invention of writing and its evolution]. Paris:
Klincksieck.
Coulmas, F. (1989). The writing systems of the world. Oxford, UK: Basil Blackwell.
Croft, W. (1990). Typology and universals. Cambridge, UK: Cambridge University Press.
Davies, W.V. (1987). Reading the past: Egyptian hieroglyphs. London: British Museum Publications.
DeFrancis, J. (1984). The Chinese language: Fact and fantasy. Honolulu: University of Hawaii Press.
DeFrancis, J. (1989). Visible speech: The diverse oneness of writing systems. Honolulu: University of Hawaii Press.
Desbordes, F. (1990). Ides romaines sur l'criture [Roman ideas on writing]. Lille: Presses Universitaires de Lille.
Doblhofer, E. (1959). Le dchiffrement des critures [The decipherment of scripts]. Paris: Arthaud.
Esteve Serrano, A. (1982). Estudios de teora ortogrfica del espaol. Murca, Spain: Universidad de Murcia.
Fvrier, J. (1959/1984). Histoire de l'criture [The history of writing]. Paris: Payot.
Gaur, A. (1984). A history of writing. London: The British Library.
Gelb, I.J. (1973). Pour une thorie de l'criture [A study of writing]. Paris: Flammarion.
Godart, L. (1990). Le pouvoir de l'crit [The power of writing]. Paris: Editions Errance.
Gras, M., Rouillard, P., & Teixidor, J. (1989). L'univers phnicien [The phoenician universe]. Paris: Arthaud.
Gruaz, C. (1987). Le mot Franais cet inconnu: Prcis de morphographmologie [The unknown French word: Treatise of
morphographemology]. Rouen: Universit de Rouen.
Guion, J. (1973). L'institution orthographe [The orthographic institution]. Paris: Le Centurion.
Hagge, C. (1985). L'homme de paroles. Paris: Fayard.
Harris, R. (1986). The origin of writing. London: Duckworth.
Harris, W. (1989). Ancient literacy. Cambridge, MA: Harvard University Press.
Havelock, E.A. (1981). Aux origines de la civilisation crite en ccident [Origins of western literacy]. Paris: Maspro.
Hill, A.A. (1967). The typology of writing systems. In W.M. Austin (Ed.), Papers in linguistics in honor of Lon Dostert (pp.
92-99). The Hague: Mouton.
Hjelmslev, L. (1966). Le langage [The language]. Paris: Minuit.
Hjelmslev, L. (1985). Nouveaux essais. Paris: Presses Universitaires de France.
Jaffr, J.-P. (1993). La phonographie: sa gense et ses limites [Phonography: Genesis and limits]. In J.-P. Jaffr, L. Sprenger-
Charolles, & M. Fayol (Eds.), Lecture-criture: Acquisition (pp. 22-37). Paris: Nathan.
Jaffr, J.-P. (1994). Le principe smiographique et la morphologie des langues [Semiographic principle and the morphology of
languages]. Paris: Comit National Franais de Liason pour la Radaptation des Handicaps.
Jaffr, J.-P. (1995). Une approche gntique de l'criture: de l'invention l'acquisition [Genetic approach to writing: From
invention to acquisition]. Linx, 31, 49-64.
Kornfilt, J. (1987). Turkish and Turkish languages. In B. Comrie (Ed.), The major languages of Eastern Europe (pp. 227-252).
London: Routledge & Kegan Paul.
Kramer, S.N. (1986), L'histoire commence Sumer [History begins at Sumer]. Paris: Arthaud.
Lyons, J. (1970). Linguistique gnrale: Introduction la linguistique thorique [General linguistics: Introduction to the
linguistic theory]. Paris: Larousse.
Michalowski, P. (1994). Writing and literacy in early states: A Mesopotamianist perspective. In D. Keller-Cohen (Ed.), Literacy:
Interdisciplinary conversations (pp. 49-70). Cresskill, NJ: Hampton Press.
Naveh, J. (1973). Word division in West Semitic writing. Israel Exploration Journal, 23, 206-208.
Naveh, J. (1987). Early history of the alphabet. Jerusalem: The Hebrew University.

Page 20
Pasques, L. (1992). Les grands courants orthographiques au XVIIme s. et la formation de l'orthographe moderne [Major
orthographic trends in 18th c. and the birth of modern orthography]. Tbingen, Germany: Niemeyer.
Powell, B.B. (1991). Homer and the origin of the Greek alphabet. Cambridge, UK: Cambridge University Press.
Pulgram, E. (1976). The typologies of writing systems. In W. Haas (Ed.), Writing without letters (pp. 1-28). Manchester, UK:
Manchester University Press.
Rieben, L., & Perfetti, C.A. (Eds.). (1991). Learning to read: Basic research and its implications. Hillsdale, NJ: Lawrence
Erlbaum Associates.
Robins, R.H. (1973). Linguistique gnrale: Une introduction [General linguistics: An introduction]. Paris: Armand Colin.
Saenger, P. (1982). Silent reading: its impact on late medieval script and society, Viator, 13, 363-414.
Sampson, G. (1985). Writing systems. London: Hutchinson.
Saussure (de), F. (1972). Cours de linguistique gnrale. Paris: Payot.
Scholfield, P.J. (1994). Writing and spelling: the view from linguistics. In G.D.A. Brown & N.C. Ellis (Eds.), Handbook of
spelling: Theory, process and intervention (pp. 51-71). Chichester, UK: Wiley.
Scragg, D.G. (1974). A history of English spelling. Manchester, UK: Manchester University Press.
Segarra, M. (1985). Histria de l'ortografia catalana. Barcelona: Empries.
Shibatani, M. (1987). Japanese. In B. Comrie (Ed.), The major languages of East and South-East Asia (pp. 127-152). London:
Routledge & Kegan Paul.
Uldall, H.J. (1944). Speech and writing, Acta Linguistica, 5, 11-16.
Vachek, J. (1939/1989). On the problem of written language. In P. Luelsdorff (Ed.), Josef Vachek, written language revisited
(pp. 103-116). Amsterdam: Benjamins.
Vachek, J. (1973). Written language: General problems and problems of English. The Hague: Mouton.
Vachek, J. (1987). Written language seen from the functionalist angle. In R. Dirven & V. Fried (Eds.), Functionalism in
linguistics (pp. 395-405). Amsterdam: Benjamins.
Veltman, R. (1992). An orthography as a theory of language. In C.M. Sterling & C. Robson (Eds.), Psychology, spelling, and
education (pp. 18-29). Clevedon, Avon, UK: Multilingual Matters Ltd.
Wingo, E.O. (1972). Latin punctuation in the Classical Age. The Hague: Mouton.
Ziegler, C. (1982). Les critures gyptiennes [Egyptian scripts]. In Naissance de l'criture: Cuniformes et hiroglyphes (pp.
117-170). Paris: Editions des Muses nationaux.

Page 21
Chapter 2
The Psycholinguistics of Spelling and Reading
Charles A. Perfetti
University of Pittsburgh
One would expect that psycholinguistics, as the study of the psychological processing of linguistic structures, is a natural home
for the study of spelling. Instead, spelling has been rather neglected within mainstream psycholinguistics. The surge of research
in spelling since the 1980s, to be sure, was important in correcting this neglect, even if this research was only implicitly
psycholinguistic (Ehri, 1980; Frith, 1985; Henderson, 1981). Moreover, Treiman's (1993) work on spelling development shows
an explicitly psycholinguistic approach: It links the acquisition of spelling to linguistic structures by specific hypotheses about
linguistic units. Nevertheless, it is fair to say that there has been a relative neglect of the psycholinguistic components of
spelling. One finds little coverage of the topic in textbooks on psycholinguistics. More telling is the Handbook of
Psycholinguistics (Gernsbacher, 1994), which covers the basic topics of the field in 34 chapters, none of which include spelling.
Indeed, the reader searches in vain for an entry in the index under spelling.
The reasons for this relative neglect are manifold but include at least the scientific privilege, inherited from linguistics, given to
spoken language. (Ironically, however, written language has actually dominated the empirical studies of adult psycholinguistics.)
Spelling may be seen less as a scientific problem of language use than as a literacy convention or a school subject. A second
reason may be that the deceptive simplicity of spelling has masked its intriguing problems, making it seem less challenging than
the higher

Page 22
mental processes that have preoccupied the cognitive sciences. In any case, neglect is gradually being remedied by research that
addresses the processes and acquisition of spelling in relation to linguistic structures (Berent & Frost, chap. 11; Bosman & Van
Orden, chap. 10; Totereau, Thevenin, & Fayol, chap. 6; Treiman, 1993; Treiman & Cassar, chap. 4).
The aim of this chapter is to further contribute to the psycholinguistic study of spelling by delineating some of the relevant
linguistic structures involved in spelling and by showing how psycholinguistic processes involving these structures are central to
both spelling and reading. These processes and structures connect spelling with reading in a highly interdependent way. In
making this case, an essential first step is to demonstrate the importance of linguistic structures that are the foundation of
spelling: those that are manifest in writing systems and orthographies.
Writing Systems and Languages

Spelling is the encoding of linguistic forms into written forms. The linguistic units phonological strings, morphemes, and words
are provided by the spoken language. The writing units are provided by the writing system and its inventory of graphic devices.
Figure 2.1 illustrates the general form of the relationship between languages and writing systems. The linkages between the units
of the two systems define the essential meaning of reading and writing. Both require that graphic units be connected to linguistic
units.
Unlike the linguistic system, which is assumed to be systematic, multilayered, and abstract, writing is often treated as
unsystematic, strictly linear, and highly conventional. Spellings are often said to be arbitrary or conventionalized. The idea of
convention a social contract enforced by appeal to authority does capture an important part of formal spelling. More important
from a psycholinguistic perspective, however, is the fact that spelling also reflects the general principles of a writing system the
writing system design and the specific orthography that embodies the general design plus its distinctive orthographic features.
In Fig. 2.1, there is a distinction between the writing system and the orthography that is often ignored in psychological
treatments of reading and spelling. Generally speaking, the writing system dictates the level at which general linguistic
constraints are imposed on writing, whereas the orthography dictates the specific details of these constraints (see Jaffr, chap. 1,
for a further discussion of these distinctions). Accordingly, writing systems can be said to be logographic, syllabic, and
alphabetic, depending on the smallest linguistic unit encoded in the basic writing unit. Orthographies within writing systems
show minor but significant variation concerning the rules that relate graphic units to linguistic units. Thus, written

Page 23
Fig. 2.1.
The general relationship among writing systems, orthographies, and languages.
Languages provide multilevel units in phonology (phonemes, syllables, onsets,
rimes not represented and morphology). Writing system principles select
one or more levels of language for encoding into elementary writing units.
Orthography is the system of constraints on the writing units, including their
mapping to language units.
Chinese (logographic or morphophonological), Japanese kana (syllabic), and Italian (alphabetic) belong to different writing
systems. However, written English, Italian, and French (all alphabetic systems) can be said to have different orthographies.1
The study of spelling (and reading as well) has been dominated by studies of one kind of writing system, the alphabetic, and one
orthography, English. The effect of this one-orthography research has been to give a distinctive character to conclusions about
spelling. For example, one of the most influ-
1 Further distinctions are sometimes confused. Thus, alphabetic systems are not the same as alphabets: Russian, written in
the Cyrillic alphabet, and French and English, written in the Roman alphabet, are equally alphabetic, and the differences
in orthographies are no greater, in principle, across than within alphabets. Finally, the actual visual form of the written
units, as opposed to the rules that map them to language units, is a matter of script. Thus one can refer to the Cyrillic-
Roman contrast as a difference in script as well as the contrast between Hebrew and Arabic, which are both consonant-
based "incomplete" alphabetic systems. Korean is alphabetic, but its arrangement of graphic units (letters) into uniform
nonlinear spaces is a variation in script.

Page 24
ential accounts of spelling is Frith's (1980, 1985) dual-route model. As with dual-route models for reading (Besner, 1987;
Coltheart, 1978; Paap, Noel, & Johansen, 1992), such models seem necessary because of the unreliability of English
orthography. These models assume that there are two different mechanisms for spelling words and for reading them, one that
makes reference to the phoneme constituents of the word and one that does not. In reading English, the phonemic route works
fine for words such as safe, but it fails for cafe. Thus, there must be a mechanism for looking up words directly and bypassing
the phonology that would otherwise produce the wrong word or no word at all. In spelling, the route is reversed, from
phonology to orthography. Applying the phonological (or merely phonemic) route to /kefa */ may lead to a spelling something
like kufay. Even spelling /seyf/, however, could be in jeopardy along the phonemic route, producing, perhaps, saif as well as
safe.
The two-route approach solves this problem by allowing words (in skilled spelling) to be spelled by a mechanism that looks up
lexical spellings directly, with no reference to phonemes. Or, more faithful to the modern dual-route concept, a conversion of
phonemes to possible letters goes on in parallel to the lexical look up, with the two sources of information competing or
converging to various degrees. A parallel argument is made for reading. Although both lexical and grapheme phoneme routes
may be used in word reading, the lexical route is more reliable and will ''win'' most of the time, relegating the phonemic route to
backup duty for low-frequency regular words. Whatever the scientific status of these dual-mechanism models, the point here is
merely how much their development has been dependent upon a particular orthography notorious for its "irregularity."
In the case of spelling, there is an important additional observation that reinforces the impression that a lexical look up is
necessary. Even if the spelling is generated from phoneme grapheme conversion, the candidate spelling must be compared with
the speller's representation of the written word. This verification process, usually implicit, becomes explicit and visible when a
speller writes down a candidate spelling to see whether it "looks right." This verification process is likely to be quite general,
perhaps universal, across orthographies. By contrast, the initial process of generating spellings on the basis of phonology may be
influenced by linguistic constraints to different degrees across different orthographies. This is because orthographies differ
exactly in the degree to which they directly represent the phoneme strings of the language they encode.
In particular, orthographies vary in the degree to which they encode the surface phonology of the language relative to the
morphology. Those orthographies that reflect relatively faithfully the surface phonology of the language (i.e., its linear string of
phonemes) are referred to as shallow or transparent; those that reflect more the morphology of the language (at the

Page 25
expense of the phonology) are considered deep orthographies (Frost, Katz, & Bentin, 1987). English tends toward the deep end
of the continuum. Its spellings are often unfaithful to surface phonology while reflecting the deeper morphology and
morphophonology. At least, that is the common assumption. In fact, there is more systematic phonology in English spelling than
is commonly assumed especially when positional and letter context constraints are considered; however, it is still a relatively
deep orthography.
Orthographies Reflect Linguistic Structure

Within the family of alphabetic writing systems, Italian provides an interesting contrast to English. It is a shallow orthography,
one in which the spelling of words mirrors their pronunciation in a more reliable fashion. The reliable mapping from
orthography to phonology makes Italian a transparent, shallow system for reading, and its relatively reliable mapping from
phonology to orthography makes it a transparent, shallow system for spelling. This symmetrical transparency contrasts with
orthographies that show a significant asymmetry between the spelling and reading transparency. For example, the orthography
phonology (reading) mapping can be more reliable than the mapping for phonology to orthography (spelling). This is the case
for French, for example, with very interesting consequences for spelling (see Totereau, Thevenin, & Fayol, chap. 6, this
volume).
To appreciate the extent to which the orthography phonology mapping functions can be made quite reliable, consider the Italian
-ci- spelling, as in ciao. To the reader of Italian, the -ci- pattern signals a reliable mapping rule: -ci- maps to /c/ in all
environments. So ciao is chow /cau/ and bocci is /boci/. But bocca, identical except for the a, is /boka/. It doesn't matter that in
bocci, the i is pronounced, whereas in ciao, it is not pronounced. (Whether the i is pronounced is also quite predictable from the
grapheme environment i before a, o, and u is inserted to signal the pronunciation of the preceding c, but is not pronounced.) In
both cases, the i signals that the c is /c/ and not /k/. The same signal is given by -e-. Thus, cello, liceo, and piacere all contain
the c before the e (which is not pronounced in liceo /lico/, but is in cello and piacere). Both -i- and -e- contrast with -a-, -o-,
and -u-, which signal a preceding /k/: cara, ancora, cucina (the last shows both sides of the rule within a single word).
This much may appear to be an arbitrary, if rather convenient, conventionalization having nothing to do with linguistic
structures. However, consider the fact that g behaves the same way as c. Its pronunciation depends on whether it precedes -i- or
-e-, in which case its phonemic value is /j/, or whether it precedes -a-, -o-, or -u-, in which case its phonemic value is /g/. Thus:
giorno, baggio, Georgio, and gellato (all /j/), but galleria, albergo, and

Page 26
pagare (all /g/). That c and g pattern together is, of course, no coincidence. The four phonemes that map onto the two letters are
cousins, paradigmatically related. /k/ and /g/ are full stop consonants that share place of articulation (the velum) and contrast in
voicing /k/ is unvoiced, /g/ is voiced. /c/ and /j/ are affricates (stops followed by sudden release) that share place of articulation
(the hard palate) and also contrast on voicing /c/ is unvoiced, /j/ is voiced. Thus, the paradigm is one that pairs phonemes on two
features place and manner of articulation but contrasts them on one feature voicing. Spelling patterns cohere around the
paradigm: Loosely speaking, -i- and -e- signal the tongue to move from the velum to the hard palate, shifting the place of
articulation slightly forward to where the affricate sounds are made. The other vowels define the default case, the velar
consonant voiced if there is a g and unvoiced if there is a c.
To summarize, this example illustrates how a transparent orthography can map abstract phonology, not merely surface
(phoneme) phonology. The spelling rules for the four consonants reflect an underlying articulation-defined paradigm. They are
conventions, but they are not mere conventions. Underlying the patterns of co-occurrence in the letters is a regularity in the
phonological system. Nor is this pattern seen only in Italian: French and other Romance languages have orthographies that show
a parallel pattern for c and g (associated with different pairs of phonemes). And English, despite its irregularities, gives the
identical pattern, but imperfectly: city versus cutey; candy versus Cindy; cell versus call; gender versus gander, barge versus
bargain, and so on. These examples, which reflect Latinate influences in the evolution of English, need to be kept in mind when
the irregularity of English spellings is lamented. To say, in English, that "c is sometimes /k/ and sometimes /s/" is misleading as
an example of irregularity.
So far, the argument has been that spelling can reflect abstract phonology. This is good for reading, decoding orthography to
phonology. But what about spelling itself? In the Italian case, the mapping from phonology to orthography is itself very reliable.
If a spoken word contains a /c/, the speller knows that there is a c followed by an e or an i. There is no confusion as to which it
is: Unlike English, the vowel that is heard is the vowel that is spelled, and there is no mistaking the phoneme values /i/ with /e/
or /ey/. What happens if no /i/ or /e/ is heard? Then the c must be followed by -i-. It alone has the duty to mark the c spelling
without phonemic value (before a, o, u). Unlike English, if a /k/ is heard, it must be spelled with a c because Italian has no k.
However, what happens if a /ki/ or a /ke/ is heard, as in bichierri? The skilled speller will not be tempted to spell the sound as
-ci- because -ci- has to be a /c/, not a /k/. Instead the Italian speller knows that h serves that unique purpose: Inserted between
the c and the -i-, it signals /k/, and, of course, with g, it is the same. The plural of albergo is alberghi. The pronunciation is
maintained.

Page 27
The alberghi example illustrates one final point about abstract systematic linguistic influences in spelling. It is often said that
English sacrifices surface phonology to preserve morphology. For example, the root morpheme in national is spelled the same
as nation even though there is a pronunciation shift in the first vowel. This serves the reader and probably the speller by drawing
attention to the morphemes of the language at the cost of not representing the phonemes. (Much of the phoneme alteration is
predictable, so there is little cost to phonological representations.) Notice, however, how the Italian example satisfies
morphology and phonology both by a simple spelling rule. The spelling of alberghi directly signals the pronunciation, which is
constant across morphological variants in this case. Other Italian examples illustrate the complementary case: pagare, an
infinitive form, converts to pagiamo in the second person plural. Notice that the pronunciation of g must change because of the
rule controlled by vowel spellings. The phonological shift, of course, is not imposed by the spelling but is present in the
inflection system, which is part of every speaker's linguistic competence. Thus, spelling always conforms to pronunciation,
which in turn reflects the morphological system.
In summary, I have taken a simple example to make a point about how linguistic systems can constrain spelling. Comparisons
across orthographies are tricky, however, even within those using the same alphabet, such as Italian and English. Indeed,
English has a lot more linguistic constraints in its orthography than is sometimes implied (Venezky, 1970). The mere linear
arrangement of letters reflects phonotactic constraints on permissible phonemic sequences. There is a difference between deep
and shallow orthographies, but even deep ones within the family of alphabetic writing systems embed linguistic constraints
within their conventions.
The Case of Chinese
From the examples so far, one might conclude that linguistic structures can be expressed in the orthographies of alphabetic
writing systems (to various degrees) but draw no conclusion about a related and broader question: Do all writing systems reflect
linguistic structures or only alphabetic ones? A widely shared assumption seems to be that writing systems reflect different
principles of design and that linguistic units play a role in alphabetic and syllabary systems but not in all systems. In particular,
Chinese, as the prime example of a meaning-based system, is sometimes assumed to reflect strictly semantic designs, with
written units corresponding to concepts (ideographs) rather than to units of the language. Indeed, the origins of written Chinese,
like those systems originating in the Middle East, appear to have been in pictorial elements. The long history of development of
the Chinese system over more than 3,000 years, however, has yielded a system in which pictorial elements are negligible.
Students of Chinese writing, although perhaps at odds over

Page 28
some aspects of writing system classification (cf. DeFrancis, 1989; Gelb, 1963; Sampson, 1985; Wang, 1973), certainly agree on
the preponderance of abstract form meaning relations in modern Chinese. Even more important for the question at hand,
DeFrancis argued that Chinese has developed into a morphophonological system, one in which compound characters, which
include the vast majority of words, contain both a meaning (morphological) and a phonetic (phonological) component. Although
there is agreement that the writing system has phonological elements, some writers continue to focus almost exclusively on the
semantic component, ignoring the phonological component that is critical in the system. One may speculate that the scriptal
difference between Chinese and alphabetic systems has made its semantic component more visible (by contrast to alphabetic
systems).
That Chinese relies on phonetic elements is very important in establishing it as a system based on language rather than on ideas
or concepts. Because of its morpheme component, it has a dual-level mapping, one that simultaneously and with different
graphic units within the word signals meanings and pronunciations (see Perfetti & Zhang, 1995, for more on this aspect of
Chinese from a reading perspective). In the scheme depicted in Fig. 2.1, written Chinese provides graphic units that are single
syllables; simultaneously, they are morphemic elements that give meaning information.2
That writing systems use a mix of features rather than purely implemented principles may be a valid generalization, as Jaffr
(chap. 1) argues in this volume. At the same time, the Chinese case also illustrates the remarkable degree to which all writing
systems depend on language. To be more precise, although encoding the morphemes of a language into a writing system is a
linguistic mapping, it is the reliance on phonological mappings in all writing systems that is truly remarkable and important in
understanding the relationship between writing systems and language. Writing systems, however variably and unevenly, rest on
spoken language. This fact has strong implications for how reading and spelling work as general human cognitive processes.
The Spelling-Reading Relationship Revisited

Spelling and reading are two sides of a coin from the perspective I take here. The first converts speech forms to written forms;
the second converts written forms to speech forms. This logical symmetry, however, does not
2 An interesting question is what constitutes Chinese spelling. One reasonable answer is that there is no spelling in
Chinese. There is the retrieval of graphic forms and their composition for specific words, but this is not what we usually
mean by spelling. On the other hand, it is possible to suggest that Chinese spelling, as in an alphabetic writing system, is a
matter of providing written word constituents, whatever the psychological source of these constituents.

Page 29
demonstrate that the psycholinguistic processes are mirror images. Indeed, the consensus view has been that spelling and reading
are rather different things. One of two observations is typically made on this point: (a) Spelling is more difficult than reading.
Thus, reading does not require the degree of knowledge about words that spelling requires. In this view, spelling subsumes
reading; (b) spelling and reading can be completely disassociated. Depending on the circumstances, spelling can be easier or
more difficult than reading. On this view, there are different mechanisms involved in the two processes.
The first of these corresponds to a commonsense view of the spelling reading relationship, conforming to the irrefutable fact that
most people can read words that they cannot spell. The second possibility derives from dual-mechanism accounts of reading and
spelling. When words can be spelled but not read, this is because reading relies on a lexical look-up procedure that fails for a
particular word; however, the same word, when spoken, can be spelled on the basis of phoneme grapheme conversion
mechanisms. Bryant and Bradley (1980; Bradley & Bryant, 1979) reported observations on children who showed this kind of
disassociation. Similarly, Cossu and Marshall (1985) reported data on two Italian children who could read only a fraction of the
words they could spell. Interestingly, these children showed this large advantage for both words and nonwords, even while
performing better on words than nonwords in both reading and spelling.
The representational assumptions of the restricted interactive (R-I) model of reading outlined in Perfetti (1991, 1992) provides a
different view on this issue. In this model, reading and spelling are the same in the lexicon; they are different in important
processing details. Their sameness comes from their mutual dependence on the quality of representations of printed word forms.
This quality has two components that are enhanced with increased experience in reading and spelling: precision and redundancy.
Precision is the probability that specific letter constituents are represented as part of a word in the reader's lexicon. Even
experienced readers can be characterized as "missing" some letters for some words; such letters are represented as variables
rather than constants. A speller, in such a case, may or may not produce a correct spelling. Redundancy is the formation of
word-specific, grapheme phoneme connections. These connections are developed by the convergence of generalized grapheme
phoneme correspondences and specific orthographic forms. They supplement the connection that can exist between a printed
word and a spoken word form. Only one set of print phonology connections is logically necessary; two provide a safety net that
assures successful and fluent word reading in a variety of situations. In fact, form-form connections (orthography-phonology)
can develop at several levels (phoneme, syllable, onset, morpheme), producing, at high levels of skill, a large amount of helpful
redundancy.

Page 30
There is an important learning or reading acquisition assumption in the R-I model relevant for spelling: Improvements in both
lexical precision and redundancy depend on phonemic knowledge (Perfetti, 1991). The reader's knowledge of speech segments
draws attention to the letters of a word and encourages their full representation. The unpalatable alternative is to imagine the
child developing strictly nonlinguistic (presumably visual) strategies for attending to the letters of a word and representing them.
A similar idea was developed more thoroughly by Share (1995) in a theoretical review of beginning reading issues (see Share &
Stanovich, 1995). Based on earlier work by Jorm and Share (1983; Share & Jorm, 1987), Share argued that phonological
knowledge, specifically phonological recoding, is the fundamental mechanism for the acquisition of orthographic knowledge. In
terms that combine Share's model with the R-I model, phonological recoding is a self-teaching procedure that produces more
precise and more redundant lexical representations.3
These two principles of lexical quality determine the lexical expertise of the reader. More precision and more redundancy add up
to more expertise. Most relevant for the present purpose is the assumption that spelling is the purest indicator of lexical quality.
Lexical expertise, the claim went, is best assessed by idealized spelling performance (Perfetti, 1991, 1992).
This assumption seems at odds with the disassociationist view of the spelling-reading relationship. It is more compatible
however, with the commonsense spelling-subsumes-reading view. Spelling is more difficult than reading. Reading can be
accomplished with incomplete word representations (i.e., representations with letters as variables). Deciding which word is
accessed by a letter string is a matter of choosing one word over a cohort of similar words (neighbors), and this is possible even
with imperfect word representations. In addition, decoding of letter sequences to phonemes can provide a match to the
phonological representation of the target word. By contrast, spelling requires the retrieval rather than the recognition of the
graphemes. Retrieval processes are errorful because (a) memory representations are low quality (imprecise in this case), or (b)
interference is encountered (with competing letter sequences, in this case). On this account, errorless spelling comes with ample
practice many successful word recognition events and enough successful spelling retrieval events to strengthen the precision and
redundancy of the word representations. We expect, as the evidence reviewed by Bosman and Van Orden (chap. 10) suggests,
that reading by itself will not dramatically improve spelling because reading does not practice the full orthographic retrieval
process demanded
3 The general idea that phonological knowledge has the role of assisting in the construction of orthographic knowledge
was earlier proposed by Venezky and Massaro (1979), who argued that phonics instruction might help children acquire
knowledge of orthographic structure.

Page 31
by spelling. Moreover, it is spelling itself that is most effective at improving the quality of the word representation. Practice at
spelling should help reading more than practice at reading helps spelling.
Also on this account, spelling words that cannot be read should be rare and should occur only during the stages of reading
acquisition (and perhaps certain rare acquired dyslexias). Indeed, the data on the disassociation of spelling and reading are
consistent with this view. In Bryant and Bradley (1980), the pattern of spelling-better-than-reading was restricted to younger and
less skilled readers. Similarly, the case studies reported by Cossu and Marshall (1985) were studies of low-ability readers. Thus,
what needs to be explained is not a general disassociation pattern but rather the occasional case in which a child who is very
poor in reading seems to spell better than one should expect. If the claim that spelling is the indicator of high quality lexical
representations is correct, then these cases are failures in which reading falls short of the potential provided by lexical
representations. This conclusion appears at odds with the usual assumption that spelling in cases of disassociation occurs only by
translation processes and not by access-to-word representations. However, this latter assumption does not square with the
observation that nearly all children, including children who spell words they fail to read, spell real words more accurately than
nonwords (Cossu, Gugliotta, & Marshall, 1995; Cossu & Marshall). Lexical representations play a part in the spelling of words
for all children who have acquired some degree of literacy.
So what is the source of the spelling advantage for those relatively few children who show one? It is quite possible that there is
no single mechanism but rather a range of ineffective word reading strategies. When they are having trouble acquiring reading,
children may develop conflicting strategies for reading words. The recognition of a regular word (i.e., one with predictable
grapheme phoneme connections) may fail because the child's lexicon does not contain the word, preventing verification, or
because the child is reluctant to implement imperfect knowledge of grapheme phoneme correspondences. The occasional success
in spelling for the same word presented orally can be explained as follows. The spoken form does what the written form cannot
do for a struggling reader: It activates a lexical representation. The spelling can be generated from this representation with the
help of phoneme grapheme correspondences. Whatever the correct set of explanations, it is important to remember that what
must be explained is poor reading in the presence of fair (not good) spelling; there is no need to try to explain good spelling
because this is not what is observed in the disassociation.
To place things in perspective, we return to the dominant pattern of reading-better-than-spelling. Children read words more
successfully than they spell them, even in a transparent orthography such as Italian, in which

Page 32
phonology-spelling mapping is very consistent. Indeed, Cossu et al. (1995) found that among 70 first- and second-grade
children, nearly all children read words they could not spell, and no children showed the opposite pattern. In general, the
assumptions of the R-I model can handle the observed patterns of spelling-reading asymmetry. The common pattern is reading-
better-than-spelling with the reversed pattern restricted to readers of low reading skill. Both patterns can be explained in terms
of processes that operate on lexical representations of varying quality. Reading is usually more successful because it can tolerate
noise in the representation to a greater extent than spelling can. When the representations are low quality, both reading and
spelling are poor. Occasionally, ineffective reading processes fail to take advantage of the lexical representations that spelling
success suggest are established.
A theory of spelling, on this account, requires a distinction between spelling words and spelling phoneme sequences. Word
representations, including their letter constituents, must be consulted in spelling familiar words (i.e., words whose
representations include graphemes). The spelling of unfamiliar words and nonwords may occur, in principle, without consulting
the lexicon. I say "in principle" because the process of spelling a pronunciation is likely to always include a search of the
lexicon. Finding a lexical representation invokes a process that is both orthographic and phonological because both kinds of
information are intrinsic to the word representation. Failing to find a lexical representation triggers a process that converts
phonemes to graphemes, probably in consultation with lexical entries that share phonemes with the unfamiliar target word or
nonword.
This account appears compatible with a dual-route mechanism, and in a general way, it is, but there is a difference in the
assumption of the R-I model that phonological information is bonded to orthographic representation for a skilled reader. (This
assumption is closely related to Ehri's, 1978, amalgamation idea.) The interconnection, within a single word, of graphemes and
phonemes, means that lexical processes themselves refer to sublexical information, including phonology. Thus, to the extent that
the representation of the word is redundant containing context sensitive, grapheme sequence phoneme sequence connections
phonology has an intrinsic role in any lexical process on that word, including spelling.
This emphasis on phonology in the word representation is not to minimize the role of visual representations, which are important
in spelling. The compelling tendency to verify spellings by writing them down for observation is probably only a surface
indicator of the extent to which visual representations are used routinely in spelling. The verification process itself, however,
seems to provide a link between reading and spelling. Notice that verifying a candidate spelling is the correct spelling of the
target word is a process of reading. Thus, spelling itself entails reading.

Page 33
Reading in Spelling, Spelling in Reading
In this final section, I address a more specific implication of the R-I model's assumption that spelling and reading rely on the
same mental representation, one that consists of tightly interconnected letters and phonemes. If there is a single representation
and if this representation is consulted during reading and during spelling, then there is a sense in which spelling a word entails
reading it and reading a word entails spelling it.
The suggestion that spelling entails reading was made in the preceding section. Going from phonological form to orthographic
form begins with a search of the lexicon and ends, in the case of a word, with the verification of the candidate spelling. This
verification is a "reading" of the constituent letters. However, this verification involves phonology as well as spelling.4 The
reading is verified when the phonological-orthographic forms that comprise the lexical representation are retrieved and
compared with the candidate spelling. For a regular word such as /reyk/, the process finds a lexical entry {rake-/reyk/} (omitting
for simplicity the redundant letter-phoneme connections). It reads out the orthographic constituents of the representation r-a-k-e.
It then reads this form as /reyk/, completing a cycle of spelling and reading that confirms the spelling. If the spelling yields
some other pronunciation, a common condition in an opaque orthography, then there will be competition (interference) during
this verification process. Notice that, in verifying the spelling pattern -ake, there is no competition.
Spelling an irregular word is not so different if the word is familiar. Suppose the word is /kefa */. A search of the lexicon with
/kefa */ as an input yields {cafe-/kefa*/}. The spelling process reads out the orthographic constituents of the representation: c-a-
f-e. It then reads this form as /kefa */, although now with some interference from /keyf/. However, partly because /keyf/ is not a
word, /kefa */ easily wins out, and the spelling is confirmed. A more important reason it wins is that /kefa */ has just been
activated in the oral spelling process, and this phonological form is primed to be connected with c-a-f-e, which, of course,
activates /kefa */ on its own. Thus, there are convergent cycles of spelling and reading that go on when both regular and
irregular words are spelled.
I have already indicated how some spelling errors might be made. The main difficulty arises from words that are unfamiliar. The
verification process fails and the speller is stuck with the form generated by phoneme-spelling mappings, analogies, or by
whatever set of strategies is used in a particular case. Another error that gives a glimpse of the spelling mechanism is
homophone errors. To spell /riyd/, a search of the lexicon can turn up {reed-/riyd/} or
4 Bosman and Van Orden (chap. 10) provide a detailed argument for the idea that spelling and reading reciprocally rely
on a phonological process.

Page 34
{read-/riyd/} or both. Because these are two distinct lexical entries, sentence context will tell the speller which word it is, and
the homophone will be suppressed. (The homophone will not be successfully suppressed all the time, however, as the numerous
homophone errors in writing, e.g., right instead of write, demonstrate.) In isolation, however, either entry can be accessed and
the reading check will not distinguish between them.
The involvement of spelling in reading is a mirror image of the process just described, one of verifying a generated candidate
string against a lexical entry. This involvement is less obvious than the reading-in-spelling case, where one can appeal to an
occasional explicit verification process (i.e., writing in order to read). In reading, the idea that there is a spelling verification
stage is implied by complex systems resonance models of word identification (Stone & Van Orden, 1994; Van Orden &
Goldinger, 1994).5 In such models, the identification of a printed word comes about by a convergence of phonological
information generated by the letters of a word (a feedforward process), with the letter information generated by the phonology (a
feedback process). These rapid cycles of feedforward phonology (given a letter string) and feedback spelling (given the
activated phonology) constitute a resonance that singles out one word candidate from the host of word candidates that are
partially activated from the initial letter string.
An especially interesting implication of this model is that reading efficiency depends on spelling variability. Roughly, the more
ways a phoneme sequence can be spelled in English, the longer it takes to read any word that contains that phonological
sequence. This is because more possible spellings lead to more possible word candidates (actually letter sequences); with fewer
possible spellings, the system settles more quickly on the target word. For example, the word shelf is more efficiently read than
the word sneer. This is because the pronunciation /elf/ is always spelled elf, so there is no competition in the spelling feedback
phase; by contrast, the pronunciation /ier/ associates with several spellings (eer, ear, ier, and er), creating competition in the
spelling feedback phase of reading. Evidence favoring the operation of this spelling feedback mechanism rests on reading
experiments with four- to six-letter words (Stone, Vanhoy, & Van Orden, 1997).
In summary, the suggestion that spelling and reading processes make reference to a single lexical representation that contains
interconnected orthographically specified and phonemically specific constituents now has some evidence as well as some
intuitive appeal. Spelling processes seem to use reading in a verification stage; reading processes seem to use spelling
5 Earlier word reading models also contain the idea of verification. For example, the activation-verification model of
Paap, Newsome, McDonald, and Schvaneveldt (1982) appealed to a verification stage in which a set of word candidates
activated during letter encoding is tested for their goodness of fit to the encoded letter string. The verification stage in
such models, however, does not rely specifically on spelling as the verification criterion.

Page 35
in a verification stage. These assumptions lead to a general model of the reading-spelling relationship depicted in Fig. 2.2.
The reading-spelling relationship, as shown in Fig. 2.2, is one based on access to lexical representations shared by the two
processes. These representations (for skilled readers) contain bonded orthographic and phonological elements that are contacted
by orthographic inputs (reading) and phonological inputs (spelling). Both reading and spelling contain a verification process in
this model. In reading, orthographic forms activate phonological forms (often competing forms) that activate orthographic
forms. The orthographic form verifies the word identity. In spelling, phonological forms activate orthographic forms that
activate phonological forms that verify the word identity. In both cases, the verification is ordinarily rapid and unconscious;
however, the fact that spelling is the retrieval of letter constituents rather than their recognition means that spelling is effortful
enough in some situations to create a slow and conscious trace of this process. In skilled reading this is seldom the case.
However, in reading unfamiliar words, one occasionally can observe the full-blown version of verification when a reader tries a
pronunciation, compares it with the visible spelling, and, recognizing that the pronunciation does not look like the spelling (i.e.,
this pronunciation is usually spelled some other way), tries again.
In the context of this model, we can reconsider the question of why spelling is generally more difficult than reading. Bosman
and Van Orden
Fig. 2.2.
The spelling-reading relationship. Both processes use a lexical representation
that contains orthographic (O) and phonological (Ph) constituents. Both
spelling and reading use both sets of constituents, and both have a
verification stage.

Page 36
(chap. 10, this volume) argue that English spelling is more difficult because of the greater variation in phonological-to-spelling
mappings than in spelling-to-phonological mappings. This assumption is reflected in the difference between the O->Ph and the
Ph->O connections in Fig. 2.2. However, this difference is only part of the story, and it is not clear that it is the major part. As
argued previously, the dramatic difference in the retrieval frequency of orthographic lexical forms compared with the recognition
frequency of orthographic lexical forms can produce the advantage of reading over spelling.
Finally, it is interesting to return to the issue of orthographic differences in this context. English, Dutch, German, and French all
alphabetic orthographies that are topics of discussion in other chapters of this volume vary in their spelling regularity (i.e., the
direction of letter sequence to phoneme predictability). Dutch and German are very regular (shallow orthographies); French is
comparable in this respect but in the other direction phonology to spelling is highly inconsistent (see Totereau, Thevenin, &
Fayol, chap. 6, this volume). Italian, discussed earlier in this chapter, is quite consistent in both directions; English has
considerable inconsistency in both directions. In comparisons of reading and spelling across orthographies, it may not be enough
to note just the shallowness of the orthography as it has been defined in recent years since the work of Frost, Katz, and Bentin
(1987). It is necessary to recognize the bidirectional nature of spelling-phonology relationships.
One might try to correct the situation by referring to shallow phonologies (e.g., Italian) and deep phonologies (e.g., French and
English) as well as to deep and shallow orthographies. The depth of the phonology would be the extent its pronunciations, as
opposed to its morphology or other factors, control spellings. To a considerable degree, it is the phonology-spelling conventions
over the centuries that have led to spelling-phonology inconsistencies. (The influence goes the other way as well; the way words
get spelled influences the course of their pronunciation changes.) Nevertheless, we should probably resist the temptation to
expand the deep-shallow distinction in this way because it suggests a misleading view of phonology that it mainly serves
spelling, and if it serves it well, then phonology is shallow. This appears to get things backward to some extent. Instead, we can
refer generally to the transparency of an orthography. This transparency has two components: the transparency of the phonology
through the spelling and the transparency of the spelling through the phonology. Orthographies can be further characterized as
symmetrically transparent or asymmetrically transparent. It is interesting that there seem to few cases in which the transparency
asymmetry favors phonology to spelling over spelling to phonology. Most cases seem to be exemplified by French, in which a
reasonably transparent reading process accompanies an opaque spelling process. This

Page 37
appears to be a consequence of the historical development and pattern of use of full (consonant plus vowel) alphabetic writing
systems. (The vowelless orthographies, e.g., Hebrew, may reverse this asymmetry.)
Summary
Spelling is a psycholinguistic problem involving multilayered linguistic units (morphology and phonology) and writing units.
Writing systems constrain the general form of the mapping between units, but all writing systems, including nonalphabetic ones,
have evolved to encode speech as well as meaning. Within alphabetic writing systems, linguistic constraints are embedded at
different levels. Well-known examples from Italian illustrate how these constraints can affect reading and spelling at fairly
abstract levels. The relationship between spelling and reading is one of shared lexical representations and different processes.
Implications of a model of reading, the R-I model, suggest that both orthographic and phonological constituents play roles in
both reading and spelling. Furthermore, a verification process appears to be part of both spelling and reading. Because this
verification process combines information from spelling and reading, it gives specific definition to the claim that reading and
spelling are similar.
References
Besner, D. (1987). Phonology, lexical access in reading, and articulatory suppression: A critical review. Quarterly Journal of
Experimental Psychology, 39A, 467-478.
Bradley, L., & Bryant, P.E. (1979). The independence of reading and spelling in backward and normal readers. Developmental
Medicine and Child Neurology, 21, 504-514.
Bryant, P.E., & Bradley, L. (1980). Why children sometimes write words which they do not read. In U. Frith (Ed.), Cognitive
processes in spelling (pp. 355-370). London: Academic Press.
Coltheart, M. (1978). Lexical access in simple reading tasks. In G. Underwood (Ed.), Strategies of information processing (pp.
151-216). New York: Academic Press.
Cossu, G., Gugliotta, M., & Marshall, J.C. (1995). Acquisition of reading and written spelling in a transparent orthography: Two
nonparallel processes? Reading and Writing, 7, 9-22.
Cossu, G., & Marshall, J.C. (1985). Dissociation between reading and written spelling in two Italian children: Dyslexia without
dysgraphia? Neuropsychologia, 23, 697-700.
DeFrancis, J. (1989). Visible speech: The diverse oneness of writing systems. Honolulu: University of Hawaii.
Ehri, L.C. (1978). Beginning reading from a psycholinguistic perspective: Amalgamation of word identities. In F.B. Murray
(Ed.), The development of the reading process. (International Reading Association Monograph No. 3, pp. 1-33). Newark, DE:
International Reading Association.
Ehri, L.C. (1980). The development in orthographic images. In U. Frith (Ed.), Cognitive processes in spelling (pp. 311-388).
London: Academic Press.
Frith, U. (1980). Unexpected spelling problems. In U. Frith (Ed.), Cognitive processes in spelling (pp. 495-516). London:
Academic Press.

Page 38
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K.E. Patterson, J.C. Marshall, & M. Coltheart (Eds.),
Surface dyslexia: Neuropsychological and cognitive studies of phonological reading (pp. 301-330). London: Lawrence Erlbaum
Associates.
Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and orthographical depth: A multilingual
comparison. Journal of Experimental Psychology: Human Perception and Performance, 13, 104-115.
Gelb, I. J. (1963). A study of writing (rev. ed.). Chicago: University of Chicago Press.
Gernsbacher, M.A. (Ed.). (1994). Handbook of psycholinguistics. San Diego: Academic Press.
Henderson, E.H. (1981). Learning to read and spell: The child's knowledge of words. DeKalb: Northern Illinois University
Press.
Jorm, A.F., & Share, D.L. (1983). Phonological recoding and reading acquisition. Applied Psycholinguistics, 4, 103-147.
Paap, K.R., Newsome, S.L., McDonald, J.E., & Schvaneveldt, R.W. (1982). An activation-verification model for letter and
word recognition: The word-superiority effect. Psychological Review, 89, 573-594.
Paap, K.R., Noel, R.W., & Johansen, L.S. (1992). Dual-route models of print to sound: Red herrings and real horses. In R. Frost
& L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 293-318). Amsterdam: North-Holland.
Perfetti, C.A. (1991). Representations and awareness in the acquisition of reading competence. In L. Rieben & C.A. Perfetti
(Eds.), Learning to read: Basic research and its implications (pp. 33-44). Hillsdale, NJ: Lawrence Erlbaum Associates.
Perfetti, C.A. (1992). The representation problem in reading acquisition. In P.B. Gough, L.C. Ehri, & R. Treiman (Eds.),
Reading acquisition (pp. 145-174). Hillsdale, NJ: Lawrence Erlbaum Associates.
Perfetti, C.A., & Zhang, S. (1995). The universal word identification reflex. In D.L. Medin (Ed.), The psychology of learning
and motivation (Vol. 33, pp. 159-189). San Diego: Academic Press.
Sampson, G. (1985). Writing systems: A linguistic introduction. London: Hutchinson.
Share, D.L. (1995). Phonological recoding and self-teaching: Sine qua non of reading acquisition. Cognition, 55, 151-218.
Share, D.L., & Jorm, A.F. (1987). Segmental analysis: Co-requisite to reading, vital for self-teaching, requiring phonological
memory. European Bulletin of Cognitive Psychology, 7, 509-513.
Share, D.L., & Stanovich, K.E. (1995). Cognitive processes in early reading development: Accommodating individual
differences into a model of acquisition. Issues in Education, 1, 1-57.
Stone, G.O., Vanhoy, M., & Van Orden, G.C. (1997). Perception is a two-way street: Feedforward and feedback phonology in
visual word recognition. Journal of Memory and Language, 36, 337-359.
Stone, G.O., & Van Orden, G.C. (1994). Building a resonance framework for word recognition using design and system
principles. Journal of Experimental Psychology: Human Perception & Performance, 20, 1248-1268.
Treiman, R. (1993). Beginning to spell: A study of first-grade children. New York: Oxford University Press.
Van Orden, G.C., & Goldinger, S.D. (1994). The interdependence of form and function in cognitive systems explains perception
of printed words. Journal of Experimental Psychology: Human Perception and Performance, 20, 1269-1291.
Venezky, R.L. (1970). The structure of English orthography. The Hague: Mouton.
Venezky, R.L., & Massaro, D.W. (1979). The role of orthographic regularity in word recognition. In L.B. Resnick & P.A.
Weaver (Eds.), Theories and practice in early reading (Vol. 1, pp. 85-107). Hillsdale, NJ: Lawrence Erlbaum Associates.
Wang, W.S.-Y. (1973). The Chinese language, Scientific American, 228, 50-60.

Page 39
Chapter 3
The Cognitive Neuropsychology of Spelling
Pascal Zesiger
Marie-Pierre de Partz
University Hospital of St.-Luc, Brussels
Neuropsychologists began to get truly interested in spelling in the 1980s. Until then, the dominant position was that the
production of written language1 was essentially dependent on the mechanisms responsible for the production of oral language.
Consequently, the disorders of written production the agraphias2 were mainly viewed as being secondary to the deficits
affecting oral language the aphasias resulting themselves from lesions of the left hemisphere. Nonfluent agraphias were usually
associated with frontal lesions and fluent agraphias with more posterior lesion sites (Benson & Cummings, 1985; Hcaen,
Angelergues, & Douzenis, 1963). However, even if most authors admitted that the intensity and the types of disorders affecting
the oral and the written domains could vary (Marcie & Hcaen, 1979), the existence of selective deficits of written production,
called pure agraphias, was difficult to conciliate with this perspective. That is why some authors, following Exner (1881),
hypothesized the existence of an autonomous center for writing located in the base of the second frontal convolution.
1 The production of written language requires a set of pragmatic, syntactical, semantic, and lexical processes as well as
perceptuomotor mechanisms allowing the movements necessary for the production of letters in the graphic space of the
page. We limit ourselves to considering the processes responsible for the generation of strings of graphemes forming
words or pseudowords.
2 The term agraphia has been used to qualify the disorders of different aspects of written production. A distinction was
usually drawn between aphasic agraphia (deficits of the linguistic aspects), apraxic agraphias (disorders related to the
production of letters), and spatial agraphias (disruption affecting the arrangement of letters in graphic space).

Page 40
After the publication of Marshall and Newcombe's (1973) study on reading, researchers started to analyze differently the errors
produced by brain-damaged patients by referring to an information-processing model. The adoption of a cognitive perspective in
neuropsychology, more focused on the specification of the processes involved in a certain behavior than on the definition of the
relation between faculty and substrate, modified the viewpoint on spelling considerably. The presentation of these new concepts
constitutes the essential part of this chapter. First, we discuss the domain specificity of spelling and describe the cognitive
architecture underlying this capacity. Second, we examine the developments that have occurred in this domain over the last 15
years. Third, we present some issues that are currently subject to discussion.
The Specificity of the Spelling Processes

Several studies have shown that acquired disorders of spelling were not necessarily associated with disruptions of oral language.
For instance, the analysis of the files of hundreds of patients, done by Basso, Taborelli, and Vignolo (1978), revealed that some
patients admittedly, a minority were affected by a selective alteration of written production without impairment of oral
production or, conversely, presented a deficit of oral production preserving written production. These data are consistent with
the description of other forms of dissociation, qualitative in nature. Assal, Buttet, and Jolivet (1981) described a patient
presenting both oral and written production disorders. These deficits expressed themselves very differently in the two modalities,
with an oral impairment being similar to a Wernicke's aphasia and a written disorder recalling the characteristics of a Broca's
aphasia. Finally, we point out that various studies showed that spelling disorders were not systematically associated with reading
impairments (i.e., Baxter & Warrington, 1985). Overall, these investigations converged towards the notion of specificity of at
least a part of the spelling processes that appear to be liable to be selectively affected by a brain lesion.
The Cognitive Architecture of Spelling Processes

The next step was made possible thanks to the publication of two cases that became famous in neuropsychological literature.
Beauvois and Drouesn (1981) described a French-speaking patient (RG) presenting an important spelling deficit. Detailed
investigations of RG's difficulties revealed the following characteristics: (a) Words produced by RG contained numerous errors,
whereas the written transcriptions of dictated pseudowords were

Page 41
all correct; (b) the more the words contained orthographic ambiguities,3 the more they were prone to errors; (c) orthographic
regularity interacted with word frequency low-frequency words were more sensitive to orthographic regularity than high-
frequency words; (d) neither the syntactical class (nouns, verbs, function words, etc.) nor the length of words seemed to affect
the patient's performance; and (e) the paragraphias produced by RG were mostly phonologically plausible errors (soeur-> seur,
rameau-> ramo). This patient presented a disorder that was limited to the production of familiar words with a preservation of the
capacities to generate pseudowords. The terms surface dysgraphia (or lexical or orthographic dysgraphia) are generally used to
refer to this type of deficit, later reported by several other authors (Baxter & Warrington, 1987; Goodman & Caramazza, 1986b;
Hatfield & Patterson, 1983; Roeltgen & Heilman, 1984).
On the other hand, Shallice (1981) described an English-speaking patient (PR) presenting the opposite pattern, known as
phonological dysgraphia. PR experienced considerable difficulties in writing dictated pseudowords even if they were simple
(18% correct responses), whereas his production of familiar words was mostly spared (94% correct responses). The errors
produced by PR were nonphonological (*bup-> coz). Complementary experiments enabled Shallice to show that PR's disorder
in the production of pseudowords was due to a double deficit: (a) segmenting auditorally presented pseudowords into isolated
phonemes; and (b) transcoding isolated phonemes into graphemes. Other patients have been similarly described (Bolla-Wilson,
Speedie, & Robinson, 1985; Roeltgen, Sevush, & Heilman, 1983).
Patients RG and PR constitute what is called in neuropsychology a double dissociation: In one of them, the capacity to perform
task X is impaired, but his capacity to execute task Y is preserved. In the other, the opposite pattern is observed. On the basis of
this double dissociation, following Morton (1980), several authors suggest that spelling processes are made up of two main
procedures, or routes: an addressing procedure (or lexical route), responsible for the generation of familiar words, and an
assembling procedure (or phonological route), responsible for the production of nonfamiliar words and of pseudowords (see Fig.
3.1). It was initially admitted that these two routes function independently from each other.
According to this model, a set of processes is necessary in order to generate a sequence of graphemes in response to a dictated
stimulus. The first processing level, which would be common to both procedures, analyzes the phonetic or acoustic features of
the stimulus. When it is a familiar word,
3 The orthographic ambiguity of a word corresponds to the presence of sounds liable to be translated in various ways in
the graphemic code (e.g., beak could be spelled beek). It is distinguished from orthographic irregularity, which denotes
the presence of exceptional phoneme-grapheme correspondences in the language (e.g., the /m/-> mn in autumn). In what
follows, we do not draw a distinction between these two forms of irregularity.

Page 42
Fig. 3.1.
Schematic diagram representing the minimal cognitive architecture
of the spelling processes.
the signal resulting from this analysis activates the phonological representation of the word in a phonological input lexicon. The
activated unit (or, more correctly stated, the unit for which activation level is the highest because this model rests on the
principle of passive parallel activation) within this lexicon transmits the information necessary for the activation of the meaning
of the word stored in a semantic system.4 This activation is then transmitted to an orthographic (or graphemic) output lexicon
that would contain the orthographic representations of words. The selected sequence of graphemes is temporarily stored in an
orthographic (or graphemic) buffer whose function is to maintain a trace of this sequence during the processing time necessary
to the processes responsible for the transformation of graphemes into sequences of movements leading to the production of
letters (peripheral processes).
4 This would be the first level activated in spontaneous writing production.

Page 43
When the dictated stimulus is unknown to the participant, it would temporarily be stored in a phonological buffer. According to
some authors, before the representation is placed in this buffer, it has to be transformed from the phonetic code into the
phonological code. It is most likely during this operation that the segmentation of the stimulus into phonemes is performed.
Once the string of sounds is stored in this working memory, a process of conversion from sounds to letters (or from phonemes
to graphemes) constitutes (or assembles) a string of graphemes that is temporarily stored in the orthographic buffer before being
processed by motor processes.
This model of the normal functioning of spelling processes allows interpretation of the two classes of deficits reported
previously (Ellis, 1982; Margolin, 1984). Surface dysgraphias can be explained by a disruption of the addressing procedure, and
phonological dysgraphias can be explained by an impairment affecting the assembling procedure. It should be highlighted that
the damage can be located at various levels of these procedures (and possibly at more than one level) within the components
themselves or within the connections between modules. Depending on the affected level, the expected pattern of performance
following a damage of a procedure can considerably vary.
Besides, impairments involving a dysfunction of the two routes have also been described. Several authors reported the existence
of a deficit, under the name deep dysgraphia, that affects the production of words as well as the production of pseudowords
(Bub & Kertesz, 1982; Hatfield, 1985; Nolan & Caramazza, 1983). Several types of errors are observed with words: (a)
semantic errors, in which the produced word has a related meaning with the dictated word (star-> moon, grass-> green); (b)
morphological errors, in which the produced word is a morphological derivative of the dictated word (relation-> relating); and
(c) visual errors, in which the produced word is visually similar to the dictated word (charge-> chance). Finally, two additional
phenomena have been described in this type of disorder: the effect of the degree of imagery of words and the effect of their
syntactical class. For the first one, words with a high degree of imagery (e.g., forest) are generally less affected by a lesion than
words with a low degree of imagery (e.g., norm). For the second one, the proportion of errors varies as a function of the
syntactical class they belong to: nouns, verbs, adjectives, function words, and so on. The interpretation of these phenomena
necessitates a more in-depth examination of the model and of the hypotheses underlying it.
The Specification of Processes

In this section we focus on the multiple processes of spelling that we try to define, on the one hand, as a function of the type of
errors observed after their damage and, on the other hand, as a function of their sensitivity to various factors.

Page 44
The Addressing Procedure
We limit ourselves to considering the semantic system and the orthographic output lexicon. An impairment of the semantic
system leads, within the framework of spelling disorders, to various deficits. The apparition of semantic errors is indeed
expected. These errors have been used to support the hypothesis of a categorical organization of this system. The production of
semantic errors can also be interpreted as a disruption of the connections between the semantic system and the orthographic
lexicon. What distinguishes these two possibilities is that, in the first case, semantic errors should be observed in all language
modalities (oral and written comprehension and production), but in the second case, such errors should only appear in written
production. The occurrence of semantic errors specific to written production seems to imply an alteration of the assembling
procedure as well (Miceli, Capasso, & Caramazza, 1994).
The effect related to the degree of imagery of words is usually thought to reflect another characteristic of the organization of the
semantic system. Thus, words with a high degree of imagery have more specific or more detailed meanings or are represented
by more constituting features than words with a low degree of imagery. Alternatively, they possess a different code than words
with low levels of imagery (visual vs. propositional). These properties make words with a high degree of imagery more resistant
to a lesion.
It is usually considered that an impairment of the orthographic lexicon results in the production of two types of errors. If the
representation of the word is entirely lost (or unaccessible), its generation is assumed by the assembling procedure.
Consequently, the production contains phonologically plausible errors that give us information about the functioning of the
assembling procedure, which is analyzed later. On the other hand, if the representation is only partially damaged, we should
observe errors that reflect the degradation of this representation (pyramid->pyminal, orchestra->orchestra) (Ellis, 1982, 1984;
Morton, 1980). For Patterson (1988), however, this last error type instead reflected problems of information degradation during
its transfer from the orthographic lexicon to the orthographic buffer.
A functional lesion of the orthographic lexicon necessarily leads to a disorder characterized by the presence of an effect of
orthographic regularity. This effect expresses the capacity of the assembling procedure to correctly spell regular words by the
application of phoneme-grapheme (P-G) conversion rules. The frequency effect is also associated with a lesion of this
component. It expresses a certain form of resistance to damage to the most frequent words. Finally, the effect of the syntactical
word class has been interpreted as reflecting the categorical organization (by syntactical class) of this lexicon (Baxter &
Warrington, 1985; Caramazza & Hillis, 1991).
A last phenomenon regarding the addressing procedure has been reported: It is homophonic confusions (their-> there, none->
nun) produced

Page 45
by some patients. Such errors can also be observed in the written production of normal participants (what is called ''slips of the
pen''). The occurrence of this type of paragraphias suggested the idea that the representations stored in the orthographic output
lexicon can be activated by phonological representations, without semantic mediation. The hypotheses concerning this
nonsemantic lexical route differ, depending on the authors: Goodman and Caramazza (1986a) postulated the existence of a
direct connection between the phonological input lexicon and the orthographic output lexicon (route a on Fig. 3.1), whereas
Patterson (1986) suggested that this route would pass via the phonological output lexicon (route b). For the time being, data
from the literature do not allow a choice between these alternative hypotheses.
It should also be mentioned that recent data collected with patients presenting an impairment of the orthographic buffer
suggested complementary hypotheses concerning the structure of orthographic representations. We present them in "The
Orthographic Buffer," the section concerning the disorders associated with this buffer.
The Assembling Procedure
The phonological route can also be altered at diverse loci. As reported previously, patient PR experienced difficulties in
segmenting auditorally presented pseudowords into isolated sounds and in converting isolated sounds into letters. These two
difficulties do not necessarily coexist. Several patients displaying the same type of deficit as PR seem to only suffer from an
inability to segment an acoustic continuum into isolated sounds without having difficulties in converting isolated sounds into
letters (Bolla-Wilson et al., 1985; Roeltgen et al., 1983). Moreover, a different impairment of this procedure was described by
Caramazza, Miceli, and Villa (1986). Their patient was affected by very specific spelling disorders. He produced errors on
pseudowords but not on words. These errors were almost all letter substitutions that were characterized by a high phonological
similarity (/d/-> /t/, /p/-> /b/). According to Caramazza and colleagues, this deficit was attributed to a dysfunction of the
phonological buffer that resulted in a deterioration of the representation that is stored in it. This phenomenon was also found in
reading and in repeating pseudowords, which would indicate that this buffer is also involved in these behaviors.
With the aim of specifying the functioning of the P-G conversion system, in-depth studies were performed with patients
presenting a quasi-abolition of the addressing procedure. The analysis of the productions of such a patient allowed Goodman
and Caramazza (1986b) and Goodman-Schulman and Caramazza (1987) to show evidence of two main phenomena: (a) In cases
when different graphemes (or sets of graphemes) correspond to the

Page 46
same phoneme (/i:/-> ea, ee, etc.), the P-G conversion system operates in a probabilistic manner by reproducing the frequency of
occurrence of the P-G correspondences in the language; (b) this component is able to select the adequate graphemic option as a
function of the within-syllable position of the grapheme as well as of the context in which the grapheme is placed. These
constraints allow the hypothesis that this module does not process each phoneme independently. Goodman and Caramazza
(1986a, 1986b) suggested that the processing unit is at least as large as a syllable. It should be noted that the research done on
this topic did not always result in such clear evidence (see Baxter & Warrington, 1987, 1988). These discrepancies could be
related to the use, in some patients, of residual capacities of the addressing procedure (Sanders & Caramazza, 1990).
The Orthographic Buffer
It was originally thought that the orthographic buffer coded two aspects: the graphemes' identity and their order. Given the
temporary nature of this stock, it was supposed that the trace held in this working memory underwent a relatively rapid decay.
The main factor associated with this level was word length: the longer the word, the longer the time necessary for preparing and
executing the movements tracing the letters.5
Caramazza, Miceli, Villa, and Romani (1987) defined the criteria characterizing a deficit of the orthographic buffer. Thus, an
impairment of this component should alter the production of words as well as pseudowords. Similarly, it should affect written
production whatever the response modality (writing, oral spelling, typing, etc.). These phenomena are related to the fact that the
orthographic buffer is common to both routes and to the various production modalities. Moreover, the factors relative to the two
procedures (lexical frequency, syntactical class, frequency of the P-G correspondences, etc.) should not affect the patient's
performance. The errors should be mainly letter substitutions, omissions, transpositions, and additions. Finally, an effect of word
length should be observed (more errors on long words than on short ones). Several patients presenting exactly this type of
pattern have been described in the literature. Furthermore, Hillis and Caramazza (1989) reported the existence of an interesting
double dissociation concerning two patients displaying a damage to the graphemic buffer: The first patient committed more
errors on the initial part of words, and the second one produced more of them on the final part of words. On this basis, these
authors suggest that, although graphemes are represented in this buffer in an abstract manner (that is to say in a code that is not
5 However, this simple principle is questioned by some authors who hypothesize a possible refreshing (or reactivation) of
this trace via one or the other routes (see Miceli, Silveri, & Caramazza, 1985; Nolan & Caramazza, 1983).

Page 47
physical or visual), their order is coded in a spatial structure liable to be neglected either on the right or on the left.
Several studies that offer more precise hypotheses on the representations contained in the orthographic buffer have been put
forward. Badecker, Hillis, and Caramazza (1990) studied the performance of an English-speaking patient (DH) displaying an
impairment of the orthographic buffer. DH is the patient mentioned earlier who produced most of his errors on the final part of
words. The authors submitted DH to writing and dictated oral spelling tasks of words possessing various morphemic structures.
These words either contained a single morpheme (e.g., ballad), two morphemes (root and affix, e.g., fixed, games) or were
compound words (newspaper, everything). The results showed, on the one hand, bimorphemic words of equal length generally
contained fewer errors than monomorphemic ones. On the other hand, the errors observed on bimorphemic words were usually
located at the end of morphemes and not at the end of words. This proved true only in cases when the composition rules of
morphemes were transparent (affixes called productive: e.g., motionless = motion + less). These results suggested that the unit
processed at this level was a morpheme rather than a word. For Badecker and collaborators, this also applied for the
representations contained in the orthographic output lexicon.
Moreover, Caramazza and Miceli (1990) analyzed the performance of an Italian-speaking patient (LB) presenting an
orthographic buffer deficit in the written production of a large corpus of six-letter words. Their analysis revealed the following
facts: (a) Words with a complex consonant vowel (CV) structure (e.g., scuola = CCVVCV) contain more errors than words with
a simple CV structure (e.g., tavolo = CVCVCV); (b) letter substitutions and transpositions by LB always conform to the rule "a
consonant for a consonant, a vowel for a vowel"; (c) the errors on groups of letters that form one or more sounds (e.g., in
Italian, the sequence SC can be pronounced either /sk/ or / /) are identical, indicating that the basic unit would be the
grapheme and not the phoneme; and (d) geminate letters are extremely resistant to errors. On the basis of these results, and by
analogy with current descriptions of phonological representations, the authors suggested that the orthographic representations
possess a multidimensional structure defined by four tiers: the graphemes' identity, their CV status, the grapho-syllabic
boundaries, and an indicator of geminate graphemes (see Fig. 3.2a). Very recently, McCloskey, Badecker, Goodman-Schulman,
and Aliminosa (1994) described an English-speaking patient with a lesion of the orthographic buffer (and most likely also of the
orthographic output lexicon) who produced, conversely to LB, more errors on words containing geminate letters than on words
that did not contain doubled letters (e.g., dress-> drees). In order to account for this patient's performance pattern, the authors
proposed a model of the structure of orthographic representations that is a bit

Page 48
Fig. 3.2.
Structure of the orthographic representation of the word
witness according to Caramazza and Miceli (1990) and
McCloskey et al. (1994).
different from the previous one. It is a multidimensional structure coding the following tiers (see Fig. 3.2b): the graphemes'
identity, their position (represented by X), their number, and their CV status. Consequently, McCloskey et al. did not make any
assumption on the representation of the grapho-syllabic boundaries and suggested that the position of the graphemes within the
word be explicitly defined, even for geminate letters. It is probably too early to try to choose between these two alternative
models, which are very similar to each other. Nevertheless, it should be noted that data collected by other researchers fit quite
well in one or the other of these frameworks, in particular the ones concerning the specification of the CV status (Cubelli, 1991)
and the special mode of representation dedicated to geminate letters (Venneri, Cubelli, & Caffarra, 1994).
Some Themes for Discussion

Phonological Route or Lexical Analogy?
The first questions about the dual-route model were proposed by Campbell (1983) who argued that normal subjects did not use a
simple phoneme to grapheme transcription in order to spell pseudowords. In her experiment, she submitted students to a
modified lexical decision task. The participants

Page 49
first had to decide if an auditorally presented stimulus was a word. If it was not a word, they had to orally spell the heard
nonword. Campbell's demonstration rests on the fact that when a word such as brain is presented before a pseudoword such as
/prein/, a majority of participants have a tendency to spell the pseudoword prain. Conversely, if the word crane is presented
shortly before, they spelled the same pseudoword prane. According to Campbell, this result indicates that in order to spell
pseudowords, participants used their knowledge concerning words and not a sound-to-letter conversion system. Pseudowords
are consequently spelled by analogy with words existing in the participants' orthographic lexicon.
Barry and Seymour (1988) criticized several methodological issues of this experiment and replicated it with some modifications
of the aspects that troubled them. The results that they obtained differ a bit from those of Campbell: They observed not only an
effect related to lexical influence but also an effect of the frequency of the sound-to-letter correspondences. There is no
interaction between these two effects. According to Barry and Seymour, these results are compatible neither with the model of
lexical analogy nor with the classical dual-route model. Therefore, they propose an alternative model, called interactive, which
is based on the existence of two routes functioning in parallel. The processing of any stimulus (familiar or not) simultaneously
involves the lexical route and the phonological route. The presentation of a word activates its representation in the orthographic
lexicon as well as the various P-G correspondences that characterize it in the P-G conversion system. This activation explains
why these correspondences are more likely to be selected in the subsequent generation of a pseudoword.
The notion of a parallel activation of both routes was also proposed by Hillis and Caramazza (1991) and by Miceli, Capasso, and
Caramazza (1994) under a different form. These authors suggest the existence of a connection between the P-G conversion
system and the orthographic output lexicon. Their hypothesis is that the activation and the selection of an orthographic
representation in the orthographic output lexicon is realized on the basis of the sum of the activations coming from both the
semantic system and the P-G conversion module. The influence of the assembling procedure is small in normal time.
Conversely, it could be determining when, for one reason or another, the information coming from the semantic system is
decayed or delayed. The corollary of this proposition is that semantic errors specific to the written output modality should be
observed only in cases of joint damage to the addressing procedure and to the assembling procedure. This prediction seems to be
confirmed by an examination of the published cases of patients committing semantic errors specific to written production: All of
them presented difficulties affecting the sound to letter conversion.

Page 50
Are Orthographic Input and Output Lexicons Independent?
In the cognitive models of reading, the existence of an orthographic input lexicon allowing recognition of visually presented
familiar words is usually postulated. The independence of this lexicon with regard to the orthographic-output lexicon (used in
spelling) is discussed in the normal subject as well as in the brain-damaged patient. Experimental results obtained with normal
subjects did not until now provide a clear answer to this question. Indeed, Monsell and Banish (cited in Monsell, 1987) found
that the writing of words, without seeing them, did not facilitate performance by the participants in a visual, lexical-decision
task performed later. This suggests that orthographic representations used for input and for output are independent. On the other
hand, in a second experiment, the authors observed a facilitative effect of writing a dictated word on the ability of the
participants to match it with a visual definition and to recognize it in a visual lexical decision task. These results suggest that
having participants produce an orthographic form of a word facilitated its recognition only when it could be viewed. In any case,
whether a single orthographic representation serves both input and output becomes a complex question in view of such data.
Data coming from pathology have not necessarily settled this question. Some authors plead in favor of a single orthographic
lexicon (Allport & Funnell, 1981; Coltheart & Funnell, 1987). From this point of view, all the representations are stored in a
central orthographic lexicon and the errors observed in reading and/or in writing are associated with an alteration of the
functionally distinct procedures of input and output. The main argument evoked by these authors is the frequent association
between surface dyslexia and dysgraphia.
However, authors favoring the hypothesis of independent lexicons cite the existence of dissociations between reading and
writing skills in the same patient (Baxter & Warrington, 1985; Beauvois & Drouesn, 1981). For instance, patient RG,
mentioned earlier in this chapter, was presenting, in addition to a surface dysgraphia, a deep dyslexia. Among all the
dissociations registered in the literature, it seems that surface dysgraphias can appear in the absence of a surface dyslexia,
whereas the opposite pattern does not appear to have clearly been observed so far.
There are, however, two studies (Howard & Franklin, 1988; Kremin, 1989) that reported the cases of patients presenting a
surface dyslexia associated with a deep dysgraphia, a dissociation that, at first sight, seems to indicate that the orthographic
representations can be preserved for output but not for input. However, it seems difficult for us to support the hypothesis of the
independence of orthographic lexicons for input and for output on the basis of this dissociation, for two reasons. First, the
presence of a surface dyslexia does not necessarily imply the presence of a deficit at the level of the

Page 51
orthographic representations; the patient can indeed be led to use the assembling procedure if he or she experiences difficulties
in accessing the output phonological lexicon, which would leave the orthographic representations intact. Second, deep
dysgraphia seems a bad candidate for dissociation given the fact that in this type of disorder, there is not only an important
alteration (of variable location and degree) of the assembling procedure but also additional deficits of the addressing procedure,
in particular at the level of the output orthographic representations. Thus, the alteration of a unique orthographic system could
very well be expressed by regularization errors in reading because the assembling procedure can, up to a certain extent,
compensate for a deficit in accessing orthographic representations and, at the same time, not appear in writing. Indeed, in deep
dysgraphia, the alteration of orthographic representations can lead to the production of nonresponses or to the production of
semantic or visual paragraphias because the assembling procedure is unable to compensate for the deficits of the addressing
procedure. Thus, the only satisfying demonstration of the independence of the two lexicons appears to result from the report of a
surface dyslexia (involving the level of orthographic representations) that either is not associated with a spelling disorder or is
associated with a phonological dysgraphia in which the spelling of words is almost entirely correct.
For the supporters of a unique orthographic lexicon, the sense of this dissociation is compatible with the hypothesis that they
defend. Indeed, the association of a surface dyslexia and of a surface dysgraphia depends, according to them, on the severity of
the deficit at the level of the unique orthographic lexicon: The more severe the deficit, the more it impairs the activation of
orthographic representations, simultaneously affecting writing and reading. Contrary to this, if the deficit is moderate, it could
appear only in writing because the activation level required for matching a visually presented item with its stored representation
(recognition) is less important than the activation level required for producing a representation in the absence of an external
sensory input (recall). The orthographic predictability is then higher in the reading task than in the writing task and accounts for
the fact that a deficit in a unique orthographic lexicon has more impact on writing than on reading, with a larger number of
words enabling efficient processing by an assembling procedure.
Acquired and Developmental Spelling Disorders
Besides acquired disorders that have been discussed to this point, developmental disorders have also been studied in the domain
of spelling (Campbell & Butterworth, 1985; Seymour, 1987; Temple, 1986). Most researchers interested in this theme use a
double theoretical framework: functional models, described in this chapter, and specific models of reading and spelling

Page 52
acquisition. These are presented in several chapters in this volume (see Treiman & Cassar, chap. 4; Ehri, chap. 13; Ellis, chap.
14; and Rieben and Saada-Robert, chap. 15). We recall that the supporters of these models hypothesize that the progressive
mastery of reading and spelling involves several phases characterized by the use of different strategies. For instance, Frith
(1985) assumed the existence of three successive strategies: a logographic strategy based on a global, visual representation of
the word; an alphabetic strategy involving a process of phonological mediation (which can be set in parallel with the assembling
procedure); and an orthographic strategy based on a morphemic representation of words (addressing procedure). There are
divergences between authors concerning whether all strategies are used in spelling (the logographic strategy in particular could
be specific to reading), whether these strategies are elaborated and used successively or simultaneously, and whether alternative
strategies (such as the strategy of lexical analogy) can be identified.
The literature concerning developmental spelling disorders indicates that some features described in brain-lesioned adults can
also be found in learning disabled children. Thus, the performance of several participants, reported by Temple (1986) and by
Seymour (1987), is characterized by the production of phonologically plausible errors and by an advantage of regular words over
irregular words. This type of deficit evokes the surface dysgraphia described in adults. In the framework of developmental
models, this disorder has generally been interpreted in terms of use of an alphabetic strategy related to the child's inability to
move on to an orthographic strategy. The opposite pattern has also been reported under the same name for adults (phonological
dysgraphia). This disorder is characterized by considerable difficulties in spelling pseudowords and by the production of mostly
nonphonologically plausible errors (Campbell & Butterworth, 1985; Seymour, 1987; Temple, 1986). According to these authors,
these participants are not able to elaborate an adequate alphabetic strategy. Those whose difficulties are the most severe use a
logographic strategy, whereas those presenting more limited deficits use an orthographic strategy.
Beyond these very general considerations, it seems that a more detailed examination shows evidence of numerous divergences
between acquired and developmental disorders. It should be noted that the deficits reported in disabled children are rarely as
selective as those described in brain-damaged adults even if participants reported in the literature are not representative of the
daily clinical practice. As mentioned by Seymour (1987), it is necessary with children to distinguish the primary source of the
deficit from its effects on later development. The presence of a disrupted process can be a handicap, but it is also liable to stop
or to modify the development of other processes that rely on incomplete or incorrect information generated by this component.
Besides, current data suggest that the child ("normal"

Page 53
as well as disabled) uses more than one strategy at a time each strategy being more or less functional. This phenomenon could
also explain part of the differences observed between acquired and developmental disorders.
Future studies should aim at going beyond these general analogies to supply detailed indications on the constitution of the
processes responsible for spelling and on the qualitative and quantitative changes occurring throughout the process of
acquisition.
Spelling and Connectionism
A last point of discussion concerns the connectionist approach, regularly presented as an alternative of the classical cognitive or
symbolic models (Rumelhart & McClelland, 1986). Several authors developed models that simulate various aspects of human
performance, both normal and pathological, in the field of spelling. For instance, two recent studies have been dedicated to the
capacities of neural networks, possessing neither a lexicon nor a set of explicit P-G conversion rules, to generate sequences of
graphemes in response to sequences of phonemes (Brown & Loosemore, 1994; Olson & Caramazza, 1994). The architecture of
these models is formed by three layers of units: an input layer representing the phonological form of words, an output layer
coding the orthographic forms, and a hidden layer functioning as an aid in the establishment of input and output
correspondences. A learning algorithm enables the network to modify the weight of the connections between units so as to
progressively increase the probability of generating correct responses.
The various implementations of these types of models proved able, after a period of learning during which they were presented
with a corpus of words, to produce the correct spelling of these words, both regular and irregular, with a high percentage of
success. Furthermore, when the spellings generated by the network were not accurate, the networks usually proposed
orthographic forms that were phonologically plausible (e.g., syntax-> sintax), thus simulating the behavior of a human
participant. Besides, these networks also proved liable to generalize their knowledge to words that had not been previously
presented. Finally, lesions of these networks (performed either by introducing sources of noise in the weight of the connections
after learning or by limiting the number of hidden units before learning) resulted in performance profiles that were similar to
certain deficits described in brain-damaged patients (Olson & Caramazza, 1994) or in dysgraphic children (Brown &
Loosemore, 1994). In particular, a dissociation between performance on regular words (relatively preserved) and performance on
irregular words (affected) was observed.
Houghton, Glasspool, and Shallice (1994) proposed a connectionist model simulating the phenomena associated with the serial
order of spelling at-

Page 54
tributed, in symbolic models, to the existence of an orthographic buffer. These authors showed that it was possible to account
for some aspects of normal and pathological performance by a three-layer model coding the word, the graphemes, and a
competitive filter enabling the successive activation of the different graphemes forming the word.
The capacity of connectionist networks to learn to spell regular and irregular words and, to a certain extent, to generalize their
knowledge to new words constitute an important challenge for the models based on the existence of multiple routes involving
separate processes. The connectionist approach is not without problems however. In particular, as noted by Olson and
Caramazza (1994), the performance of the networks appears to be highly related to the mode of representation chosen.
Moreover, "lesioning" these networks results in the emergence of only certain aspects of the deficits observed in brain-damaged
patients. For the time being, it cannot be claimed that the other characteristics and the other patterns described in the preceding
pages will emerge from "lesions" made to these networks. The dialogue between symbolic and subsymbolic models in the field
of spelling is now open. As suggested by some authors, these two types of models are not necessarily incompatible, and the
dynamic that is now established between the supporters of the two approaches will certainly be beneficial to the advancement of
concepts of spelling in the domain of neuropsychology and, more generally, of psychology.
Conclusion
This brief examination of the studies done with a cognitive perspective in the neuropsychology of spelling allows a first
evaluation. Within less than two decades, researchers have managed to elaborate consistent, empirically testable, models that
present the advantage of integrating a considerable number of observations and experimental facts. As we have attempted to
show, these models are still being elaborated and, are, therefore, liable to be modified and enriched at any moment. It is obvious
that much research is still necessary in order to define more precisely the structure and the functioning of each of the
components as well as their connections.
Nevertheless, these models begin to have an impact on clinical neuropsychological practice, in particular on the evaluation and
remediation of spelling disorders. A methodology of clinical evaluation in which the theoretical models are used by clinicians to
make inferences about the functional origin of individual, pathological behaviors is developing. A number of recent studies
(Behrmann, 1987; de Partz, Seron, & Van der Linden, 1992; Hatfield, 1982) showed that a cognitive interpretation can be used
in developing a remediation goal in two ways: By specifying the impaired processes

Page 55
to which the training should apply (on the one hand), and by specifying the preserved components that can (possibly?)
compensate for the impaired components (on the other hand).
References
Allport, D.A., & Funnell, E. (1981). Components of the mental lexicon. Philosophical Transactions of the Royal Society
(London), B295, 397-410.
Assal, G., Buttet, J., & Jolivet, R. (1981). Dissociations in aphasia: A case report. Brain and Language, 13, 223-240.
Badecker, W., Hillis, A., & Caramazza, A. (1990). Lexical morphology and its role in the writing process: Evidence from a case
of acquired dysgraphia. Cognition, 35, 205-243.
Barry, C., & Seymour, P.H.K. (1988). Lexical priming and sound-to-spelling contingency effects in nonword spelling. The
Quarterly Journal of Experimental Psychology, 40A(1), 5-40.
Basso, A., Taborelli, A., & Vignolo, L.A. (1978). Dissociated disorders of speaking and writing in aphasia. Journal of
Neurology, Neurosurgery, and Psychiatry, 41, 556-563.
Baxter, D.M., & Warrington, E.K. (1985). Category specific phonological dysgraphia. Neuropsychologia, 23, 653-666.
Baxter, D.M., & Warrington, E.K. (1987). Transcoding sound to spelling: Single or multiple sound unit correspondence? Cortex,
23, 11-28.
Baxter, D.M., & Warrington, E.K. (1988). The case for biphoneme processing: A rejoinder to Goodman-Schulman. Cortex, 24,
137-142.
Beauvois, M.F., & Drouesn, J. (1981). Lexical or orthographic agraphia. Brain, 104, 21-49.
Behrmann, M. (1987). The rites of righting: Homophone remediation in acquired dysgraphia. Cognitive Neuropsychology, 9,
209-251.
Benson, D.F., & Cummings, J.L. (1985). Agraphia. In J.A.M. Frederiks (Ed.), Handbook of clinical neurology, 1(45), 457-472.
Clinical neuropsychology. Amsterdam: Elsevier.
Bolla-Wilson, K., Speedie, L.J., & Robinson, R.G. (1985). Phonological agraphia in a left-handed patient after a right
hemisphere lesion. Neurology, 35, 1778-1781.
Brown, G.D.A., & Loosemore, R.P.W. (1994). Computational approaches to normal and impaired spelling. In G.D.A. Brown &
N.C. Ellis (Eds.), Handbook of spelling: Theory, process, and intervention (pp. 319-335). Chichester, UK: Wiley.
Bub, D., & Kertesz, A. (1982). Deep agraphia. Brain and Language, 17, 146-165.
Campbell, R. (1983). Writing nonwords to dictation. Brain and Language, 19, 153-178.
Campbell, R., & Butterworth, B. (1985). Phonological dyslexia and dysgraphia in a highly literate subject: A developmental
case with associated deficits of phonemic processing and awareness. The Quarterly Journal of Experimental Psychology, 37A,
435-475.
Caramazza, A., & Hillis, A.E. (1991). Lexical organization of nouns and verbs in the brain. Nature, 349, 788-790.
Caramazza, A., & Miceli, G. (1990). The structure of graphemic representations. Cognition, 37, 243-297.
Caramazza, A., Miceli, G., & Villa, G. (1986). The role of the (output) phonological buffer in reading, writing, and repetition.
Cognitive Neuropsychology, 3, 37-76.
Caramazza, A., Miceli, G., Villa, G., & Romani, C. (1987). The role of the graphemic buffer in spelling: Evidence from a case
of acquired dysgraphia. Cognition, 26, 59-85.
Coltheart, M., & Funnell, E. (1987). Reading and writing: One lexicon or two? In D.A. Allport, D.G. MacKay, W. Prinz, & E.
Scheerer (Eds.), Language perception and production: Relationships between listening, reading and writing. London: Academic
Press.

Page 56
Cubelli, R. (1991). A selective deficit for writing vowels in acquired dysgraphia. Nature, 353, 258-260.
de Partz, M.-P., Seron, X., & Van der Linden, M. (1992). Re-education of a surface dysgraphia with a visual imagery strategy.
Cognitive Neuropsychology, 9, 369-401.
Ellis, A.W. (1982). Spelling and writing (and reading and speaking). In A.W. Ellis (Ed.), Normality and pathology in cognitive
functions (pp. 113-146). London: Academic Press.
Ellis, A.W. (1984). Reading, writing and dyslexia: A cognitive analysis. Hove & London: Lawrence Erlbaum Associates.
Exner, S. (1881). Untersuchung ber die Lokalisation der Funktionen in der Grosshirnrinde des Menschen [Investigations on
the localization of functions in the cerebral cortex of humans]. Vienna: Wilhein Braumller.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K.E. Patterson, J.C. Marshall, & M. Coltheart (Eds.),
Surface dyslexia: Neuropsychological and cognitive studies of phonological reading (pp. 301-330). Hove & London: Lawrence
Erlbaum Associates.
Goodman, R.A., & Caramazza, A. (1986a). Aspects of the spelling process: Evidence from a case of acquired dysgraphia.
Language and Cognitive Processes, 1(4), 263-296.
Goodman, R.A., & Caramazza, A. (1986b). Phonologically plausible errors: Implications for a model of the phoneme-grapheme
conversion mechanism in the spelling process. In G. Augst (Ed.), New trends in graphemics and orthography (pp. 300-325).
Berlin: Walter de Gruyter.
Goodman-Schulman, R.A., & Caramazza, A. (1987). Patterns of dysgraphia and the nonlexical spelling process. Cortex, 23,
143-148.
Hatfield, F.M. (1982). Diverses formes de dsintgration du langage crit et implications pour la rducation [Various forms of
written language and implications for remediation]. In X. Seron & C. Laterre (Eds.), Brain Remediation (pp. 135-154).
Bruxelles: Mardaga.
Hatfield, F.M. (1985). Visual and phonological factors in acquired dysgraphia. Neuropsychologia, 23(1), 13-29.
Hatfield, F.M., & Patterson, K.E. (1983). Phonological spelling. Quarterly Journal of Experimental Psychology, 35A, 451-468.
Hcaen, H., Angelergues, R., & Douzenis, J.A. (1963). Les agraphies [The agraphias]. Neuropsychologia, 1, 179-208.
Hillis, A.E., & Caramazza, A. (1989). The graphemic buffer and attentional mechanisms. Brain and Language, 36, 208-235.
Hillis, A.E., & Caramazza, A. (1991). Mechanisms for accessing lexical representations for output: Evidence from a category-
specific semantic deficit. Brain and Language, 40, 106-144.
Houghton, G., Glasspool, D.W., & Shallice, T. (1994). Spelling and serial recall: Insights from a competitive queueing model.
In G.D.A. Brown, & N.C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 365-404). Chichester, UK:
Wiley.
Howard, D., & Franklin, S. (1988). Missing the meaning. Cambridge, MA: MIT Press.
Kremin, H. (1989). Case study of a patient with a surface dyslexia and deep dysgraphia with a special attention to the
noun/verb distinction. International Conference on Cognitive Neuropsychology. Harrowgate, England.
Marcie, P., & Hcaen, H. (1979). Agraphia: Writing disorders associated with unilateral cortical lesions. In K.M. Heilman, & E.
Valenstein (Eds.), Clinical Neuropsychology (pp. 92-127). New York: Oxford University Press.
Margolin, D. I. (1984). The neuropsychology of writing and spelling: Semantic, phonological, motor, and perceptual processes.
The Quarterly Journal of Experimental Psychology, 36A, 459-489.
Marshall, J.C., & Newcombe, F. (1973). Pattern of paralexia: A psycholinguistic approach. Journal of Psycholinguistic
Research, 2, 175-199.
McCloskey, M., Badecker, W., Goodman-Schulman, R.A., & Aliminosa, D. (1994). The structure of graphemic representations
in spelling: Evidence from a case of acquired dysgraphia. Cognitive Neuropsychology, 11, 341-392.

Page 57
Miceli, G., Capasso, R., & Caramazza, A. (1994). The interaction of lexical and sublexical processes in reading, writing, and
repetition. Neuropsychologia, 32, 317-333.
Miceli, G., Silveri, M.C., & Caramazza, A. (1985). Cognitive analysis of a case of pure dysgraphia. Brain and Language, 25,
187-212.
Monsell, S. (1987). Nonvisual orthographic processing and the orthographic input lexicon. In M. Coltheart (Ed.), Attention and
Performance XII: The psychology of reading (pp. 299-323). London: Lawrence Erlbaum Associates.
Morton, J. (1980). The logogen model and orthographic structure. In U. Frith (Ed.), Cognitive processes in spelling (pp. 117-
133). London: Academic Press.
Nolan, K.A., & Caramazza, A. (1983). An analysis of writing in a case of deep dyslexia. Brain and Language, 20, 305-328.
Olson, A., & Caramazza, A. (1994). Representation and connectionist models: The NETspell experience. In G.D.A. Brown, &
N.C. Ellis (Eds.), Handbook of spelling: Theory, process, and intervention (pp. 337-363). Chichester, UK: Wiley.
Patterson, K.E. (1986). Lexical but nonsemantic spelling? Cognitive Neuropsychology, 3, 341-367.
Patterson, K.E. (1988). Acquired disorders of spelling. In G. Denes, C. Semenza, & P. Bisiacchi (Eds.), Perpectives on cognitive
neuropsychology (pp. 213-229). Hove, UK: Lawrence Erlbaum Associates.
Roeltgen, D.P., & Heilman, K.M. (1984). Lexical agraphia: Further support for the two-system hypothesis of linguistic agraphia.
Brain, 107, 811-827.
Roeltgen, D.P., Sevush, S., & Heilman, K.M. (1983). Phonological agraphia: Writing by the lexical-semantic route. Neurology,
33, 755-765.
Rumelhart, D.E., & McClelland, J.L. (1986). Parallel distributed processing. Cambridge, MA: MIT Press.
Sanders, R.J., & Caramazza, A. (1990). Operation of the phoneme-to-grapheme conversion mechanism in a brain-injured
patient. Reading and Writing, 2, 61-82.
Seymour, P.H.K. (1987). Developmental dyslexia: A cognitive experimental analysis. In M. Coltheart, G. Sartori, & R. Job
(Eds.), The cognitive neuropsychology of language (pp. 357-395). London: Lawrence Erlbaum Associates.
Shallice, T. (1981). Phonological agraphia and the lexical route in writing. Brain, 104, 413-429.
Temple, C.M. (1986). Developmental dysgraphias. The Quarterly Journal of Experimental Psychology, 38A, 77-110.
Venneri, A., Cubelli, R., & Caffarra, P. (1994). Perseverative dysgraphia: A selective disorder in writing double letters.
Neuropsychologia, 32, 923-931.


Page 59
PART II
THE ACQUISITION OF SPELLING


Page 61
Chapter 4
Spelling Acquisition in English
Rebecca Treiman
Marie Cassar
Wayne State University
Having been given the opportunity to write a short chapter about spelling acquisition in English, we are pleased to report that
this task is now near to impossible. The reason we are happy is that, in the past, the field of spelling suffered from a dearth of
research. There was much more work on the acquisition of reading and word recognition than on the acquisition of writing and
spelling. However, the body of research on spelling in English-speaking children is now so large that it would be foolish to
presume to do it justice within a single chapter. Given this limitation, we chose three main topics for discussion. These are the
use of letter names as an entry into the writing system, young children's knowledge of the orthographic patterns of their
language, and children's use of morphological strategies to guide their spelling. We chose these three topics because each plays
a major role in existing theories of spelling development. We review experimental evidence to suggest that young spellers have
more knowledge in each of these areas than portrayed by existing theories. Children's knowledge, we argue, provides a
foundation for the development of spelling skills.
To accomplish our goal, it is first necessary to review what the existing theories have to say about spelling development.
Current research on the topics of letter-name spellings, knowledge of orthography, and morphological knowledge is then
discussed. In the final section, we address the implications of the findings for views of spelling development.

Page 62
Theories of Spelling Development
Henderson (1985) outlined five stages of spelling development. The first stage, preliterate writing, begins almost as soon as a
child starts to make marks with a crayon or pencil. A child's ''writing'' at this stage may resemble scribbles more than writing.
The child understands that writing is different from drawing but not that writing represents speech. When children do begin to
realize that writing is a way to communicate speech and that letters symbolize the sounds in words, they are considered to be
entering the second stage of spelling development. This second stage is called letter-name spelling. Gradually, the child has
become aware of sounds and of the letters used to represent them. The child has also learned the names of the letters of the
alphabet. Often a child will write a letter to represent the sounds of the letter's name. For example, the words mess and help may
be spelled MS and HLP, respectively, with the letters s and l standing for their names, /es/ and/el/. (Throughout this chapter, we
use capital letters to refer to children's spellings of words or nonwords.) Children do not represent in their spellings sounds that
they have difficulty accessing as separate units, such as the /m/ and /n/ in the words bump and want. For the child, the sound of
the vowel plus the /m/ or /n/ appears to be one. Because children do not yet know all of the conventions of the writing system,
their spellings contain these apparently odd errors. Henderson stated that children use a phonetic or sound-based strategy to
spell words that they do not know throughout the first and second spelling stages and into the third stage, called the within-word
pattern stage.
By the time children reach the within-word pattern stage, they have begun to learn words by sight through their reading.
Basically, the term sight words refers to an acquired pool of memorized spellings. According to Henderson (1985), children
usually spell sight words correctly once they have acquired these words. Not until a considerable number of sight words are
learned do they begin to influence children's spellings of novel words. As children's writing starts to change as a result of this
influence, children are considered to be entering the within-word pattern stage. This stage is marked by the correct spelling of
short vowels, the use of silent markers such as final e for long vowels, and the correct spelling of clusters such as st, dr, and tr.
The letters m and n that were previously missing in words such as bump and want now appear. Children's knowledge of sight
words helps them to realize that writing does not always involve a one-to-one matching of sound to letters. Children begin to
use frequent letter patterns that correspond to sequences of sounds, for example the -ight pattern found in light, might, and fight.
After children have begun to learn patterns in conventional spelling and start to use them in their own spellings, they become
able to understand

Page 63
how meaning relations among words are marked in print. Before this, according to Henderson (1985), children's spelling is
based on sounds. For instance, letter-name spellers represent the past tense ending -ed with either d, t or id according to its
sound, as in PED for pinned, STAPT for stepped, and PLATID for planted. During the initial part of the within-word period,
children still choose different patterns depending on the morpheme's phonetic form. Later, children come to understand that the
past tense ending has a consistent spelling. "From the third grade on, . . . the role of meaning becomes rapidly more
conspicuous" (Henderson, p. 63).
The use of double consonants to mark a short vowel characterizes Henderson's (1985) fourth stage, called the syllable juncture
stage. Children who are making rapid progress in learning to spell may reach this point during the middle years of elementary
school; other children do not reach this stage until later. Children now learn that the presence versus the absence of a double
consonant marks the difference between the first vowels of words such as little and title. Other learning at this stage includes the
understanding of when and when not to double consonants with the addition of suffixes such as -ed and -ing.
The last stage in Henderson's (1985) outline of spelling development is called derivational principles. This stage may begin as
early as the late elementary grades for children who are progressing quickly; it begins later for other children and continues
throughout a writer's lifetime. As the title of this stage implies, the spelling relations among words in terms of roots, origins, and
meanings are further explored. Knowledge of the meaning connections among words such as confide, confident, and confidential
aids the speller. These words have similar spelling patterns even though their sounds vary slightly.
A very similar sequence for spelling development has been outlined by Gentry (1982), who analyzed the spellings of a
precocious youngster who began to write at home without formal instruction (Bissex, 1980). Gentry's depiction begins with the
precommunicative stage, when the child has no knowledge of letter-sound correspondence. During this stage, spellings are
random letter strings that may include numbers or letter-like marks. Next, the child begins to understand that letters are used to
represent sounds. However, spellings start out as incomplete phonetic representations of words. The letter-name strategy
mentioned earlier is the hallmark of this second stage, which Gentry called the semiphonetic stage. "Where possible the speller
represents words, sounds, or syllables with letters that match their letter names (e.g., R [are]; U [you]; LEFT [elephant]) instead
of representing the vowel and consonant sounds separately" (p. 194). Once children can symbolize the entire sound structure of
words in their spellings, they are considered to have reached the phonetic stage. All of the surface sound features of words are
now represented in spellings. However, "letters are assigned

Page 64
strictly on the basis of sound, without regard for acceptable English letter sequence or other conventions of English
orthography" (p. 195). The fourth stage, called the transitional stage, starts when children become more aware of conventional
spellings. Children now employ visual and morphological information in their spelling rather than depend wholly on sound, as
they did earlier. Gentry's last stage is simply called the correct stage. Knowledge of the orthographic system and of basic rules
is firm. A writer will continue to master uncommon and irregular forms and so errors will occur, but the majority of spellings
will be correct.
Ehri (1986) outlined three stages describing the development of orthographic knowledge: the semiphonetic, phonetic, and
morphemic stages. The characteristics of these stages are very similar to the semiphonetic, phonetic, and transitional stages
described by Gentry (1982). Ehri did not include the period before actual alphabetic writing as the initial stage of spelling
development, in contrast to Gentry's first precommunicative stage. Although Ehri acknowledged that very young children have
some primitive notions about writing, she preferred to begin her description at the point at which children are attempting to
write words. Another difference between the stages outlined by Ehri and by Gentry is that Ehri did not include a final stage
characterized by correct spelling. She argued that a number of regularities are learned during the morphemic stage, leading to the
correct spellings of many words. This learning process is part of the morphemic stage, which continues throughout the writer's
lifetime.
These three descriptions of spelling development (Ehri, 1986; Gentry, 1982; Henderson, 1985) share some important
characteristics. All depict beginning spellers as focused on representing speech sounds in their spellings. Common phoneme
grapheme mappings and letter names are believed to guide a child's use of letters and letter sequences at this time. Knowledge
about the variety of letter sequences that are possible in English and about when each one should be used develops only after
children have established a corpus of sight words. The understanding of meaning relations among words and the use of this
knowledge to guide spelling is also a late development. Thus, the theories portray spelling development as a sequence of stages
that involve the ability to use different types of information. Young spellers are limited to a few sources of information, namely
sounds and letter names. As children progress through the stages, they gain the ability to use other types of information,
including orthographic regularities and morphological relations among words.
We argue that the existing theories underestimate young children's abilities. Although phonology and letter names play
important roles in early spelling, young children also have some budding knowledge about the orthographic regularities of the
written language and about the role of morphology in spelling. Children begin to use this knowledge earlier than they

Page 65
are given credit for in the existing theories. In addition, we argue that theories of spelling development need to pay more
attention to the interactions among various sources of knowledge. For example, we suggest that children's tendency to make use
of letter names is affected by the phonological structure of the letter names themselves. To exemplify these notions, let us begin
by examining how children use their knowledge of letter names to guide their spellings.
Letter Names and Spelling

Middle-class North American children typically learn to recite the alphabet (often by singing the alphabet song) and to name
many of the letters well before they begin formal schooling. Parents, children's television programs, and preschools are just
some of the sources from which children acquire this information. In one study (Mason, 1980), almost two thirds of 4-year-olds
were said by their parents to "very often" recite the alphabet without error. Over half of the children were said to recognize more
than 20 letters of the alphabet by name. In another study, which assessed acquisition of alphabet skills directly rather than rely
on parental report, children could recite or sing five or more letters of the alphabet by age 4 and were almost perfect by age 5
(Worden & Boettcher, 1990). Shown uppercase letters and asked to name them, 4-year-olds were correct on about 14 of the 26
letters. Five-year-olds were correct on about 22 letters. Children's knowledge of the sounds of letters was not as good as their
knowledge of the names. Four-year-olds could provide the sounds for about 6 letters of the 26 and 5-year-olds for 8.
Given that children enter school with a good deal of knowledge about letter names, what role, if any, does this knowledge play
in learning to write and read? One might argue that letter-name knowledge cannot directly benefit spelling or reading because
written letters symbolize phonemes and not the letters' names. After all, bat is pronounced /bt/ rather than /biti/. However,
most English letter names contain the phoneme that is commonly represented by the letter. For instance, b has the name /bi/;
this letter typically symbolizes the phoneme /b/, which is the first sound of its name. Durrell (1980), Ehri (1983, 1986), and
others argued that letter-name knowledge plays an important role in the early acquisition of literacy. In fact, the theories of
spelling development that we have discussed depict beginners as relying so heavily on a letter-name strategy that they spell a
phoneme or sequence of phonemes that matches a letter name with the corresponding letter whenever it is possible to do so
(Ehri, 1986; Gentry, 1982; Henderson, 1985). However, recent evidence suggests that such letter-name spellings may not be as
widespread as previously thought. Children's tendency

Page 66
to use letter names as a guide to spelling appears to be influenced by the phonological properties of the letter names themselves.
Evidence to support this point comes from a series of experiments conducted by Treiman (1994). In the first experiment, first
graders were asked to spell a series of monosyllables, each of which contained a sequence of phonemes that matched the name
of an English consonant letter. Examples are /var/, which contains the name of the letter r, /vel/, which contains the name of the
letter l, and /pem/ and /kef/, which contain the names of m and f. If children use a letter-name spelling strategy whenever it is
possible to do so, as the theories of spelling development claim, they should spell these nonwords as VR, VL, PM, and KF. The
first graders, who were tested in October and November, did make some of these consonant-consonant (CC) errors. However,
the errors were by far most frequent for nonwords containing the name of r. For nonwords, such as /var/, 41% of the children's
spellings were letter-name errors such as VR. The letter-name errors occurred less often for l, with a rate of 9% errors such as
VL for /vel/. Vowel omissions in spellings of nonwords containing the names of m, n, f, and s were even less common,
occurring between 2% and 4% of the time. It appears that some vowel-consonant (VC) letter-name sequences are more likely to
be spelled as units than others.
Treiman's (1994) second experiment was designed to investigate these differences further. In addition to examining letters with
VC names, the second study also included letters with consonant-vowel (CV) names. Children in preschool, kindergarten, and
the very beginning of first grade were tested; these children were screened to ensure that they knew the names of the critical
letters. The children were asked to spell syllables containing phoneme sequences that matched the names of English letters. The
letters considered were r, l, m, n, f, s, t, p, and k. For example, /gar/ contains the letter name for r, or /ar/, /zef/ contains the
letter name for f, or /ef/, and /tib/ contains the letter name for t, or /ti/. The kindergartners and first graders produced most CC
spellings for syllables containing the name of r. Indeed, kindergartners made errors such as GR for /gar/ 61% of the time, and
first graders did so 50% of the time. Syllables containing the letter name l were the next most likely to be spelled as CCs, with
41% such errors for kindergartners and 19% for first graders. CC spellings were less common for syllables that contained the
names of the letters m, n, f, s, t, p, and k. However, children did produce more CC spellings for these syllables than for control
syllables that did not contain letter-name sequences. Thus, although the kindergartners' and first graders' spellings were most
influenced by their knowledge of letter names for r and, to a lesser extent, l, there was some effect for a broad range of other
letters. (The preschoolers showed a different pattern of performance than the older children; their results are discussed further
on.)

Page 67
Two additional experiments reported by Treiman (1994), as well as data on first graders' spellings of real words reported by
Treiman (1993), provide further evidence that letter-name spellings are not equally likely for all consonants. These errors are
most common for r, next most common for l, and less frequent for other consonant letters. These differences speak against the
idea that children use a letter-name spelling strategy whenever it is possible to do so.
Why are children more likely to spell /ar/ as r than to spell /el/ as l or /ti/ as t? The answer to this question, Treiman (1993,
1994) suggested, may lie in the phonological structure of the letters' names. In particular, some letter names may be easier for
children to segment into phonemes than others. Children have particular difficulty dividing the sequence /ar/ into phonemes and
so tend to spell this sequence as a single unit. The difficulty with /ar/ may reflect, in part, the special properties of postvocalic
/r/. Evidence from speech errors and from the learning of games that divide spoken syllables in various ways suggests that a
vowel and a following /r/ form an especially tight bond, stronger than that between a vowel and a following /l/ or between a
vowel and another following consonant (Derwing & Nearey, 1991; Laubstein, 1987; but see Stemberger, 1983). Thus, based on
the properties of the consonants in the VC letter names, the letter-name sequence /ar/ may be more cohesive than /el/. Both of
these in turn may be more cohesive than sequences such as /es/ (Derwing & Nearey, 1991; Hindson & Byrne, 1997; Treiman,
1984; Treiman, Zukowski, & Richmond-Welty, 1995). A child attempting to spell the spoken syllable /gar/ may analyze it into
/g/ and /ar/ and, being unable to segment the /ar/ further, may spell the syllable as GR.
So far, we have focused on the special properties of postvocalic /r/ that may make it hard for children to divide the letter name
/ar/ into phonemes. However, the nature of the vowel may be important as well. The letter r is unusual in that the vowel
phoneme in its name, /a/, occurs in no other English letter name. The lack of letter names such as /ap/ or /ag/ leaves children
without a mate to compare with /ar/. In contrast, the repetition of vowel phonemes across other consonant letter names may help
children to understand that the letters are used to represent the distinctive consonant phonemes rather than the repeated vowels.
Thus, a child may come to realize that the letter names /bi/, /di/, /pi/, /ti/, and so on, share the vowel /i/ and that the letters b, d,
p, and t symbolize the consonants that distinguish these letter names. The same may be true for the VC letter names /el/, /ef/,
/em/, /en/, and /es/. How much of a contribution the uniqueness of the vowel in /ar/ among the letter names of English makes to
children's use of r letter-name spellings remains to be investigated.
We have seen that first graders and kindergartners sometimes symbolize a difficult-to-segment sequence such as /ar/ or /el/ with
the letter that has

Page 68
this name, producing errors such as CR for car and HLP for help. Additional evidence points to letter-name effects in even
younger children. Consider the results for the preschoolers in the second experiment by Treiman (1994), in which children
attempted to spell syllables such as /gar/, /zef/ and /tib/. A number of the preschoolers originally screened did not know the
names of all of the critical consonant letters and so did not participate in the spelling test. Those who knew the letter names,
however, made some intriguing errors. Unlike the kindergartners and first graders, these preschoolers produced many single-
letter spellings. The single letter that they used was often the consonant letter suggested by the letter-name sequence in the
spoken syllable. For example, the preschoolers often spelled /gar/ as R, /zef/ as F, and /tib/ as T. These errors are interesting
because children generally have more success spelling initial consonants than final ones (Stage & Wagner, 1992; Treiman, 1993;
Treiman, Berch, & Weatherston, 1993). Letter-name knowledge is apparently potent enough to override this tendency, making
preschoolers more likely to spell /gar/ as R, the final letter, than as G, the initial letter. Thus, children who have a strong
knowledge of letter names before they begin formal schooling may be able to take advantage of this knowledge to produce
sound-based spellings. A child who spells /gar/ as R is demonstrating a knowledge that the spelling of a word is not arbitrary
that the letters in a spelling have some connection to a word's sound. A child may be able to spell /gar/ as R based on the overall
similarity in sound between /gar/ and /ar/; the child need not be able to divide the spoken syllable /gar/ into smaller units of
sound (see Treiman & Breaux, 1982). Although spellings like R for /gar/ appear very primitive, the children who produce these
spellings may be able to appreciate that certain aspects of conventional spelling, such as the r in car or the p in pizza, make
sense given the sounds in words. In this way, letter-name knowledge may help children take their first steps toward
understanding that writing is connected to speech.
The names of letters may also help children to discover and remember the sounds that letters make, at least for those letters
whose names suggest their sounds. Evidence to support this claim comes from a study by Treiman, Weatherston, and Berch
(1994, Exp. 2). These researchers asked preschoolers and kindergartners to supply the first letters of syllables such as /ba/, /la/,
and /ga/ and, in another condition, the last letters of syllables such as /ab/, /al/, and /ag/. Children were simply asked to say the
first or last letters of the spoken syllables; they were not asked to write them. Knowledge of letter names could help children to
remember the link between /b/ and b and the link between /l/ and l. This is because /b/ is found in the name of b and /l/ is found
in the name of l. Letter names could not help children with the mapping between /g/ and g because the name of the letter g does
not contain the phoneme /g/. Children indeed performed better in the supply-the-first-

Page 69
letter task on syllables for which letter-name knowledge could help them (e.g., /ba/, /la/) than on syllables for which letter-name
knowledge would be useless or even misleading (e.g., /ga/). Similarly, performance in the supply-the-last-letter task was better
for syllables such as /ab/ and /al/ than for syllables such as /ag/. An additional finding is that performance was better with letters
like b, where the phoneme that the letter symbolizes is at the beginning of a CV letter name, than with letters like l, where the
phoneme that the letter symbolizes is at the end of a VC letter name. This difference may arise because /b/ is in the salient onset
position of the CV letter name /bi/. Children's ability to divide the letter name into its onset and rime (Bowey & Francis, 1991;
Kirtley, Bryant, Maclean, & Bradley, 1989; Treiman, 1985, 1992) helps them access the /b/ in the syllable /bi/. In contrast, the
/l/ of /el/ is in a less salient position in the spoken syllable and is more difficult to separate from the vowel.
Thus, the phonological properties of a letter's name that make it easier or harder to segment into phonemes seem to have at least
two effects. First, these properties influence children's tendency to use the letter to spell the sequence of phonemes in its name.
Second, they affect children's ability to use the letter's name as a clue to its sound. As an example, the name of b, /bi/, is
relatively easy for children to divide into phonemes. Its segmentability means that errors like BT for beat will be uncommon but
that children will be able to use b's name to remember its sound. The name of /l/, /el/, is harder to break apart. Hence, children
will make a fair number of errors like BL for bell and will have more trouble using the letter's name as a clue to its sound.
Children's tendency to use letter names to discover the letters' sounds sometimes leads them astray. Consider the letter w, which
is used in English to spell the phoneme /w/. However, the name of w ("doubleyou") does not contain /w/. The letter name that
contains the /w/ phoneme is instead the name of y, /wai/. Not surprisingly, the kindergartners tested by Treiman, Weatherston,
and Berch (1994) sometimes said that y makes the sound /we */. Also, they sometimes misspelled words with initial /w/ with y,
as in YAT for wet or YRM for warm. These errors were nearly absent by the spring of first grade (see also Read, 1975;
Treiman, 1993). Thus, young children who can divide spoken syllables into onsets and rimes may analyze the CV name of y
into /w/ and /ai/. They may conclude that y should be used to spell /w/ just as other letters with CV names, such as b and k, are
used to spell the first phonemes of their names. As children learn the spellings of common words such as went and will, in
which /w/ is spelled with w, or as they are taught that w makes the sound /w/, they abandon their early belief that y corresponds
to /w/.
Exposure to print may thus be one factor that leads to the eventual disappearance of errors such as YAT for wet and CR for car.
Children see

Page 70
that words beginning with /w/ generally start with w rather than y. They observe that printed words almost always contain a
vowel. Indeed, first graders' knowledge that printed words must include vowels leads them to sometimes produce spellings such
as GRE for the nonword /gar/ (Treiman, 1994). A child's belief that /ar/ is an indivisible phonological unit suggests that the rime
be spelled with single r; the child's orthographic knowledge suggests that the word contain a vowel. The child may solve this
conflict by placing an e, which occurs as a "silent" letter at the ends of words such as came and give, at the end of the spelling.
Thus, even in the early stages of spelling development, children may notice what words look like and may use this information
when constructing their own spellings. In the next section, we focus more directly on children's orthographic knowledge.
Orthographic Knowledge
By orthographic knowledge, we mean children's understanding of the conventions used in the writing system of their language.
Proficient writers possess a good deal of information about the orthography, including knowledge about the spacing of words,
the orientation of writing, acceptable and unacceptable letter sequences, and the variety of ways in which certain phonemes may
be represented, depending on such factors as their position in a word. It is widely held that individuals develop orthographic
knowledge based on their experiences with the printed language. This knowledge, in addition to phonology, then influences
spelling.
The theories of spelling development we have described (Ehri, 1986; Gentry, 1982; Henderson, 1985) maintain that
orthographic knowledge does not begin to affect a child's spelling until the child has accumulated a considerable number of
words that are recognized by sight. These theorists describe orthographic knowledge as the learning of complex sequences such
as the -ight of words such as light and learning when to double consonants in polysyllabic words. Very likely, knowledge of
complex orthographic patterns such as these does develop late relative to other spelling skills. However, the theories fail to
account for the possibility that young children possess knowledge about simpler conventions of the orthography.
Treiman (1993) examined early spellings for evidence of adherence to relatively simple orthographic conventions. She looked at
the writings produced over the course of the school year by first graders. These first graders were in a curriculum in which they
were encouraged to write but their spelling was not corrected. The children tended to produce spellings that were consistent with
the regularities of the English writing system. For instance, vowels and consonants that are more likely to occur as doublets in
English, such as e and l, were more likely to occur as doublets in the

Page 71
children's writing than letters that infrequently or never double, such as u and h. Another example of children's adherence to
orthographic patterns involves ck. The ck spelling never occurs at the beginning of a word in written English; initial /k/ is
instead spelled with c or k. The children in Treiman's study did use ck but rarely at the beginning of a word. These and other
examples suggest that beginning spellers develop a sensitivity to simple orthographic patterns earlier than expected given current
models of spelling development.
Treiman's (1993) evidence of orthographic knowledge came from the writings that first graders produced as part of their
everyday classwork. Treiman also experimentally investigated children's knowledge of orthography by devising an orthographic
constraints test. This test included 16 pairs of nonwords. Each pair tested a constraint or regularity of the English writing
system. One nonword in a pair conformed to the regular pattern; the other word did not. However, both nonwords were
pronounceable. For example, one pair was ckun and nuck. Children were to choose which item looked more like a real word. If
children make their judgments on the basis of sound only, both items would be equally likely to be chosen. However, if children
consider orthographic acceptability in making their judgments, the item that conformed to the orthographic constraint should be
chosen more often. Treiman found that middle-class kindergartners (tested in May), first graders (tested in March), and second
graders (tested in May) all chose the conforming item significantly more than 50% of the time. The percentages of correct
responses were 56% for kindergartners, 62% for first graders, and 83% for second graders. The above-chance performance of
kindergartners and first graders supports the idea that knowledge of orthography begins to emerge earlier than previously
thought.
Treiman's (1993) orthographic constraints test contained 16 pairs of nonwords testing a variety of different constraints.
Knowledge of no one constraint was explored in detail. Some further studies, however, specifically investigated children's
knowledge of double consonants and vowels (Cassar & Treiman, in press). The first study focused on consonants. In English,
consonant doublets may occur in the middle or at the end of a word but not at the beginning. In addition, medial doublets
normally follow short vowels in polysyllabic words, as in latter. Single consonants normally follow long vowels, as in later.
Pairs of nonwords were designed to test children's knowledge of these conventions. In pairs that tested knowledge of position,
one nonword contained an initial doublet and the other nonword contained a final doublet. An example is nnus and nuss. If
children know where consonant doublets may occur, they should judge that nuss is more likely to be a word than nnus. In pairs
that examined the phonetic environment for doubling, one nonword contained a medial single consonant, as in salip, and the
other contained a medial doublet, as in sallip. Cassar and Treiman asked

Page 72
whether participants listening to pronunciations for the nonwords chose salip for a pronunciation with ''long a'' or /e/ in the first
syllable and sallip for a pronunciation with "short a" or //.
The participants in Cassar and Treiman's (in press) first study were children in kindergarten, first, second, third, sixth, and ninth
grades as well as college undergraduates. One group at each grade level was assigned to an auditory condition in which they
chose which spelling looked best for the word they heard. The other group was assigned to a visual condition in which they
viewed pairs of spellings and chose the one they thought looked more like a real word. For the pairs that tested knowledge of
position (e.g., nuss and nnus), even kindergartners tested in the first semester of the school year picked the final doublet
spellings significantly more often than chance. Not until sixth grade and above did children evidence knowledge of the
correspondence between short vowels and spellings with medial doublets. That is, only the older children were able to reliably
choose sallip for the pronunciation /s'lip/ and salip for the pronunciation /s'elip/. These latter results agree with Henderson's
(1985) claim that knowledge about the complex pattern of short vowel plus double consonant in two-syllable words develops
later than the first few years of elementary school. However, young children's above-chance performance on the test of
knowledge about the positions of consonant doublets suggests that some types of orthographic knowledge emerge much earlier
in the course of spelling acquisition. This finding fits with Treiman's (1993) observation that the first graders in her study did
not often make errors such as BBAL for ball. It appears that beginning spellers not only represent speech sounds when they
write but that they also attempt to honor the kinds of letter sequences to which they have been exposed.
Cassar and Treiman's (in press) second study further investigated children's knowledge of double consonants as well as double
vowels. An attempt was also made to identify the spelling development of each child in terms of the stages outlined in the
theories of spelling development. An examination of children's knowledge of doublets in relation to their level of spelling
development was expected to yield insights into the types of information that children use when they are beginning to spell. As
in the first study, an orthographic choice test employed pairs of nonwords. One nonword in each pair contained an acceptable
vowel or consonant doublet; the other nonword contained an unacceptable doublet. The doublets occurred in the medial or final
positions of the spellings, both of which are acceptable for doublets. Sample pairs are noss and novv and geed and gaad. If
children know which letters are allowed to double, they should choose noss and geed as more word-like than novv and gaad.
To identify each child's level of spelling development, the child was asked to spell a list of common words such as farm and
people. The spellings were

Page 73
scored on the basis of features thought to predominate at particular stages of spelling development. Beginning spellers
considered to be at the semiphonetic and phonetic stages as described by Ehri (1986) and Gentry (1982), which correspond to
the letter-name stage of Henderson (1985), were targeted in this study. According to the theories, these children focus only on
sounds or phonetic mappings when they spell.
Kindergarten, first- and second-grade children were tested during February. First and second graders, but not kindergartners,
chose spellings containing allowable doublets over spellings containing unallowable doublets. That is, they judged that noss and
geed were more like real English words than novv and gaad. Importantly, children classified as phonetic spellers performed
significantly better than chance on the orthographic choice test. Children at the semiphonetic level did not show above-chance
levels of performance. These results suggest that children begin to learn about which letters may double in English earlier than
previously thought. Gentry (1982) claimed that during the phonetic stage, letters are assigned strictly on the basis of sound,
without regard for the conventions of English orthography. The results of Cassar and Treiman's (in press) study dispute this
claim by showing that children whose spellings have many of the hallmarks of the phonetic or letter-name stage possess some
knowledge of orthographic conventions. Children may begin to learn what words look like earlier than commonly believed.
They use this knowledge, together with their knowledge of phonology and letter names, to construct spellings.
Morphological Knowledge
The theories of spelling development we have reviewed depict the use of morphology to guide spelling as a late development. In
fact, Henderson (1985) advised teachers that children need to be taught about the morphological relations among words. The
relations to which Henderson referred are those such as the relation between courage and courageous. These connections may
not be obvious to children, and so it may be best to point them out. However, not all morphological relations are this complex.
Some, such as the relations between rain and rained or bar and bars, are simpler. Children can handle these kinds of
morphological alternations in their speech from an early age (Berko, 1958). Hence, young children might take advantage of such
relations in their spelling as well.
Consider the verb wait. By adding the suffix -ed, we create the past tense form waited. The addition of -ed also changes the
pronunciation of the /t/ to a flap, a quick tap of the tongue against the top of the mouth. The flap of waited, which is voiced, is
similar to /d/, which is also voiced. If young spellers use only sound-based information, they should spell the flap of waited with
a d. However, if children understand the relation in meaning

Page 74
between wait and waited, and if they realize that this relation is marked in spelling, they should use a t in waited based on the
spelling of the stem. This was the reasoning behind two studies conducted by Treiman, Cassar, and Zukowski (1994).
The first study to be discussed (Treiman, Cassar, & Zukowski, 1994, Exp. 3) examined children's spellings of words containing
flaps that were conventionally spelled with either t or d. The words contained either one or two morphemes. For instance, duty
and attic are one-morpheme words with t flaps, and waited and dirty are two-morpheme words with t flaps. Sturdy and louder
are examples of one- and two-morpheme words containing flaps that are conventionally spelled with d. If children use meaning
relations among words to aid their spelling, they should do better on the flaps of the two-morpheme words (for which the stems
can help them) than on the flaps of the one-morpheme words (for which there are no embedded morphemes that can aid
performance). The results for first and second graders confirmed this hypothesis. These children were more likely to use t for
two-morpheme words such as dirty than for one-morpheme words such as duty. The first graders, especially, often misspelled
the flaps of both types of words as d, pointing to an influence of sound on spelling. However, the children were less likely to
make this error for words like dirty than for words like duty. Children did not use morphological information to the full extent
possible. Whereas they almost never misspelled the t of dirt with d, they sometimes misspelled the t of dirty in this manner.
There is room for improvement in children's use of morphological relations among words to guide their spelling, but this
strategy is by no means beyond their capabilities.
The youngest children in the study just described (Treiman, Cassar, & Zukowski, 1994, Exp. 3) were second-semester first
graders. Another study (Treiman, Cassar, & Zukowski, 1994, Exp. 4) was designed to explore whether even younger children
can also benefit from meaning relations when spelling. Kindergarten, first-, and second-grade children were tested twice during
the school year, once in the fall and then again in the spring. As before, the stimuli included words with medial flaps that were
conventionally spelled with either t or d. Of the words containing each type of flap, half contained a single morpheme and the
other half contained two morphemes. Instead of spelling the whole words, as in the preceding study, children were asked to fill
in the missing t or d in each word's spelling.
If young children's spelling is based only on sound, as the theories suggest, then children should spell the flaps of both one- and
two-morpheme words with d. However, if children use morphology to supplement their performance, then they should perform
better on the two-morpheme words than the one-morpheme words. This is what Treiman, Cassar, and Zukowski (1994, Exp. 4)
found. Overall, children did better on d flap words than t flap words. Because flaps, being voiced, are closer in sound to /d/ than
to /t/,

Page 75
children tend to spell them with d. This leads to correct responses for d flap words but incorrect responses for t flap words.
Importantly, the kindergartners and first graders performed significantly better on the flaps of the two-morpheme words than
those of the one-morpheme words. (The second graders did not show a significant difference, probably due to their relatively
high overall levels of performance.) Evidently, the children used their knowledge of the spellings of the stems to help spell the
flaps in the two-morpheme words. For instance, the link in meaning between float and floated suggested to children that the flap
of floated should be spelled with a t. Children were less likely to use a t in floated than in float, showing that they did not use
morphological information to the full extent of their ability. Nevertheless, these kindergartners and first graders could use
meaning relations to guide their spelling even when so doing meant overriding sound.
Together, the results of the two studies with flaps (Treiman, Cassar, & Zukowski, 1994, Exps. 3 and 4) indicate that children as
young as kindergarten age have some ability to use simple morphological relations among words to guide their spelling. This
result speaks against the view that the use of morphological information in spelling is a late development. To determine whether
children are affected by morphology more generally, Treiman and Cassar (1996) turned to the case of final consonant clusters.
Children sometimes fail to spell the first consonants of final clusters, especially when these consonants are liquids (/r/ or /l/) or
nasals (/m/, /n/, or /h/) (Marcel, 1980; Read, 1975; Treiman, 1993; Treiman, Zukowski, & Richmond-Welty, 1995). For
instance, first graders may spell horse as HOS or the nonword /vLnt/ as VUT. If children have trouble including the n in one-
morpheme words such as brand, they should also have problems with the n in two-morpheme words such as rained unless they
use their knowledge of the stem, rain, to aid them. However, if children are able to use morphology, they should omit the first
consonant of a final cluster less often in two-morpheme words than in one-morpheme words. Thus, the underlying logic of the
experiments with final clusters was the same as in the flap studies described earlier.
The children in the experiments by Treiman and Cassar (1996) were asked to spell words with final consonant clusters,
including one-morpheme words such as brand and two-morpheme words such as rained. The one- and two-morpheme words
had the same final clusters in their spoken forms; for instance, brand and rained both end with /nd/. About a week after spelling
the one- and two-morpheme words, the children were asked to spell the stems of the two-morpheme words (e.g., rain). The
children's spellings of final cluster words were scored according to how they represented the final consonants. An A spelling
was one that contained a reasonable representation of the first consonant in the cluster but not of the second consonant, as in
RAN for rained. A B spelling represented only the second consonant, as in BRAD for brand. AB spellings symbolized both
consonants in the final clus-

Page 76
ter, as in RAND for rained. Given previous findings that children sometimes fail to spell the first consonants of final clusters,
we expected to see a fair number of B spellings and fewer A spellings for the final cluster words. We asked whether B spellings
would be less common for two-morpheme words than for one-morpheme words, suggesting that children benefited from the
additional meaning information in the two-morpheme words.
In the first study by Treiman and Cassar (1996), children in first, second, and fourth grades were tested in March. Overall,
children were more likely to omit the first consonant of a final cluster (B errors) than the second consonant of a final cluster (A
errors). However, B errors were less frequent for two-morpheme words such as rained than for one-morpheme words such as
brand. That is, children were less likely to spell rained as RAD than to spell brand as BRAD. Children apparently benefited
from the stem rain when spelling the two-morpheme word rained. However, the first and second graders did not use their
knowledge of the stem to the full extent possible because they left out the n of rained more often than the n of rain. Errors in
which the final consonant of the cluster was omitted, or A errors, were more common for two-morpheme words than for one-
morpheme words. That is, children misspelled rained as RAN more often than they misspelled brand as BRAN. In assembling
the spelling of a two-morpheme word from the spellings of the component morphemes, children sometimes seemed to forget the
second morpheme.
Our second study (Treiman & Cassar, 1996) included children in kindergarten as well as first grade. The kindergartners were
tested once, late in the school year. The first graders were tested both in the fall and the spring. The children were asked to
complete each word's spelling. For instance, bra ___ was printed on the test sheet and the children were told to finish this
spelling for the word brand. As in the first study, omissions of the first consonant of the final cluster (B errors) were more
common for one-morpheme words such as brand than for two-morpheme words such as rained. That children were more likely
to omit the n of brand than the n of rained suggests that they derived some benefit from the stem rain. In contrast, omissions of
the second consonant of the final cluster (A errors) were more common on two-morpheme words than on one-morpheme words.
Children sometimes spelled rained without a final d, suggesting that they stopped writing after they had symbolized the first
morpheme.
The results of the experiments with flaps and final clusters indicate that children's spelling can be influenced by morphology
from an early age. This conclusion is different from that drawn by Treiman (1993) based on her naturalistic study of first
graders' classroom writings. In that study, first graders' spellings were seemingly unaffected by morphology. For example, the
children in Treiman's naturalistic study were not significantly more likely to include the n in a two-morpheme word such as
rained than in a one-mor-

Page 77
pheme word such as brand. The different results in the two studies might reflect task differences. In Treiman's naturalistic study,
children wrote words as part of texts that they produced during the classroom writing period. In the experiments described so
far, children spelled or completed spellings for isolated words presented by an experimenter. Hoping that the different findings
might be related to these task differences, we designed a third study using the same two-morpheme and one-morpheme words
with final clusters.
Treiman and Cassar's (1996) third study included two groups of first-semester second graders. One group of children spelled the
words using a whole word spelling task. The other group of children was asked to compose a sentence using each word and
then to write out the entire sentence. If composing a sentence is more like the writing task in the Treiman (1993) study, then
children in the sentence condition should produce similar types of spellings for the final consonant clusters in one- and two-
morpheme words. However, this did not turn out to be the case. The children in the sentence condition, like those in the word
condition, showed different patterns of performance on the one-morpheme and two-morpheme words. Omissions of the first
consonant of the final cluster (B errors) were more common on one-morpheme words than two-morpheme words, and omissions
of the second consonant of the cluster (A errors) were more common on two-morpheme words.
There are several plausible explanations for the failure of the sentence condition to replicate the results of Treiman (1993). The
most obvious is that the sentence task did not duplicate the conditions of the naturalistic setting. In the experiment, the children
were required to use a particular word to compose a complete idea. In the classroom, children wrote their ideas without
restrictions as to word usage. In this situation, words may become part of the meaning of the message without having to be
individually analyzed in terms of meaning. In contrast, a child who is required to use a specified word in a sentence may need to
analyze the word's meaning to determine how it should be used. This may promote morphological analysis of the two-
morpheme words. Of course, this explanation is conjecture without more evidence. The most striking and important finding
from Treiman and Cassar's (1996) third experiment, as well as from the other experiments on flaps and final clusters, is that
young children have some ability to use morphological strategies in spelling. In this way, children avoid errors that they would
otherwise make if they were bound solely to a phonological spelling strategy.
Conclusions
Existing theories of spelling development (Ehri, 1986; Gentry, 1982; Henderson, 1985) portray children as passing through a
series of qualitatively different stages as they learn to spell. During an early period, children are

Page 78
said to treat writing as a task of symbolizing the sounds in words. They progress through each word from left to right, using
their knowledge of phoneme grapheme correspondences and of letter names to represent the word's sounds. Only later, according
to the theories, do children begin to symbolize a sound differently depending on its position or its context. Only later do children
use morphological relations, symbolizing meaningful words instead of sounds.
The results that we have reviewed suggest that these views of spelling development are too limited. There is no doubt that
young spellers rely on sound to a large degree, but this is not all that they do. For example, children often spell flaps as d,
testifying to the importance of sound. However, children's knowledge of the relation in meaning between dirty and dirt makes
them less likely to misspell the flap of dirty with a d than to misspell the flap of duty with a d. Children have some ability to
divide dirty into dirt and -y, some notion that spelling represents meaning as well as sound. This is a long way from being able
to retain the spelling of courage in courageous, but it is a start in that direction. Along with their attention to meaning, young
children also attend to the kinds of letter sequences that make up words. Early on, they notice that English words may start but
not end with capital letters and that words may end but not start with double consonants. They notice that letters such as e and s
may double but that letters such as a and v rarely do. Children attempt to reproduce what they have observed, so even first
graders do not usually produce such errors as BBAL for ball or HAAT for hat.
In our view, spelling involves an interaction among different sources of knowledge from the beginning. Although sound plays a
central role in early spelling, young children also have other kinds of knowledge, including knowledge of morphological
relations among words and information about the kinds of letter sequences that typically occur in print. Theories that treat
learning to spell in terms of a sequence of stages that involve qualitatively different types of information are too simple. We
must acknowledge that spellers, even young ones, are able to use various types of knowledge. We must learn more about how
these different sources of information interact.
Acknowledgments
Preparation of this chapter was supported by NSF Grant SBR-9020956. We thank Ruth Tincoff for her comments on a draft of
the manuscript.
References
Berko, J. (1958). The child's learning of English morphology. Word, 14, 150-177.
Bissex, G.L. (1980). Gnys at wrk. Cambridge, MA: Harvard University Press.

Page 79
Bowey, J.A., & Francis, J. (1991). Phonological analysis as a function of age and exposure to reading instruction. Applied
Psycholinguistics, 12, 91-121.
Cassar, M., & Treiman, R. (in press). The beginnings of orthographic knowledge: Children's understanding of simple letter
patterns. Journal of Educational Psychology.
Derwing, B.L., & Nearey, T.M. (1991, August). The "vowel-stickiness" phenomenon: Three experimental sources of evidence.
Paper presented at the Twelfth International Congress of Phonetic Sciences, Aix-en-Provence, France.
Durrell, D.D. (1980). Letter-name values in reading and spelling. Reading Research Quarterly, 16, 159-163.
Ehri, L.C. (1983). A critique of five studies related to letter-name knowledge and learning to read. In L.M. Gentile, M.L. Kamil,
& J. Blanchard (Eds.), Reading research revisited (pp. 143-153). Columbus, OH: Merrill.
Ehri, L.C. (1986). Sources of difficulty in learning to spell and read. In M.L. Wolraich & D. Routh (Eds.), Advances in
developmental and behavioral pediatrics (Vol. 7, pp. 121-195). Greenwich, CT: JAI.
Gentry, J.R. (1982). An analysis of developmental spelling in GNYS AT WRK. The Reading Teacher, 36, 192-200.
Henderson, E. (1985). Teaching spelling. Boston: Houghton Mifflin.
Hindson, B.A., & Byrne, B. (1997). The status of final consonant clusters in English syllables: Evidence from children. Journal
of Experimental Child Psychology, 64, 119-136.
Kirtley, C., Bryant, P., Maclean, M., & Bradley, L. (1989). Rhyme, rime, and the onset of reading. Journal of Experimental
Child Psychology, 48, 224-245.
Laubstein, A.S. (1987). Syllable structure: The speech error evidence. Canadian Journal of Linguistics, 32, 339-363.
Marcel, T. (1980). Phonological awareness and phonological representation: Investigation of a specific spelling problem. In U.
Frith (Ed.), Cognitive processes in spelling (pp. 373-403). London: Academic Press.
Mason, J.M. (1980). When do children begin to read: An exploration of four year old children's letter and word reading
competencies. Reading Research Quarterly, 15, 203-227.
Read, C. (1975). Children's categorization of speech sounds in English (NCTE Research Report No. 17). Urbana, IL: National
Council of Teachers of English.
Stage, S.A., & Wagner, R.K. (1992). The development of young children's phonological and orthographic knowledge as
revealed by their spellings. Developmental Psychology, 28, 287-296.
Stemberger, J. P. (1983). The nature of /r/ and /l/ in English: Evidence from speech errors. Journal of Phonetics, 11, 139-147.
Treiman, R. (1984). On the status of final consonant clusters in English syllables. Journal of Verbal Learning and Verbal
Behavior, 23, 343-356.
Treiman, R. (1985). Onsets and rimes as units of spoken syllables: Evidence from children. Journal of Experimental Child
Psychology, 39, 161-181.
Treiman, R. (1992). The role of intrasyllabic units in learning to read and spell. In P.B. Gough, L. Ehri, & R. Treiman (Eds.),
Treiman, R. (1994). Use of consonant letter names in beginning spelling. Developmental Psychology, 30, 567-580.
Treiman, R., Berch, D., & Weatherston, S. (1993). Children's use of phoneme-grapheme correspondences in spelling: Roles of
position and stress. Journal of Educational Psychology, 85, 1-12.

Page 80
Treiman, R., & Breaux, A.M. (1982). Common phoneme and overall similarity relations among spoken syllables: Their use by
children and adults. Journal of Psycholinguistic Research, 11, 569-598.
Treiman, R., & Cassar, M. (1996). Effects of morphology on children's spelling of final consonant clusters. Journal of
Experimental Child Psychology, 63, 141-170.
Treiman, R., Cassar, M., & Zukowski, A. (1994). What types of linguistic information do children use in spelling? The case of
flaps. Child Development, 65, 1310-1329.
Treiman, R., Weatherston, S., & Berch, D. (1994). The role of letter names in children's learning of phoneme-grapheme
relations. Applied Psycholinguistics, 15, 97-122.
Treiman, R., Zukowski, A., & Richmond-Welty, E.D. (1995). What happened to the "n" of sink? Children's spellings of final
consonant clusters. Cognition, 55, 1-38.
Worden, P.E., & Boettcher, W. (1990). Young children's acquisition of alphabet knowledge. Journal of Reading Behavior, 22,
277-295.

Page 81
Chapter 5
How Learning to Spell German Differs from Learning to Spell English
Heinz Wimmer
Karin Landerl
University of Salzburg
Psychological research on the acquisition of spelling is quite anglo-centric due to the simple fact that the majority of findings
come from children who learn to spell English. The present study questions some of the English conclusions by contrasting them
with German findings. German offers an interesting test case for conclusions about the role of orthographic consistency and of
phonology in the development of spelling because, due to the partly common roots of German and English, there are similarities
with respect to phonology, although German is written in a more consistent manner than English.
The consistency with which phonemes map onto graphemes is of obvious importance for spelling both within and between
writing systems. For example, Treiman (1993), in her landmark study on early spelling, concluded from the regularity effect
observed in the spellings of her first-grade sample "that a writing system with one-to-one relations from phonemes to graphemes
would be easier for children to learn than a writing system with one-to-many relations from phonemes to graphemes" (p. 59).
German is certainly not a writing system with one-to-one relations from phonemes to graphemes. For nearly every phoneme
there are at least two possible spellings. For example, /a:/ can be spelled as a, aa, and ah, and for /b/, as for most other
consonants, possible spellings are b and bb. However, doubling of consonant letters only occurs after short vowels (e.g., Sommer
[summer], Butter [butter]), and most of the different spellings for vowels have to do

Page 82
only with vowel length. Another inconsistency is that / / is spelled with sch and with s. However, the use of s for / / is
restricted to cases of / / before /t/ and /p/, as in stehen (stand) or Sport (sport).
One obvious difference between English and German is that the latter makes little use of silent letters. For example, in words
such as Knie (knee) the k, in contrast to English, is not silent and, similarly, the final -e in words such as Nase (nose) or Liebe
(love) is pronounced. However, the main difference between German and English lies in the consistency of vowel spelling.
Identical spellings of different vowels such as in sign - signature or heal - health hardly exist in German, but, as already noted,
there are different spellings for the same phoneme, for example, for the /a:/ in Zahl versus in Saal or for the / / in Blle versus
in Welle. The diacritical marker in the case of Blle (balls) indicates vowel change due to pluralization while preserving
morpheme identity (e.g., Ball - Blle, Ofen - Ofen). Even this comparatively moderate way to preserve morpheme identity
creates spelling problems as it has the effect that there are two possible graphemes ( and e) for / /. Furthermore, vowel
graphemes with diacritical markers are not only used to preserve morpheme identity. In Br (bear), for example, the is not
due to a morphological process. However, despite these ambiguities with which vowels are represented in the German writing
system, the difference from English is massive. Most often, there are only two possible ways to represent a vowel and, in
addition, the use of phonological and morphological constraints often reduces the possibilities to one.
Study 1 explores the effect of these differences between the German and the English writing systems with respect to vowel
representation. For this comparison, we used words with similar pronunciation and with similar spelling in the two orthographies
(e.g., rose - Rose, friend - Freund). Expectations are that German children should produce an incorrect vowel spelling only if
phoneme-grapheme correspondences are ambigous and that the error should consist of the alternative spelling possibility. For
English children, it was expected that the overall inconsistency of the phoneme-grapheme relations for vowels may lead to
general uncertainty about how to spell vowel sounds.
Besides phoneme-grapheme inconsistency, Treiman (1993) identified phonological aspects of spoken English as causing
spelling problems: ''Not all of children's difficulties in learning to spell stem from the nature of our writing system. In addition to
the problematic aspects of the English writing system its one-to-many relations from phonemes to graphemes, its use of
multiple-letter graphemes, and its tendency to spell morphemes consistently despite changes in their phonemic forms the nature
of spoken English also contributes to children's difficulties'' (p. 59). She concluded that "beginning spellers would have less
difficulty, at least in the sense of producing fewer illegal errors, in languages whose spoken forms contain few or no consonant

Page 83
clusters" (p. 60). The difficulty of consonant clusters according to Treiman's analysis has to do with the fact that spelling
clusters requires phonemic segmentation of the cluster, which is difficult because the cluster is treated as a phonological unit,
namely, as syllable onset or as coda of the rime. German provides a test case for this conclusion because, similar to English,
German has many words with consonant clusters (e.g., blau - blue, drei - three, Hand - hand, Wolf - wolf), and one should,
therefore, expect that German children should have similar difficulties with the spelling of consonant clusters as English
children.
However, the other possibility is that the higher consistency of the German writing system may also affect children's ability to
spell consonant clusters. This effect of consistency of cluster spelling is expected to operate indirectly via reading acquisition. In
other studies with German first graders, it was found that the high consistency of German's grapheme phoneme relations in
conjunction with the common phonics instructional approach makes acquisition of grapheme-phoneme recoding as a reliable
word recognition strategy easy (Wimmer & Goswami, 1994; Wimmer & Hummer, 1990). For spelling acquisition, it is relevant
that word recognition via grapheme-phoneme recoding provides systematic segmentation training. For example, recoding of
blau as /b/-/l/-/ao/ and subsequent recognition as the spoken word "blau" makes prevalent the phonemic segments of the
consonantal onset cluster. The expectation is that German children may have little difficulty with spelling consonant clusters
because of their reliance on recoding as an early word-recognition strategy. Study 2 examines these differing expectations for
consonant cluster spelling among German children and also provides information on several other aspects of spelling
development among German children.
Before presenting the two studies, some information on the instructional approach experienced by the German children
participating in these studies may be useful. Due to the fact that there is no reading preparation in kindergarten, our children
begin school often with very limited letter knowledge, with no reading ability and, often, with no phonemic awareness at all
(Wimmer, Landerl, Linortner, & Hummer, 1991). In Grade 1, children are then exposed to a slowly proceeding, phonics-based
instructional approach, where letter sounds, instead of the conventional letter names, are learned. For example, for the letter f,
children learn /f/ and not /ef/ or /fe */, and, similarly, for m, they learn /m/ and not /em/ or /me */. For the letters representing
stop consonants, the letter sound is the stop plus a schwa sound, for example, /pe */ for p. The first letters children learn are for
vowels and continuants to make word recognition via sounding out and blending easier to demonstrate and to practice with
words such as Mama, Oma, im, am. The multiletter graphemes (sch, ch, au, eu, u, ei) are introduced in the same way as single
letters, and graphic markers are used to signal multiletter graphemes in

Page 84
words. Graphic markers are also used for syllable boundaries to make blending easier. In the beginning, children are induced to
utter their blending attempts as word preforms that allow the teacher to help and to correct. The word preforms quite often have
artificially lengthened phonemes and wrong stress assignments, but due to the consistency of German orthography, they are
normally close enough to the target pronunciation to allow recognition. In first grade, there is more emphasis on reading than on
spelling. Children normally spell words and sentences that they have already read. Invented spelling of new words is not
encouraged, and misspellings are always corrected.
Study I: Vowel Spelling by German and English Children

This exploration of German-English differences in children's spellings of vowel sounds was an addition to a reading study that
focused on orthography effects on reading in the first years in school. To minimize memory and attentional problems involved
in phonemic analysis and in letter production, the children's task was to simply insert the missing letters for the vowels in given
skeleton spellings. A main methodological feature of the study was that words with similar pronunciations and spellings and
with identical meanings in the two languages/orthographies were used (e.g., Nase - nose, Boot - boat, Preis - prize). From the
identical meanings in the two languages, it follows that the frequency with which these spellings were encountered by German
and English children should not be too different. From the similarity of the words, with respect to phonology and with respect to
number and kinds of letters, it follows that phonemic analysis and letter production should not account for systematic spelling
differences between German and English children.
Eighty-two English and 87 German children participated with about equal numbers of children from Grades 2, 3, and 4. The
English children had about half a year more of school experience than the German children because testing took place at the end
of the school year in London and around the middle of the year in Salzburg. At least for the younger children, it can be expected
that they may not have learned the conventional spellings to a major extent and, therefore, their spellings may be affected by the
difference between German and English in the consistency with which vowels map onto graphemes.
The spelling task was performed in class. Each child received a sheet with 29 skeleton spellings from which the letter(s) for the
main vowel was missing (e.g., gr_n for green). The experimenter first pronounced each word, then read a sentence that included
the word in order to specify the meaning,

Page 85
then repeated the word once again. For example, "green The grass is green green." After this, children had to insert the missing
letter(s) for the vowel into the skeleton spelling. The first 5 items were used to explain and practice the task, leaving 24 test
items. Table 5.1 shows the English and the corresponding German test words. The italicized letters were missing in the skeleton
spelling. As evident, the final -e was provided in both the English and German spellings. Table 5.1 also shows that, in the case
of heart - Herz and pearl - Perle, the r was deleted together with the vowel grapheme. This was not done in the case of bear -
Bar to avoid the situation in which only the first letter would have formed the skeleton spelling.
Differences in how English and German children spelled the vowel of each item pair are presented in Table 5.1. From percent
correct in Table 5.1, it is
TABLE 5.1
English and German Vowel Spellings
Percent Correct Number of Examples of Misspellingsb
Different
Spellingsa
Item English German English German English German
rose-Rose 98 100 3 1 ai-e
ghost-Geist 95 99 4 2 a-e-s eu
home-Heim 94 99 4 2 u-e-a
nose-Nase 93 100 6 1 e-h-a-r-p
stone-Stein 90 99 8 2 a-r-w-p-l-oa if
whale-Wal 87 91 8 4 e-i-o-g-r-u ah(5)-aa-o
throne-Thron 85 97 8 3 a-n-e-u-f-i oh-n
green-grun 84 94 11 4 e(5)-er-ne-ea-r-ln-ie-et u--an
prize-Preis 82 100 9 1 u-s-ii-z-a-t-y-e
slime-Schleim 79 95 7 2 e(5)-u(5)-a-r-m-l e
wasp-Wespe 74 99 6 2 o(16)-e-b-t-p
deep-tief 72 89 14 2 e(7)-ea(5)-er-iep-pe-o i(10)
bear-Br 71 66 15 4 ae(5)-ee-eu-a-e-aa e(25)-ee-a
boat-Boot 63 53 9 5 oo(10)-ou(8)-o(6)-ao-ea o(37)-u-a
bread-Brot 60 99 15 2 e(7)-ae(5)-a-er-ee t
nerve-Nerven 54 98 20 3 ur(8)-u(6)-ea-ar-oo-a e-err
coal-Kohlen 50 45 17 4 oo(13)-o(11)-ou-oe-a o(46)-oo
steel-Stahl 48 47 15 5 ea(13)-e(11)-i-ie-a-eu a(40)-aa-o
friend-Freund 48 98 22 3 ei(8)-e(5)-i(5)-ea-ine-ae e-eu
pearl-Perle 39 95 27 4 ea(10)-er(6)-ar-ru-ol-ur e-re-a
heart-Herz 32 99 28 2 ar(18)-ea-aru-ur-r-ae-aer et
soul-Seele 29 28 17 3 oa(21)-ol(9)-oo(8)-oe-o e(52)-eh(11)
rhyme-Reim 26 97 8 3 i(41)-a(9)-e-m-o-r-miss eu-ein
thief-Dieb 22 90 15 3 ei(20)-ee(15)-ea(11)-e(6)-i i(8)-ei
aThe number of different spellings includes the correct spelling.
bIn parentheses the number of children committing the misspelling is given if more
than
one child produced this spelling.

Page 86
evident that for the English words, the difficulty of correct vowel spelling was rather graded, ranging from such easy words as
rose and ghost, with over 90% correct, to such difficult ones as rhyme and thief, with less than 30% correct. In contrast, for the
German words, spelling of the vowels was either very easy or rather difficult: for 19 words, the percentage of correct vowel
spelling was about 90% and higher and, for only 5 words, percentages ranged from about 70% to about 30%. Overall, for 19
word pairs, the vowel was easier to spell for the German than for the English word, whereas for only 5 pairs, the opposite was
the case.
Why vowel spelling proved difficult for some of the German words is easy to understand from the misspellings listed in Table
5.1. In the case of Br (only 66% correct), 25 children used e for /e/. The letter e actually is the more common spelling for /e/,
and in the case of Br, morphological considerations are of no help to decide between e and . In the case of the four other
difficult words Boot, Kohlen, Stahl, and Seele it is obvious from the misspellings that children did not use the specific
orthographic marking of the long vowel in these words. For example, in the case of Kohle, 46 children inserted the single o
letter for the vowel. In the case of Seele, there were 52 children who spelled the /e:/ with the single e letter, and there were an
additional number of children who marked vowel length explicitly with eh. Presumably, it was not a failure to recognize that the
vowel is long in these words, which led to the incorrect spellings, but the inconsistency with which long vowels are
orthographically marked. This orthographic interpretation of the difficulty of the spelling of long vowels is suggested by the
observation that in the case of /i:/ in tief and Dieb, comparatively few children neglected the orthographic marking of the long
vowel. The difference is that for /i:/, there is the specific grapheme ie named long-i that rather consistently represents /i:/
(exceptions are the pronouns ihr [her] and ihm [him]), whereas in the case of /a/, /e/, and /o/, the long version of the vowel can
be represented by a single vowel letter, by a double vowel letter, and by the h after the vowel letter.
The most impressive difference between how English and German children spelled the vowels of the corresponding word pairs
is given by the number of different spellings in Table 5.1. In this number, the correct spelling is included together with the
number of different misspellings. From Table 5.1, it is evident that for each word pair, the English children produced a higher
number of different vowel spellings than the German children, and in several cases, the difference was large (e.g., for friend -
Freund: 22 different English, 3 different German vowel spellings). It is important to note that the orthography difference in the
number of different vowel spellings is not simply due to the difference in the number of correct spellings. For five words (Br,
Boot, Kohlen, Stahl, Seele), the German children produced a higher number of incorrect spellings than the English children.
Neverthe-

Page 87
less, the number of different vowel spellings produced by the German children was small, even for these words, and ranged
from 3 to 5, whereas for English children the number of different spellings for these words ranged from 9 to 16. The
orthography difference in the variety of vowel spellings quite plausibly reflects the difference in the consistency with which
vowels map onto letters in English versus German. Another indication of the uncertainty of what letters English children should
use for vowels is the finding that in 32 cases (1.6% of all words), English children simply did not insert any letter at all in the
skeleton spelling, whereas only 1 such case occurred among the German children.
However, there were a number of misspellings produced by English children that can hardly be explained by uncertainty which
letter(s) to use for a certain vowel. The main characteristic of these misspellings is that consonant letters with no relation to the
pronunciation of the word were used. Therefore, it seems plausible that in these cases, the difficulty had to do with phonemic
segmentation of the word pronunciation and/or with relating the segments to the letters of the skeleton spelling. Altogether there
were 135 (7%) such vowel spellings with consonant letter(s) for the English but only 10 (0.5%) for the German words. The
more plausible type of such consonant inclusion was that the child used the consonant letter given in the spelling skeleton after
the blank together with a vowel letter. For example, in the case of pearl, the final l was used together with one or two vowel
letters. The resulting spellings were poll, pell, parll or paell, and in the case of deep, this strategy led to depp and diepp. There
were 50 such misspellings among the English but only 3 among the German spellings. Another error type was that no vowel
letter was used at all but rather one or two consonant letters. For example, in the case of nose, the letters h, r, and p were
inserted instead of a vowel letter, resulting in the spellings nhse, nrse, npse. There were 46 English misspellings of this strange
type but only 2 German misspellings. Consonant letters were also used in conjunction with vowel letters. Examples are grern,
grren, grnen, and gretn for green and prrikl, peteal, pulel, plurchl, pklul, pulul for pearl. Again, the phonological
inappropriateness of the spellings is striking. It should be noted that for pearl, nerve, and heart, all occurrences of the letter r
either singly or in combination with vowel letters were not counted as an inappropriate use of consonant letters.
One could speculate that the reason for the phonologically inappropriate misspellings made by the English children may have
less to do with lacking knowledge of phoneme-grapheme correspondences, and more to do with difficulties inherent in the
specific form of our spelling task. To find out what letter(s) to insert into the skeleton spelling, the child has to examine the
skeleton spelling, segment the phonological form of the given word into phonemes, relate the resulting phonemic segments to
the letters given, figure out which segment is not represented and, finally, find the letter(s)

Page 88
for the misssing vowel segment. It seems possible that the phonologically inappropriate spellings of the English children result
from difficulties with the segmentation of the phonological word form or from difficulities with the matching of the segments to
the letters given. Evidence for such matching difficulties are those errors of the English children in which they inserted the
given consonant letter after the blank together with a vowel letter (e.g., diepp for deep). Of course, the question is why only the
English children, but not the German children, had difficulties with phonemic segmentation or phoneme-to-letter matching. This
question will be taken up again in the conclusion after Study 2, which shows that German children do not exhibit difficulties
with the spelling of consonant clusters that also require segmentation. The argument is that early reliance on phonological
recoding in word recognition allowed by high grapheme phoneme consistency and induced by phonics instruction provides
German children with segmentation training that is responsible for the advantage German children have with respect to spelling.
First, however, Study 2 presents more comprehensive data on the spelling performance of German children in the early phase of
spelling acquisition. In contrast to Study 1, in which children had to insert vowel letters in skeleton spellings, the participants of
Study 2 produced spellings for whole words. Specific questions were how consonant clusters were spelled and whether children
show differences in the spelling of consonants versus vowels.
Study 2: Spelling by German Children in Grade One

This study is based on spellings collected for standardization of a classroom spelling test in Austrian schools. Three classes,
totalling 68 children, participated at the end of Grade 1. These children had not more than about 9 months of reading and
spelling instruction because in kindergarten, there is no reading preparation. Testing in the classrooms was not done by the
teacher but by a collaborator who was specifically instructed not to pronounce the words in an exaggerated High German
manner, which is closer to the correct spelling than standard Austrian High German. One difference is that in Southern German
dialects and in Austrian High German, the stop consonants /b/, /p/, /d/, and /t/ are neutralized in most positions. For example, the
stops of Paul and Ball and of Rad and Rat sound more or less the same, and, therefore, pronunciation is of little help in
choosing the correct letters.
Children had a test form with 25 sentences in front of them. The test word from each sentence was missing and had to be
inserted by the child. The tester first pronounced the word, then read the sentence, then repeated the word, after which the
children had to fill it in the blank. Table 5.2 shows the 25 test words ordered with respect to difficulty. In Table 5.2, first the

Page 89

Page 90
pronunciation of the word is given, followed by the correct spelling (and the English counterpart) and by percent correct, and,
finally, all misspellings are listed. For misspellings that were produced by more than two children, the number of occurrences is
added in parentheses. Correct capitalization is not considered.
Table 5.2 shows that all the words were rather short not longer than two syllables or six graphemes and can be considered to be
of high frequency in spoken language (as evident from the English counterparts). From the pronunciations, it is evident that
several words include consonant clusters both in initial position and/or in end position. This allows examination of Treiman's
(1993) conclusion that phonological difficulties affect beginning spellers independent of orthographic consistency. Furthermore,
the words present a variety of orthographic problems as evident from the relations between pronunciation and correct spelling in
Table 5.2 in nearly no case did the spelling follow unambiguously from the pronunciation. For example, several pronunciations
end with /t/, but the spelling is d in Hund or Feld and t in Hut and Bett. Similarly, / / is spelled with sch in schwimmt and with
s in steht, and the /e/ is represented with e in Feld and with in Mdchen.
Consonant Clusters
Table 5.2 shows that there were eight words with consonant clusters in word initial position (sparen, steht, streiten, schwimmt,
schnell, Schnee, Klasse, trgt) and five words with consonant clusters at the end of the word (Wind, Hund, Feld, schwimmt,
trgt). Not included in the end cluster words was arm, because here the r lengthens the vowel and is not pronounced. The words
schwimmt and trgt have a special status because the end clusters result from the addition of a morphemic -t to the stem of the
verb (i.e., schwimm + t) to form the grammatical inflection required by third person singular.
Altogether, phonemically incorrect spellings of both onset and end clusters occurred extremely infrequently. For onset clusters,
only four such misspellings were observed (0.9%). They occurred only for schwimmt (two cases), streiten, and trgt and always
affected interior phonemes of the cluster. These phonemes were either omitted (four cases: e.g., schimt for schwimmt, schreiden
for streiten) or misrepresented (one case: schfipt for schwimmt). The phonemically correct spellings of the onset clusters
apparently were not due to the fact that children simply produced learned spellings for the cluster words from memory. As
evident from the listed misspellings of these words in Table 5.2, each of these words was misspelled by a number of children
and, quite often, the onset was orthographically misspelled: schbaren instead of sparen, schtreiten instead of streiten, and Snee
instead of Schnee. In all these examples, however, the phonemes constituting the onset cluster are present.
For end clusters, it was also the case that misspellings were extremely infrequent, with only seven (2%) misspellings. In four
cases, the interior

Page 91
consonant of the cluster was omitted (e.g., Hudd for Hund, trt for trgt) and in three cases, the final /t/ was omitted. All the
final t omissions occurred for schwimmt and may not be due to a phonological problem. Because schwimmt has quite a number
of letters, children may have forgotten the inflection represented by /t/ when they came to the end of the spelling. This
interpretation is suggested by the finding that not a single omission of the final /t/ occurred for Hund, Wind, Feld, and trgt.
Interesting spelling attempts for schwimmt were schfipt (m omitted) and sweimmnt (no end cluster error), where children
apparently transcribed articulatory features: The /p/ sound results from opening the lips after the /m/ and the /n/ results from
bringing the tongue after the /m/ in the alveolar position to articulate the following /t/. Again, it is important to note that the
close-to-perfect performance for end cluster spellings cannot be due to the fact that children may have produced learned
spellings of these words from memory. This is evident from the listed misspellings of the end cluster words in Table 5.2. For
example, in the case of trgt, the majority of the spellings were orthographically incorrect, but only two phonemically incorrect.
The near absence of difficulties with the spelling of consonant clusters was also observed for a group of 21 children who, at the
end of Grade 2, were identified as suffering from specific spelling difficulties. These children (16 boys, 5 girls) were nominated
by their teachers as showing unexpected spelling difficulties. Additional criteria were performance of at least one standard
deviation below the mean of the Grade 2 norm group on the present spelling test form and a nonverbal IQ (Raven's Colored
Progressive Matrices) of higher than 90. Among these children only eight instances of phonemic misspellings of onset clusters
and three misspellings of end clusters were observed. Again, it seems quite implausible that the near absence of difficulties with
the spelling of consonant clusters could be due to the fact that these children simply produced learned spellings from memory.
The majority of the spellings of the poor spellers (i.e., 57%) was orthographically incorrect.
Consonant Versus Vowel Spelling
In a further step, we examined how the group of 68 first graders spelled consonants and vowels. This comparison was limited to
the first occurrences of consonants and vowels in each word. That is, for all attempted spellings we examined how the initial
consonant and how the following vowel in the word was spelled. For example, in the case of Sommer, we noted how the /s/ and
the /o/ were spelled. In the case of words with initial consonant clusters, only the spelling of the first consonant of the cluster
was considered. The three words without an initial consonant were only evaluated with respect to vowel spelling (ste, arm,
ihr). Spellings were evaluated as or-

Page 92
thographically correct or incorrect (including omissions) and, in the case of an incorrect spelling, we distinguished between
phonemically acceptable and phonemically inacceptable spellings. Table 5.3 shows number and percent of spellings falling into
these categories. Not considered were the thirteen instances (0.8%) where no spelling was attempted. The majority of these
response failures came from one child.
Table 5.3 shows that initial consonants were nearly always spelled correctly, and the few errors were all phonemically
acceptable. The majority of the errors were of the type schtreiten for streiten (see Table 5.2 for misspellings of sparen, streiten,
steht), where the misspelling is actually phonemically more appropriate than the orthographically correct spelling. We also
counted as phonemically acceptable the few instances of drgt for trgt. As already noted, in standard Austrian High German
pronunciation and, even more so in dialect, the /t/ is unaspirated and pronounced more like /d/.
Table 5.3 further shows that the first vowels in the words led to more orthographic misspellings than the initial consonants. The
large majority of the orthographic misspellings were phonemically acceptable. As is evident from Table 5.3, among the
phonemically acceptable misspellings of vowels, the majority were errors related to vowel length. These errors most often were
failures to orthographically mark vowel length at all or, in a few cases, from erroneous markings of vowel length. From the
listed misspellings in Table 5.2, it is evident that doubling the consonant following the vowel, which indicates that the vowel is
short, was often neglected (e.g., Butter was spelled Buter by the large majority of children). Similarly, lengthening the vowel by
the ''silent-h'' or by doubling the vowel letter was also often neglected (e.g., faren for fahren, Schne for Schnee). In some cases,
the long vowel was marked but in an incorrect manner (e.g., Schneh for Schnee, ier for ihr). Another problem of vowel spelling
was that e was used instead of and that eu was used instead of u (e.g., trekt for trgt, Beume for Bume). This is a plausible
error because both e and stand for /e/, and eu and u stand for /oi/, and morphology constrains the possibilities. For example,
in the cases of trgt, Bume, and ste
TABLE 5.3
Consonant and Vowel Spellings
Consonant Vowel
Spellings Number Percent Number Percent
Correct 1,448 98 1,312 78
Acceptable 37 2 319 19
Vowel Length Errors 212 13
Inacceptable 0 0 56 3
Omission 0 0 5 0
Phoneme Quality 0 0 17 1
Vowel Length Errors 34 2

Page 93
consideration of the phonological form of the word stems (spelled: tragen, Baum, Ast) suggests the correct grapheme.
Table 5.3 also shows that inacceptable spellings of vowels were produced, but the number was very small. In only five spellings
was the vowel grapheme omitted. With one exception, it was the e before the l that was omitted (e.g., schnll for schnell, Fld for
Feld). The plausible reason is that in these cases children relied on the letter name /el/ for l. There were a few instances in
which vowel category was misrepresented (e.g., Hand for Hund, strieten for streiten, Bme for Bume). The majority of the
phonemically inacceptable errors as in the case of the phonemically acceptable errors were due to vowel length. For example,
trgt (long vowel) was misspelled as trckt or treckt. This can be considered as phonemically inacceptable because ck usually
marks that the preceding vowel is short. Conversely, Wind (short vowel) was misspelled as Wiend, the ie being the grapheme for
/i:/. One could, of course, argue that the classification of these spelling errors as phonemically inacceptable is overly stringent
because children at the end of Grade 1 may not have had enough learning experience to extract the partly context-dependent
rules of vowel-length marking.
Discussion
The main finding of the reported research is that after about 8 or 9 months of instruction, German children tend to produce
correct or phonemically acceptable word spellings and give little indication of phonologically caused spelling difficulties.
Because German is rather similar to English with respect to phonology, but different with respect to phoneme grapheme
consistency, it can be concluded that the latter factor is responsible for the ease with which German children, in comparison to
English children, learn to spell. The role of instruction also has to be considered but may not be fully independent from
orthographic consistency. Before discussing these broader issues, the main findings are discussed.
Spelling of Vowel Phonemes. In the Introduction, it was noted that specifically for vowels, the German writing system is much
more consistent than English. This higher consistency was reflected in two sets of findings. Study 1 compared how English and
German children spelled the vowels in words with similar pronunciation and spelling in English and German and established
two main findings. One was that German children tended to have less difficulty with correct vowel spelling than English
children. This was due to the fact that difficulties for the German children were limited to the few words where phoneme
grapheme correspondences were ambiguous. Directly related to the difference in consistency was the further finding

Page 94
that German children, faced with a certain vowel with ambigous phoneme grapheme relations, always produced only one or,
maximally, two alternative spellings with higher frequency. The most frequent spelling error of the German children was that
they neglected orthographic markers for long vowels (e.g., Sele instead of Seele, Bot instead of Boot). In contrast, English
children produced a large variety of different vowel spellings when for a word the correct spelling was not obvious. Some of
these misspellings of vowels could not be attributed to phoneme grapheme inconsistency because they consisted of, or included,
consonant letters with no relation to pronunciation. Apparently, these misspellings were due to difficulties with phonemic
segmentation of word pronunciations and/or with understanding how the segments relate to the letters of the given spelling
skeleton. Such errors were nearly absent among the German children.
Study 2 with German children at the end of Grade 1 extended the findings of Study 1 with respect to vowel spelling. Again, the
large majority of the spellings were correct, and the errors were limited to vowels in which the phoneme grapheme relations are
ambiguous. Again, the majority of the misspellings were failures to mark vowel length. Children often failed to double the
consonant letter after the vowel to mark shortness (Buter instead of Butter), and they failed to mark that the vowel is long by
doubling the vowel letter (Schne instead of Schnee) or by inclusion of h after the vowel letter (Ban instead of Bahn). Other
errors related to vowel length were that an alternative to the correct marking for long vowels was used (e.g., ier instead of ihr)
or that orthographic markers for short vowels were used when the vowel was actually long, and vice versa. An important finding
was that misspellings of the type that the category of the vowel was misrepresented (e.g., Hand for Hund) were nearly absent,
as were spelling refusals. Thus, it can be concluded that German children, after about 8 or 9 months of instruction, had no
difficulty identifying the category of the vowel in word pronunciations and used the correct letter(s) for vowel category. There
were also no difficulties when digraphs had to be used for correct vowel spelling.
The present findings on vowel spelling in German fit well with Treiman's (1993) conclusions about the sources of difficulty of
vowel spelling observed among her American first graders. Treiman found that her participants had more difficulty spelling
vowels than spelling consonants and that this difference was related to both the difference in phoneme grapheme consistency
between consonants and vowels and to the difference in the number of letters with which consonants and vowels were spelled.
When these two orthographic factors were accounted for, then the phonological category of consonant versus vowel had no
effect. The present study similarly found that word-initial consonants led to more correct spellings than first vowels. However,
this difference was due mainly to the inconsistency of vowel spelling with respect to vowel length.

Page 95
Consonant Clusters. Difficulties with the correct spelling of consonant clusters is a finding consistently reported for English
children in the first grade (e.g., Snowling, 1994; Treiman, 1993) and for English children with specific spelling problems (e.g.,
Kibel & Miles, 1994). This difficulty is attributed to the phonemic segmentation problem inherent in the spelling of consonant
clusters. It led Treiman to conclude that phonological factors, independent of the inconsistency of English, contribute to early
spelling difficulties. In the case of poor spellers, the observed difficulty of consonant-cluster spelling fits well with the general
phonological impairment account of dyslexia with phonemic awareness difficulties as a leading symptom (e.g., Stanovich,
1994). In relation to these English findings, it was quite astonishing that difficulties with the spelling of consonant clusters were
nearly absent both in normally progressing German children after about 9 months of instruction and in German children with
specific spelling difficulties at the end of Grade 2. The finding that German dyslexic children at least after about 2 years of
instruction have little difficulty with the spelling of consonant clusters is consistent with other findings from our laboratory
showing that dyslexic children exhibit quite high performance levels with respect to pseudoword spelling and phonemic
segmentation, which was tested with a vowel substitution task (Wimmer, 1993).
As already pointed out, there is a plausible explanation for the absence of difficulties with consonant clusters and of other
segmentation-based spelling difficulties among young German spellers and among German dyslexic children. This explanation
does not deny that consonant clusters pose difficulties for phonemic segmentation. It does suggest, however, that such
segmentation difficulties are easily overcome by the combination of a consistent orthography and an instructional regime that
induces children to early word recognition in reading via grapheme phoneme recoding. As previously explained, word
recognition via grapheme phoneme recoding provides systematic segmentation training that makes among other things the
phonemic composition of consonant clusters quite transparent. Obviously, grapheme phoneme recoding can only be an easy to
teach and a reliable word recognition strategy, when the orthography exhibits a high level of consistency between graphemes
and phonemes. In this respect, it is important that the consistency of German is higher in the direction from graphemes to
phonemes than vice versa. The conclusion from the present findings is that the consistency of an orthography may be a more
profound factor for spelling development than generally assumed by English theories. Specifically, the consistency of an
orthography in the reading direction, that is, from graphemes to phonemes, may be as important or even more important for
spelling development than the consistency in the spelling direction, that is, from phonemes to graphemes. The specific effect of
a high level of grapheme phoneme consistency via grapheme phoneme recoding is that

Page 96
the structural relationship between pronunciation and spelling becomes readily transparent, even for children with phonological
impairments.
Of course, the present findings do not allow a direct evaluation of this explanation that in its strongest form implies that
instructional emphasis on word recognition via phonological recoding has a positive effect on spelling only for consistent
orthographies such as German. For such an evaluation and, more generally, for the separation of the role of orthographic
consistency and instructional options on spelling acquisition, it is necessary to study the effects of different forms of reading
instruction (high versus low emphasis on grapheme phoneme coding) for consistent orthographies and also for less consistent
orthographies such as English. Nevertheless, the present findings strongly suggest that a consistent orthography when its
structure is made transparent by instruction provides a more favorable orthographic environment for children's first moves into
literacy than English, and researchers and teachers working within consistent orthographies are well advised not to base their
theories and instructional choices solely on English findings.
Acknowledgments
The manuscript profited from the comments of the editors of the present volume. The research reported in this manuscript was
supported by a grant from the Austrian Science Foundation to Heinz Wimmer (PO9911-H15).
References
Kibel, M., & Miles, T. R. (1994). Phonological errors in the spelling of taught dyslexic children. In C. Hulme & M. Snowling
(Eds.), Reading development and dyslexia (pp. 105-127). London: Whurr.
Snowling, M. J. (1994). Towards a model of spelling acquisition: The development of some component skills. In G. D. A.
Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 111-128). Chichester, UK: Wiley.
Stanovich, K. E. (1994). Does dyslexia exist? Journal of Child Psychology and Psychiatry, 35, 579-595.
Treiman, R. (1993). Beginning to spell. New York: Oxford University Press.
Wimmer, H. (1993). Characteristics of developmental dyslexia in a regular writing system. Applied Psycholinguistics, 14, 1-33.
Wimmer, H., & Goswami, U. (1994). The influence of orthographic consistency on reading development: Word recognition in
English and German children. Cognition, 51, 91-103.
Wimmer, H., & Hummer, P. (1990). How German-speaking first graders read and spell: Doubts on the importance of the
logographic stage. Applied Psycholinguistics, 11, 349-368.
Wimmer, H., Landerl, K., Linortner, R., & Hummer, P. (1991). The relationship of phonemic awareness to reading acquisition.
More consequence than precondition but still important. Cognition, 40, 219-249.

Page 97
Chapter 6
The Development of the Understanding of Number Morphology in Written French
Corinne Totereau
Marie-Genevive Thevenin
Michel Fayol
University of Bourgogne
Studying how to learn and how to use written language morphology is one of the less frequently explored fields of research in
psycholinguistics. Yet, morphology plays an essential role in written French, especially since many written markers have no
corresponding pronunciation (Catach, 1986; Dubois, 1965). This predominantly silent morphology has two general
consequences: (a) For children, the learning of these markers and of their functions must be performed without an oral reference
(e.g., the absence of phonetic realization of the nominal plural -s in les chiens (the dogs) and of the verbal plural -nt in ils jouent
(they play); (b) for adults, the implementation and the control of these markers take place only in reference to the written
language. Despite the crucial character of morphology for the mastery of writing, the studies that concern it remain rare and
mainly involve the counting of spelling errors or developing taxonomies of errors (Chervel & Manesse, 1989; Girolami-
Boulinier, 1984; Jaffr, 1992). In short, there is, to the best of our knowledge, no psycholinguistic approach of the acquisition
and implementation of written French morphology.
Number Morphology in Oral and Written French

The aim of the present series of studies is to make a first contribution to the psycholinguistic approach to the study of
acquisition of French morphology for number in relation to nouns and verbs. This work is concerned

Page 98
with the study of the two most frequent and regular cases: the marking by -s for nominal plurals and by -nt for verb plurals.
Dubois (1965), who compared the distribution of oral and written markers for number in French, showed that in the oral
modality, only the first segment of phrases and sentences (i.e., the determiner) carries a formal variation whereas in writing there
are more markers (e.g., adjective noun agreement) and these tend to apply to the entire phrases and sentences. In spoken French
the determiner the is either le or la for singular nouns and les for plural nouns. This modality difference clearly appears when
sentences (1) and (2) are compared:
(1) La petite poule picore [The small hen pecks]
(2) Les petites poules picorent [The small hens peck]
In the oral modality, only the la/les difference is perceptible because of the deletion of final consonants in spoken French,
whereas in the written modality, all the words are modified. According to Dubois, this asymmetry may be due to the fact that
the secondary function of the redundancy of markers in writing is to ensure the cohesion of texts. This cohesion is insured by the
intonation in the oral modality.
Dubois (1965) also showed that the markers for number are carried by the phrase and the sentence, not the noun or the verb:
The probability of finding a marked item depends on the number of elements from a given syntactic class that can have
variations. In the oral modality, very few nouns (about 50 out of 25,000), verbs, or adjectives have an audible singular plural
difference. Only the most frequent verbs, and especially the auxiliaries, require a phonological variation between the third person
singular and plural (e.g., a/ont [has/have]; est/sont [is/are]; fait/font [does/do]). In fact, the number markers are carried by the
determiner for the nominals, and, to a lesser extent, by the auxiliary (if present) with regard to the verbs. In writing, the situation
is the opposite; very few nouns have the same singular and plural form (e.g., nez [nose]).
The rarity of number markers in oral French and, by contrast, their high frequency in writing mean that French children must
acquire a specific linguistic subsystem at the same time that they are learning the written language. The reference to oral
language is of little use to them, unlike in English. In English, the number differences are orally marked on the noun (e.g.,
dog/dogs) as well as on the verb (e.g., sings/sing) and adjective noun agreement is not marked in oral or written language. As a
consequence, French children must discover the written system of specific features: -s for the nominal and adjectival plurals, -nt
for the verbal plurals (in contrast to in singular form); for the adverbial plurals. Once this discovery has been made, then
the children have to automatize the implementation of inflec-

Page 99
tions (Anderson, 1983). As this automatization takes place without any oral support, it is probably longer and more difficult than
in other languages such as English. It presupposes a regular and frequent practice of writing (Logan, 1988a, 1988b). However,
little is known concerning this acquisition.
The experiments reported here attempt to explore how children from elementary school, and, in particular, first, second, and
third graders, acquire and use the written markers of nominal and verb plurals in comprehension and production tasks.
Brief Review of Studies About the Acquisition of Oral Morphology for Number
Most of the studies dealing with the acquisition of number morphology have been conducted on the English language (see
Ferguson & Slobin, 1973; Kess, 1992; MacWhinney, 1978). Two types of methods have been used in these studies. The first
one consisted of following the same children during several months or several years and analyzing the occurrences of inflections
associated to number in spontaneous conversations in natural settings (Brown, 1973; Cazden, 1972; De Villiers & De Villiers,
1973; Mervis & Johnson, 1991). These longitudinal studies provided evidence that the acquisition of number morphology was
one of the earliest to develop and that it developed approximately in four phases: pre-pluralization; transitional phase without
and then with overgeneralization (i.e., extension of nominal plural markers to other syntactic categories, such as adjectives in
English); generalized use and correct extension to unknown words; and lastly adult-type use (Cazden, 1972; Mervis & Johnson,
1991).
The second method consisted in presenting drawings depicting situations involving one or several objects, persons or actions. In
comprehension tasks, the participants were required to choose the drawing that corresponded to the singular (or plural) item
supplied by the experimenter. One drawing displayed several objects or persons, whereas the other one displayed only one
object or person. In the production tasks, the children were required to produce the oral item corresponding to a drawing
depicting a singular (or plural) situation when they were provided with this same item in the plural (or singular) form (e.g., they
had to say "wugs" when the experimenter provided "wug" or to say "wug" when given "wugs" (Berko, 1958; Brittain, 1970;
Graves & Koziol, 1971; Keeney & Smith, 1971; Levy, 1987; Nicolaci-da Costa & Harris, 1983, 1984; Newfield & Schlanger,
1968; Solomon, 1972).
The results of these studies led to two conclusions and one problem. The first conclusion relates to the fact that the nominal
plural is acquired before

Page 100
the verb plural (Keeney & Wolfe, 1972; Nicolaci-da Costa & Harris, 1984). The precocity of the number marking for nouns in
comparison with verbs is interpreted as corresponding to the fact that numerosity is a property of the meaning of nouns more
than is the case of verbs. Number primarily affects nouns and is then to be copied onto verbs to secure cohesion.
The second conclusion concerns the redundancy of markers. Nicolaci-da Costa and Harris (1983) asked 3- and 4-year-old
children to point drawings out or to act out sentences in which the plural marking was more-or-less redundant (e.g., from one
marker, as in the demonstrative This sheep jumped/These sheep jumped, to several markers as in This boy is driving his
car/These boys are driving their cars). They found that the various markers presented different levels of difficulty in
comprehension but the higher the number of different syntactic markers, the better the performance. However, the redundancy
of information by repetition of the same marker (e.g., The sheep jumps and runs) had no significant effect on comprehension.
The improvement of the comprehension when several markers are involved is due to the fact that having only one marker makes
the processing highly fragile; a brief moment of inattention is enough to lose the marker. The redundancy of syntactic
information would offset the gaps in the information gathering.
The problem raised by several of the previously mentioned studies concerns the relative performance in comprehension and
production. Some authors only tested production (Berko, 1958). Others studied production, comprehension, and even imitation
(Fernald, 1972; Fraser, Bellugi, & Brown, 1963; Keeney & Smith, 1971; Lovell & Dixon, 1967; Winitz, Sanders, & Kort,
1981). However, the results differ from one study to another. Fraser et al. found that imitation (i.e., repeating a sentence)
preceded comprehension (i.e., selecting among two drawings the one which corresponded to a sentence) which preceded
production (i.e., formulating an utterance in connection with a provided drawing). Lovell and Dixon (1967) also found this
hierarchy. Yet several criticisms were directed at the methodology, particularly the manner of comparing comprehension and
production performance (Baird, 1972; Fernald, 1972). In contrast, Keeney and Smith (1971) found that production concerning
the morphology number in relation to verbs preceded comprehension.
In summary, the previous studies concerning the acquisition of oral morphology for number showed that the nominal inflections
are acquired earlier than verb inflections, even in a language such as English in which these inflections are audible. Also, the
studies found that comprehension of singularity/plurality is facilitated by the redundancy of markers. However, there is no
consensus concerning the precedence of comprehension in comparison with production. In the following studies, these
conclusions have been reconsidered and adapted with regard to the study of acquisition of morphology for number in written
French.

Page 101
The Present Study
Even though there are a number of studies dealing with the acquisition of oral morphology for number, the same does not apply
to the written modality. To the best of our knowledge, only one investigation was conducted in English. Beers and Beers (1992)
used Berko's (1958) method by adapting it to written language. As Berko did, they presented singular nonsense words and their
associated drawings to children from first to sixth grades. The students were asked to write the plural forms of these nonsense
words (the oral plural forms were provided). The present research was designed to study the progressive acquisition of -s, the
only plural marker in the written modality despite the diversity of oral plurals (/s/, /z/, /iz/; see Graves & Koziol, 1971). This is a
problem that is not very relevant to the French language because the plural markers are very consistent (almost always /s/). The
authors found that third graders had acquired the inflections corresponding to the various oral markers.
Beers and Beers (1992) provided new information; however, as in Berko (1958), they studied only production and not
comprehension. Moreover, the use of nonsense words with children older than 5 to 6 years raises problems. For example, Levy
(1987) found that, in a pluralization task, 7-year-old children treated nonsense words in a very different way than 2-year-old
children did. In particular, the 7-year-olds made pluralization errors on nonsense words that did not occur with words, whereas
the studies of the oral modality showed that the 3- to 4-year-old children succeeded in these same pluralization tasks. It was as
if the older children interpreted the task relating to nonsense words as differing from the one they had to perform with words.
Finally, Beers and Beers considered only nominal, not verb, plurals. However their study did make a contribution that is
elaborated in the present series of studies.
The present series of studies was designed to study the acquisition and the implementation of nominal and verb plurals in
written French in 6- to 10-year-old children. Berko's (1958) method was adapted to written comprehension and production tasks
using words selected from books used to teach reading in first grade. In the comprehension tasks, children were required to
match nominal groups (e.g., les chiens [The dogs]), nouns presented alone (e.g., sapin [Christmas tree]), short pronoun verb
sentences (e.g., ils sautent [they jump]), or verbs presented alone (e.g., repassent [iron]) in singular or plural form with
drawings depicting scenes involving one or several objects, persons, or actions. In production tasks, children were given orally a
nominal group or a short sentence corresponding to a drawing depicting one or several objects, persons, or actions and were
required to write the words under that drawing.
Three hypotheses were tested. First, because children master oral language before learning the written language, it was expected
that perform-

Page 102
ance on comprehension would be higher than performance on production tasks. In comprehension, children can rely on cues
such as articles; these cues indicate number in the oral modality (le-la vs. les). However, production tasks require the use of
pluralization markers (nominal -s; verb -nt) without corresponding pronunciation. Thus, it is likely that success in these
production tasks will be later and less likely than in the comprehension tasks.
The second hypothesis concerns the noun verb difference. The results of studies dealing with oral language have shown that
nominal plurals were acquired earlier than verbal plurals, even in the English language in which the singular plural difference is
audible (cat/cats; he sings/they sing). The same trend is expected in written French because the numerosity concept is more
clearly associated with the noun than with the verb meaning. This hypothesis predicts that the correct agreements will be more
frequent and occur earlier with nouns than with verbs.
The third hypothesis was developed on the basis of the results of Nicolaci-da Costa and Harris (1983). These authors showed
that a redundancy of markers made comprehension easier. The same trend is expected in written French. Specifically, the
number of correct responses should be higher with phrases (e.g., des chiens [some dogs]) than with nouns presented alone (e.g.,
chiens [dogs]) or with sentences consisting of a pronoun followed by a verb (e.g., ils chantent [they sing]) rather than with
verbs presented alone (e.g., chantent [sing]).
The three hypotheses were tested with first, second and third graders individually over several sessions in the same school year
(Exp. 1) in order to study the acquisition process. The first two hypotheses were again tested in Experiment 2 with a larger
population. In addition, another hypothesis concerning the activation of different markers was tested (for the nouns: in
singular form, -s in plural form; for the verbs: in singular form, -nt in plural form). Indeed, the results of Experiment 1 and
the review of previous studies showed that the acquisition of the comprehension of markers for nominal and verb plurals
occurred over an extended period. However, this performance was obtained with tasks in which children had to match a given
word, either singular or plural (e.g., un chien [a dog] or des chiens [some dogs]) to one of two drawings (e.g., one drawing with
one dog or another one with several dogs). In such a situation, only one condition of linguistic marking (singular or plural) was
explicitly available to the child because only one word was given. The children had to understand that the difference between
one vs. several objects depicted in the two pictures corresponded to different markers: /-s for nouns and /-nt for verbs.
Therefore, in order to successfully choose the correct picture, the child had to activate the complementary marker of the one
given and then choose the correct picture from the two choices. This task of matching only one linguistic item with two
drawings is expected to be harder than the task of matching two

Page 103
linguistic items with only one drawing. Our hypothesis was that children's performance would be better when both different
markers were explicitly given (e.g., le chien/les chiens [the dog/the dogs]) to the children, and they would have to match one
linguistic item (among the two provided items) with only one drawing depicting either one or several objects, persons, or
actions.
Experiment 1
This first experiment was designed to study the acquisition of written French morphology. Six- to 8-year-old children were
administered tests dealing with comprehension and production of written morphology.
Method
Participants. Sixty native French-speaking first, second, and third graders who never repeated a year participated in Experiment
1. There were 20 children from each grade. The mean age of children was 6.4 for first grade (range: from 6.2 to 7.1), 7.4 for
second grade (range: from 7.1 to 8.5) and 8.4 for the third grade (range: from 8.4 to 9.3) at the beginning of the experiment.
Material and Procedure. The material and the procedure were developed from Berko's (1958) and Keeney and Wolfe's (1972)
studies.
In the production tasks, the children had to produce, in writing, the number inflections for the nouns and verbs. For every trial, a
drawing depicting objects, persons, or actions, the corresponding oral formulation, and the written beginning of the phrases (e.g.,
C'est un _/Il y a deux _ [This is a _/ There are two _]) or sentences (e.g., La chemine _/ Les chemines_[The chimney _/ The
chimneys_]) were supplied. (See Figs. 6.1a and 6.1b). The children had to write down the last word of the orally given
phrases/sentences.
In comprehension tasks, children were presented with pictures depicting one or several objects, persons, or actions. Children had
to select from two pictures the one that corresponded to the one linguistic item provided with the drawings. The linguistic item
could take one of four different forms and was either a noun or a verb:
1. The noun presented alone in either the singular form or plural form (e.g., poisson vs. poissons [fish vs. fishes]).
2. The verb presented alone in either the singular form or plural form (e.g., vole vs. volent [flies vs. fly]).

Page 104
Fig. 6.1.
Types of tasks used in Experiment 1 (1a, 1b, 1c, and 1d) and Experiment 2
(1a, 1b, 1c, 1d, 1e, and 1f).

Page 105
3. The noun, either singular or plural, preceded by an article (e.g., un poisson vs. des poissons [a fish vs. some fishes]).
4. The verb, either singular or plural, preceded by a pronoun (e.g., il vole vs. ils volent [it flies vs. they fly]).
See Figs. 6.1c and 6.1d for examples. Each child was presented with six tasks consisting of ten items each, five in the singular,
five in the plural.
In order to track the learning process, the first graders were individually given these tasks twice (in April and June). The second
graders did the tasks five times (in December, January, March, April, and June). Because of their high performance, the third
graders were given the tasks only once (in April).
Scoring. Performance was assessed by counting the number of correct responses (out of 10) for each condition.
Results
Three analyses were performed on the data. First, because we know that the singular form (nominal or verbal) being unmarked
appears first and that each task consisted of five items (out of ten) in the singular, it is possible that a child using only the
singular form could have a score of 5. Thus, we first attempted to determine if the scores significantly differed from 5 (using a t-
test). The results are depicted in Table 6.1. The data show that third-
TABLE 6.1
Experiment 1: Significance of Students' Scores in Relation to 5 (out of 10)
Tasks
NP VP ANC NC PVC VC
First grade
- April no no ** * ** **
- June ** no ** ** ** **
Second grade
- December ** no ** ** ** **
- January ** no ** ** ** **
- March ** no ** ** ** **
- April ** ** ** ** ** **
- June ** ** ** ** ** **
Third grade
- April ** ** ** ** ** **
Note. NP: Noun production; VP: Verb production; ANC: Article + Noun
comprehension; PVC:
Pronoun + Verb omprehension; NC: Noun comprehension; VC: Verb
comprehension.
*p<.05 and **p<.01.

Page 106
graders' performance was always significantly higher than 5 for all tasks. The same was true for second-graders' performance
with all comprehension tasks and the noun-production task but not with the verb-production task until April. For first-graders
also, the results of the comprehension tasks always differed significantly from 5. By contrast, the results from the noun-
production and verb-production tasks did not significantly differ from 5 before June. In summary, in the production tasks, the
unmarked form (i.e., the singular marker) was used for a long time, with verbs even more than with nouns, whereas during the
same time period, the comprehension tasks revealed that a certain number of participants interpreted the presence or absence of
-s and -nt correctly.
Second, several analyses of variance and series of contrasts were performed (using the Sidak test due to the nonorthogonality of
the contrasts; see Abdi, 1987). These analyses were designed to test the significance of the differences between the modalities of
the variables. Results are displayed in Fig. 6.2. The contrasts enabled us to verify the following hypotheses: (a) Performance on
comprehension was always significantly higher than performance on production (all ps < .01); (b) scores on nouns were
significantly better than scores on verbs in production (all ps < .01, except in first grade due to floor effect). In comprehension,
no significant differences were found in third graders due to ceiling effect. In first graders, the differences were significant only
when nouns were preceded by an article and verbs preceded by a pronoun (p < .01). In second graders, the noun-verb difference
was only significant when nouns and verbs were presented alone (p < .05)
Fig. 6.2.
Mean number of correct responses out of 10 as a function of grade and type of
task.

Page 107
TABLE 6.2
Experiment 1: Summary of Results From the Different
Contrasts Testing the Hypotheses (Sidak test)
Grades
Hypotheses Grade 1 Grade 2 Grade 3
H1. Performance on comprehension is higher ## ## ##
than performance on production.
-With the nouns ## ## no(*)
-With the verbs # ## ##
H2. Performance in relation to nouns is higher ## ## ##
than performance in relation to verbs.
-In production no ## ##
-In comprehension
With articles/pronouns ## no(*) no(*)
Without articles/pronouns no # no(*)
H3. Performance on comprehension is better
when there are two marks. ## ## no(*)
-With the nouns ## no(*) no(*)
-With the verbs no # no(*)
Note. (*) is the absence of significance due to ceiling effects; # p < .05;
and ## p < .01.
because a ceiling effect occurred when the nouns were preceded by an article and verbs by a pronoun; (c) the items were
significantly easier in the presence of articles (for the nouns) in first graders (p < .01) and pronouns (for the verbs) in second
graders (p < .05) than in the absence of these markers. The main trends are displayed in Table 6.2.
Finally, according to Siegler's (1987) suggestions, individual profiles were developed session by session, in order to determine
whether the trends observed for the groups were true for individual participants as well. There are three types of patterns an
individual could display that are incompatible with our hypotheses: higher scores on production than on comprehension; better
performance in non redundant situations (noun or verb presented alone) than in redundant situations (article + noun or pronoun +
verb), or better results with verbs than with nouns. We determined that a child showing only one of these three patterns was
enough to reject it as fitting our hypotheses. Only 10 patterns out of 160 (6.25%) were rejected, that is, only 10 patterns violated
one of these expectations. This shows that our expectations were true concerning most individual patterns of performance.
In summary, the individual profiles as well as the global analyses confirm that for the acquisition of French number morphology
for nouns and verbs, comprehension precedes production, nominal plural is acquired earlier than verbal plural, and the
redundancy of markers makes comprehension easier.
The fact that comprehension of plural markers precedes production suggests that the recognition of markers is easier than recall
of the same

Page 108
markers in appropriate contexts. However, considering the simplicity of the marking ( vs. -s for nouns and vs. -nt for
verbs), it seems surprising that comprehension of the number morphology, in particular the verbal one, is late and not strongly
established at the end of first grade. The late success in comprehending written noun and especially verb morphology led us to
envision that at least some comprehension errors would be due to the children's inability to evoke the complementary marker
systematically ( when -s is present, for example, or the reverse) and, thus, to understand that the presence of -s with nouns and
of -nt with verbs (as opposed to ) means that the linguistic item corresponds to a drawing depicting several objects or actions
(instead of one object or action). The following experiment attempted to verify this hypothesis.
Experiment 2
Traditionally, tasks testing comprehension consist in presenting a linguistic item in the singular or plural form and then asking
the participants to select the corresponding drawing depicting a singular or plural situation from two available drawings. In
general, the two pictures only differ in relation to the relevant dimension (here, the numerosity of the sets of objects, persons or
actions). Thus, it is surprising to observe so many errors in written French among 6- and 7-year-old children who have been
using the oral number morphology for a long time. Indeed, even if the written nominal and verb markers are most often not
audible, their number is very limited and their occurrence is frequent and regular. Consequently, they should be acquired very
rapidly, at least in comprehension. However, at the end of first grade as well as at the beginning of second grade, performance
remains weak, in particular for nouns presented alone (about 7 out of 10), pronoun + verb configurations (6.5 out of 10) and,
especially, verbs presented alone (6 out of 10).
It could be that these weak results are due to the fact that the task consisted of the presentation of only one linguistic item
(singular or plural), so the child had to evoke the complementary marker (respectively plural or singular) in order to select the
corresponding illustration. This hypothesis predicts that performance should be better when the task presents the two different
terms (e.g., le chien/les chiens [the dog/the dogs]) and only one picture and then asks the participants to match this picture with
the relevant linguistic item.
Method
Participants. Two hundred and seventy-two French native-speaking children (136 first graders and 136 second graders) from
eight different schools participated in Experiment 2. The mean age of children was 6.6 for first grade (range: from 5.5 to 8) and
7.7 for second grade (range: from 6.8 to 8.11).

Page 109
Material and Procedure. In January, the children were collectively given a series of production and comprehension tasks
concerning the morphological markers for number. The production tasks were the same type as the ones used in Experiment 1
(see Figs. 6.1a and 6.1b). All the comprehension tasks concerned the article + noun or pronoun + verb configurations. However,
in half of the cases, the participants received one linguistic item (singular or plural) that they had to match with one of the two
provided drawings (one vs. several objects, persons, or actions; see Figs. 6.1c and 6.1d). In the other half of the cases, the
participants had two linguistic items (singular or plural) and only one picture available. Thus, the selection focused on the
linguistic item (Figs. 6.1e and 6.1f).
Results
As shown in Fig. 6.3, the findings were the same as in Experiment 1: Performance on comprehension was always better than
performance on production (all ps < .0001), and scores on nouns were better than scores on verbs (all ps < .0001). In first grade,
performance on nouns was better than performance on verbs for both comprehension (7.64 vs. 5.45) and production (3.84 vs.
3.34). The performance difference between nouns and verbs was larger in comprehension than in production. In second grade,
the intensity of the effect was greater in production (7.06 vs. 4.56) than in comprehension (7.99 vs. 7.39). Moreover, consistent
with our hypothesis, performance was better when the participants had two linguistic items to be matched with
Fig. 6.3.
Proportion of correct responses as a function of grade and type of task.

Page 110
one picture rather than one item to be chosen from two available pictures (all ps < .0001). However, this trend was significant
only with verbs, in first grade as well as in second grade (all ps < .0001).
In summary, first graders and second graders knew the number markers and were able to use them correctly (in a matching task
with a drawing) when they had the explicit singular-plural difference available. On the other hand, they failed more frequently
when they were supplied with only one marker. These data suggest that, at least at a certain period of the learning process,
children know the number morphemes, but they are not yet able to activate one of the markers (singular or plural) in the
presence of the other (plural or singular). The fact that this phenomenon occurs only for verbs in our data is probably a
consequence of the earlier acquisition of nominal markers. Studies with younger children are needed in order to find an
equivalent result for nouns.
Thus, in the middle of second grade, children seem to have acquired the French nominal and verbal morphology for number.
They make few errors in comprehension, and they have a high success rate in production when they have to write down only
one word for agreement. However, these results do not imply that these same children are able to use morphemes in more
complex conditions. Indeed, we have shown in some previous experiments that even highly educated adults make errors when
the agreement task is performed under conditions of cognitive overload (Fayol, Largy, & Lemaire, 1994).
Conclusion
The results of the two experiments indicate that there are three phases in the acquisition of written morphology for nominal and
verbal number. First, children process number markers ( vs. -s, vs. -nt) only in comprehension tasks (i.e., in linguistic item
picture-matching tasks), but they do not use them in production tasks. Even then, the use in comprehension is constrained by the
simultaneous presence of the two linguistic items ( vs. -s for nouns; vs. -nt for verbs). It is as if children had acquired the
marker-situation associations (singularity ; plurality -s/-nt), but they were able to interpret these markers only when
both singular and plural forms were present.
However, the first stage occurred only with verbs ( vs. -nt), and plural markers for nouns were understood even by the
youngest children in this study. It is likely that the same trends would appear with nouns ( vs.-s) in an earlier period.
In the second phase, performance on comprehension improved alone, without a corresponding increase on production scores.
The improvement

Page 111
in comprehension resulted from the capacity of children to interpret any number marker ( vs. -s, vs. -nt) without the other
marker being present. The students were not yet able to mobilize the number markers in production, but they had enough
knowledge to understand the meaning of the presence and/or absence of these markers.
The next phase had a different status; it corresponded to the increasing automatization of the use of markers. Indeed, Fayol,
Largy, and Lemaire (1994) found that highly educated adults made agreement errors when they wrote sentences that were
dictated at the same time that they were engaged in a secondary task. These errors did not occur when they were dealing with
the recall task (i.e., dictation) alone. Therefore, adult experts, generally writing without any mistakes, can be made to produce
errors under conditions of cognitive overload. These results suggest that acquiring the knowledge concerning number
morphology is not enough. The use of number inflections must be automatized in order to be able to use the relevant markers at
the right time. However, such an automatization develops over several years. At the very beginning, children who know the
markers and are able to use them when explicitly asked to do so, fail as soon as the inflection use must be performed with a
concurrent task. At this level, even the graphic transcription is enough to disrupt the agreement process (Bourdin & Fayol,
1994). As the automatization increases, people become more and more able to inflect nouns and verbs at the same time that they
are dealing with another activity. However, the success of the agreement implementation remains dependent on the demand
characteristics of the task at hand. Even educated adults can make errors when the written task becomes highly difficult.
The data suggest that the acquisition of written number morphology in French occurs through procedural learning (Anderson,
1983, 1992; Logan, 1988a, 1988b, 1992). The first phase consists of learning the different linguistic markers and matching these
with the corresponding conceptual notions (here singularity vs. plurality). This phase ends when the child can interpret each
marker separately (i.e., in the absence of its complementary marker). The retrieval of the complementary marker indicates a
certain level of automatization (by activation of elements belonging to the same semantic network).
The second phase concerns the transition from a controlled algorithmic application of morphology (e.g., if les or des are present
then the noun takes a -s) to the more and more automatized implementation of agreement. This automatization would result
from the frequent and regular encounter of configurations (e.g., les/des + noun [the/some + noun]; il/ils + verb [he/they + verb]
for example) allowing the matching of relatively consistent conditions with the occurrence of markers (-s/-nt). This
automatization leads to a progressive change of the regulation of the management of agreements

Page 112
from step-by-step control to compilation proceduralization (Anderson, 1983, 1992) and to a reduction of the cognitive cost of
this management as well. At the very beginning, controlled management is cognitively very costly and easily affected by any
concurrent task. This sensitivity to the addition of another task would explain the weakness of performance in written recall.
Children, having to keep the sentences presented orally in memory and to perform the transcription at the same time (Bourdin &
Fayol, 1994) do not succeed in simultaneously managing the control of agreement hence, the difference between the
performance in explicit and implicit agreement is significant. Increased experience with writing and written text can be expected
to reduce the cost of the implementation of certain components which would in turn lead to a progressive automatization. This
automatization corresponds to a direct recovery in memory for the inflection (Logan, 1988a, 1988b), recovery that would be fast
and not very costly in terms of resources. However, the use of the algorithm still remains possible, for example, for the purposes
of control or in situations consisting of contradictory or competing cues (Fayol et al., 1994).
Automatization is progressive and still relative. It remains sensitive to the cost(s) of concurrent tasks. If the participant faced an
additional task, correct agreement might not occur or occurs by default (i.e., in singular form) hence, agreement errors in
complex situations of written production appear. Finally, increasing automatization (and, in particular, with an association such
as a plural item in front of a verb activates the plural verbal inflection) leads to the occurrence of new error types (Fayol et al.,
1994).
This schema of development corresponds to the one that is observed among participants with no difficulties and among
participants in school environments that encourage writing. However, it is likely that some problems remain in certain
populations. Indeed, Nicolson and Fawcett (1990) have shown that a population of 13-year-old dyslexics was characterized,
among other problems, by difficulty in automatizing various abilities, even simple motor skills (e.g., balancing on a beam).
Thus, it can be expected that the participants will struggle to automatize agreements and, consequently, make errors more
frequently when tasks become complicated. Indeed, complication leads to the mobilization of attentional resources. These
resources become unavailable for the management of agreements, which leads to errors. This scenario is all the more plausible
because other studies have shown that morphology was particularly sensitive to stress. Indeed, several studies found that it is
possible, experimentally, to produce selective impairments (through the use of increased cognitive loads) of the morphology that
simulate aphasic disorders (Jarema & Kehayia, 1992; Kilborn, 1991).
The empirical findings of the two present studies and the interpretative context presented enable us to understand the acquisition
of morphology

Page 113
in a synthetic way, some characteristics of adult production, and even a few facts brought to the fore recently with regard to the
disorders of written language.
References
Abdi, H. (1987). Introduction au traitement statistique des donnes exprimentales [Introduction to the statistical treatment of
experimental data]. Grenoble: P.U.G.
Anderson, J. R. (1983). The architecture of cognition. New York: Academic Press.
Anderson, J. R. (1992). Automaticity and the ACT theory. American Journal of Psychology, 105, 165-180.
Baird, K. (1972). On the role of chance in imitation-comprehension-production test results. Journal of Verbal Learning and
Verbal Behavior, 11, 474-477.
Beers, C. S., & Beers, J. W. (1992). Children's spelling of English inflectional morphology. In S. Templeton & D. R. Bear
(Eds), Development of orthographic knowledge and the foundations of literacy (pp. 231-251). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Bourdin, B., & Fayol, M. (1994). Is written language production more difficult than oral language production? A working
memory approach. International Journal of Psychology, 29, 591-620.
Brittain, M. M. (1970). Inflectional performance and early reading achievement. Reading Research Quarterly, 6, 34-48.
Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press.
Catach, N. (1986). L'orthographe franaise [The French spelling]. Paris: Nathan.
Cazden, C. B. (1972). Child language and education. New York: Holt, Rinehart & Winston.
Chervel, A., & Manesse, D. (1989). Comparaison de deux ensembles de dictes 1873-1987 [Comparison of two sets of
dictations 1873-1987]. Paris: INRP.
De Villiers, J., & De Villiers, P. (1973). A cross-sectional study of the acquisition of grammatical morphemes in child speech.
Journal of Psycholinguistic Research, 2, 267-278.
Dubois, J. (1965). Grammaire structurale du Franais: Nom et pronom [Structural grammar of French: Noun and pronoun].
Paris: Larousse.
Fayol, M., Largy, P., & Lemaire, P. (1994). When cognitive overload enhances subject-verb agreement errors. A study in
French written language. Quarterly Journal of Experimental Psychology, 47, 437-464.
Ferguson, C. A., & Slobin, D. I. (1973). Studies of child language development. New York: Holt, Rinehart & Winston.
Fernald, J. (1972). Control of grammar in imitation, comprehension and production: Problems of replication. Journal of Verbal
Learning and Verbal Behavior, 11, 606-613.
Fraser, C., Bellugi, V., & Brown, R. (1963). Control of grammar in imitation, comprehension and production. Journal of Verbal
Learning and Verbal Behavior, 2, 121-135.
Girolami-Boulimier, A. (1984). Les niveaux actuels dans la pratique du langage oral et crit [The present levels in the practice
of oral and written language]. Paris: Masson.
Graves, M. F., & Koziol, S. (1971). Noun plural development in primary grade children. Child Development, 42, 1165-1173.
Jaffr, J. P. (1992). Didactiques de l'orthographe [Didactics of spelling]. Paris: Hachette.
Jarema, G., & Kehayia, E. (1992). Impairment of inflectional morphology and lexical storage. Brain and Language, 43, 541-
564.
Keeney, T. J., & Smith, N. D. (1971). Young children's imitation and comprehension of sentential singularity and plurality.
Language and Speech, 14, 372-383.

Page 114
Keeney, T. J., & Wolfe, J. (1972). The acquisition of agreement in English. Journal of Verbal Learning and Verbal Behavior,
11, 698-705.
Kess, J. F. (1992). Psycholinguistics. Amsterdam: Benjamins.
Kilborn, K. (1991). Selective impairment of grammatical morphology due to induced stress in normal listeners: Implications for
aphasia. Brain and Language, 41, 275-288.
Levy, Y. (1987). The wug technique revisited. Cognitive Development, 2, 71-87.
Logan, G. D. (1988a). Toward an instance theory of automatization. Psychological Review, 95, 492-527.
Logan, G. D. (1988b). Automaticity, resources, and memory. Human Factors, 30, 583-598.
Logan, G. D. (1992). Shapes of reaction-time distributions and shapes of learning curves. Journal of Experimental Psychology:
Learning, Memory and Cognition, 18, 883-914.
Lovell, K., & Dixon, E. M. (1967). The growth of the control of grammar in imitation, comprehension and production. Journal
of Child Psychology and Psychiatry, 8, 31-39.
MacWhinney, B. (1978). The acquisition of morphophonology. Monographs of the Society for Research in Child Development,
43 (1-2, Serial No. 174).
Mervis, C. B., & Johnson, K. E. (1991). Acquisition of the plural morpheme: A case study. Developmental Psychology, 27, 222-
235.
Newfield, M. K., & Schlanger, B. B. (1968). The acquisition of English morphology by normal and educable mentally retarded
children. Journal of Speech and Hearing Research, 11, 693-706.
Nicolaci-da Costa, A., & Harris, M. (1983). Redundancy of syntactic information: An aid to young children's comprehension of
sentential number. British Journal of Psychology, 74, 343-352.
Nicolaci-da Costa, A., & Harris, M. (1984). Young children's comprehension of number markers. British Journal of Psychology,
2, 105-111.
Nicolson, R. I., & Fawcett, A. J. (1990). Automaticity: A new framework for dyslexia research? Cognition, 35, 159-182.
Siegler, R. S. (1987). The peril of averaging data over strategies: An example from children's addition. Journal of Experimental
Psychology: General, 106, 250-264.
Solomon, M. (1972). Stem endings and the acquisition of inflections. Language Learning, 22, 43-50
Winitz, H., Sanders, K., & Kort, J. (1981). Comprehension and production of the / -z/ plural allomorph. Journal of
Psycholinguistic Research, 10, 259-271.

Page 115
Chapter 7
Lexical Spelling Processes in Reading Disabled French-speaking Children
Jesus Alegria
Philippe Mousty
Free University of Brussels
The lexical spelling processes of reading disabled French-speaking children were examined by asking the participants to spell
frequent and infrequent words containing nondominant transcriptions of inconsistent graphonemes (e.g., /k/ spelled ''qu'' as in
"quatre," the dominant spelling for /k/ being "c"). These children were matched with normal controls on their reading ability
level. Spelling at the lowest reading level already depended on word frequency in both groups of children. As reading ability
progressed, the effects of word frequency increased. However, this increase was weaker in the reading disabled than in the
control children. A tentative explanation supposes that word identification processes differ in the two groups. Disabled readers
use partial cues that allow reading but do not supply the orthographic lexicon with complete representations of words.
French orthography belongs to the family of deep orthographic systems (Gleitman & Rozin, 1977; Klima, 1972). It represents
language at the phonological level but simultaneously includes aspects of morphology and syntax. This fact often creates
conflicting situations. For instance, the morphological and syntactic aspects of language not marked at the phonological level
can, nevertheless, be orthographically represented. For example, the second and third person singular as well as the third person
plural are homophonous in several tenses of French verbs: "tu manges" (you eat), "il mange" (he eats), "ils mangent" (they eat)
are all pronounced /mase */. Different spellings, however, are systematically associated with each morpheme: "es," "e," and
"ent" (see Totereau, Thvenin, & Fayol, chap. 6, this volume).

Page 116
French orthography also presents complexities at the phoneme-grapheme translation level. For example, the "graphoneme"1 /v/-
> "v" is consistent: grapho-phonological knowledge is sufficient to correctly spell the phoneme /v/. However, the graphonemes /
/-> "an'' and / /-> "g" are inconsistent, "en" and ''j" are frequent alternatives, respectively. Specific lexical knowledge is
indispensable in order to spell "danger" (danger), instead of "denger," "danjer," or "denjer."
Therefore, in order to spell French correctly, the child must possess and use a number of linguistic abilities phonological,
morphological, lexical, and syntactic that go far beyond the sound spelling transcription rules. A computer simulation study has
shown that pure sound spelling transcription rules allow correct spelling of 88% of phonemes but only about 50% of words
(Vronis, 1988).
It is widely admitted that two basic spelling procedures are necessary to explain the empirical data: the phonological procedure
that constructs spellings using sound spelling transcription rules and the lexical procedure that retrieves spellings of known
words from an orthographic lexicon (cf. Brown & Ellis, 1994; Frith, 1980a). Evidence in favor of this view comes from the
literature about acquired spelling troubles (Zesiger & de Partz, chap. 3, this volume). On the one hand, patients who have lost
the lexical procedure but not the phonological one have been described. These surface dysgraphic patients show a severe
impairment in spelling irregular words, but a reasonably good ability to spell regular words and pseudowords (see Beauvois &
Drouesn, 1981, for a French case and Hatfield & Patterson, 1983, for an English case). On the other hand, phonological
dysgraphic patients show the opposite pattern of disorders. They have severe difficulty manipulating phonological units, as
attested to by their poor performance on pseudoword spelling, but a preserved capacity to spell frequent, even irregular, words
(see Shallice, 1981, for a rather clear case).
Evidence for the dissociation between these two procedures has been also found in developmental disorders. Boder (1973), for
example, proposed a taxonomy including dysphonetic children who have deficits similar to those of phonological dysgraphics
and dyseidetic children who have deficits similar to surface dysgraphics. Of special interest are the cases of adults who show
developmental syndromes similar to the acquired syndromes but who, nevertheless, have reached good literacy levels (Campbell
& Butterworth, 1985; Funnell & Davison, 1989; Goulandris & Snowling, 1991).
Current models of reading and spelling development propose a mechanism involving the articulation of lexical and phonological
procedures. It
1 According to Vronis (1986), a graphoneme is the couple of a grapheme and its phonematic counterpart. The concept of
grapheme corresponds to the functional spelling units of Venezky (1970). So, /s/-> s (serpent), /s/-> c (cigarette), /s/-> t
(notion), and /s/-> sc (science) are examples of French graphonemes.

Page 117
has been suggested that the elaboration of lexical representations results from the activity of the phonological process. The
identification of a particular word using phonology progressively elaborates the corresponding orthographic representation that
would allow its subsequent lexical identification and spelling. The shift proposed by Frith (1985) from alphabetic to
orthographic reading and spelling procedures is a good example of the manner in which these procedures can be articulated in a
developmental model (see Ehri, 1980, 1991; Morton, 1989; Perfetti, 1991).
The development of the phonological and lexical procedures in normal and reading disabled French-speaking children has been
already examined in earlier studies (Alegria & Mousty, 1994, 1996). This chapter further analyzes the development of the lexical
process in the context of reading and spelling disability. We concentrate on the spelling of inconsistent, nondominant
graphonemes, a procedure that inevitably requires the use of the lexical process. The dominance of an inconsistent graphoneme
refers to the frequency of its phoneme-grapheme transcription rule relative to the other possible transcriptions of the same
phoneme. For example, the graphoneme /s/-> "c" is nondominant because "c" is, in French, a less frequent transcription of the
phoneme /s/ than "s'' (/s/-> ''s" is the dominant graphoneme).
Our previous results showed that the performance on such inconsistent nondominant graphonemes was basically nil in beginning
spellers, normal as well as reading disabled, indicating that they did not possess an orthographic lexicon at this early stage of
spelling development. These children systematically used the dominant version of inconsistent graphonemes. The simplest way
to account for these results is to suppose that the first steps in spelling are based on a simplified set of rules introduced most
likely by explicit classroom instruction rather than by reading activity. Afterwards, the evolution of spelling with reading ability
differed in the control and the reading disabled groups. The performance on the inconsistent, nondominant items increased
between second and third grades much more rapidly in the normal group than in the disabled group (Alegria & Mousty, 1994).
This acceleration in the spelling development of such items unambiguously indicates that, during this period of time, the
orthographic lexicon incorporates a large number of relatively infrequent words in normal readers but, for reasons that remain to
be determined, not in disabled readers. Consistent with these speculations, a second study considering normal and disabled
children matched on reading level showed weaker word frequency effects in the latter group than in the former group (Alegria &
Mousty, 1996). If, as proposed earlier, word representations allowing lexical spelling develop as a result of reading activity,
these results suggest that reading processes might differ in normal and disabled readers.
The spelling evolution of four inconsistent, nondominant graphonemes is investigated in this chapter. Part of the data come from
the study pre-

Page 118
viously mentioned (Alegria & Mousty, 1996). Two graphonemes (/s/e, i-> "c," and /z/e, i-> "z") already examined are
reanalyzed together with two new ones: /k/a, o> ''qu," and / /#-> "ain.''2. As stated before, the correct spelling of those
graphonemes depends on the existence of an orthographic lexicon. Besides, it has been hypothesized that the development of an
orthographic lexicon should depend on the number of encounters with the individual written words, so strong frequency effects
were expected as reading level increases. An important question is whether reading disabled and normal children that have
reached the same reading level possess equivalent effects of frequency in word spelling.
Method
Participants
The sample consisted of 38 reading disabled and 75 control children. Reading disabled participants came from a special school
for dyslexic children. According to a routine procedure in Belgium, these children had been detected by an official institution
that decided to send them to this special school. They have reading and spelling problems but are of normal intelligence (they
must have an IQ of at least 85). The group of control children included 19 participants tested at the beginning of Grade 2, 22 at
the end of Grade 2, 18 in Grade 3, and 16 in Grade 5. They came from an elementary school where a phonics teaching method
was used.
For each participant, overall reading efficiency was assessed with a forced-choice sentence completion test (subtest L3 from the
ORLEC battery; Lobrot, 1973). This test consists of 36 sentences with a missing word. Five alternatives are proposed on each
trial and the children have to fill as many sentences as they can in 5 minutes. The score is the percentage of correct responses
obtained in this fixed period of time. As the test goes along, the complexity of the task increases; words become less frequent
and syntactic, cognitive and pragmatic aspects of the sentences become more complex. The test thus provides a global measure
of reading comprehension that includes both specific (efficient written-word identification processes) and nonspecific (general
linguistic and cognitive knowledge) abilities.
Three reading level subgroups of reading disabled and control children were then matched on the basis of their Lobrot reading
scores. Table 7.1 provides the characteristics of these subgroups.
2 The notation adopted here concerning the phonemic part of graphonemes specifies the orthographic context in which
phonemes are included. For example, /s/e, i represents the phoneme /s/ followed by the letter e or i Similarly, / /#
represents the phoneme / / at the end of a word.

Page 119
TABLE 7.1
Characteristics of the Disabled and Control
Subjects Matched on Reading Level
Group Reading Level N Lobrot Scores Age (yrs;mths)
Disabled 1 11 6.8 (0-14) 10;10 (9;4-12;1)
2 8 28.5 (19-33) 12;2 (10;3-13;6)
3 19 54.8 (42-81) 12;9 (10;6-14;5)
Control 1 15 7.2 (0-14) 7;4 (6;9-8;1)
2 28 25.8 (17-36) 7;9 (6;9-11;3)
3 32 57.0 (42-100) 9;9 (7;8-11'4)
Conditions
The spelling task consisted of 20 words included in sentences. Six (or four) words, half high-frequency and half low-frequency,
were selected for each of the four graphonemes: /s/e, i-> "c," /z/e, i-> "z," /k/a, o-> "qu," and / /#-> ''ain" (see Table 7.2). The
classification of words as a function of frequency was made according to frequency counts collected on a corpus of French
literary texts (Trsor de la langue franaise, 1971) and available in the computerized lexical database BRULEX (Content,
Mousty, & Radeau, 1990). Obviously, literary texts do not correspond very well to the printed materials primary school children
usually see. For this reason, a group of 50 fourth graders were presented with the whole list of words used in the experiment.
They were asked to evaluate on a 5-point scale the frequency with which they had earlier seen these printed words: (1) never
seen, (2) rarely, (3) from time to time, (4) often, (5) very often. The results are summarized in Table 7.2.
The four graphonemes examined were inconsistent and nondominant. That is, they all have a competitor that is more frequent.
The first two items, /s/ and /z/, are spelled "s" more often than "c" and "z," respectively. This has been established both on the
basis of statistical counts (Vronis, 1986) as well as empirically in a pseudoword spelling task (Alegria & Mousty,
TABLE 7.2
Mean Subjective Frequencies (on a 5-point scale) for the
Low- and High-Frequency Words Used for Each Graphoneme
Graphoneme High-Frequency Words Low-Frequency Words
/s/-> "c" ciel cirque cinma 3.72 cigale macdoine citerne 2.11
/z/-> "z" douze gazette 2.75 chimpanz trapziste 2.26
/k/-> "qu" quotidien quatre 3.24 quotient marquage 1.83
/ /-> "ain" main bain pain 4.01 fusain poulain tain 1.92

Page 120
1996). The /k/a, o-> "qu" graphoneme has a far more dominant competitor, "c," which represents 96.9% of the occurrences
(Vronis). The graphoneme / /#-> "ain" has a competitor, "in," which is dominant in word endings (42.9% and 31.0% for "in''
and ''ain," respectively). The "in" spelling is even more dominant if its position in the word is not taken into account. In word-
initial position, with the contribution of the morphological "in-" (in words such as "invisible" and "incapable") it reaches 98.8%.
This creates a strong general bias toward the / /#-> "in" graphoneme.
Procedure
Children were tested in groups in their classrooms. They were first presented with the pseudoword-spelling task not discussed in
this chapter. Then the word-spelling task was presented. Items from each condition were randomly mixed within each task. In
the word-spelling task, participants were given an answer sheet on which the carrier sentence was printed with the word
containing the target graphoneme missing. The experimenter read the whole sentence aloud and then repeated the missing word
in isolation. The child was asked to write it down. Only the letter(s) corresponding to the target phoneme was(were) considered
in the analyses (i.e., if the word "cinma" /sinema/ was used to test the /s/e, i-> "c" graphoneme, only the critical letter "c" was
taken into account).
Results
The mean percentages of correct responses per graphoneme for frequent and infrequent words as a function of group and
reading level are presented in Fig. 7.1. A first ANOVA was run with group (control vs. reading disabled children) and reading
level (1st, 2nd, and 3rd level) as between-subject factors, and frequency (low vs. high) and graphonemes (/s/-> "c," /z/-> "z,"
/k/-> "qu," and / /-> "ain") as within-subject factors. As expected, the effects of frequency and reading level were highly
significant: F(1, 107) = 550.73; p < .001, and F(2, 107) = 80.37; p < .001, respectively, as well as their interaction, F(2, 107) =
23.77; p < .001. An important question in this experiment concerned the presence of frequency effects at the lowest reading
level. Partial analysis showed that frequency was already significant at the first reading level, F(1, 24) = 28.76; p < .001. The
main graphoneme effect was significant, F(3, 231) = 34.95; p < .001, as well as its interaction with frequency, F(3, 321) =
31,88; p < .001.
The main group effect was significant, F(1, 107) = 20.89; p < .001, indicating that the percentage of correct responses was
greater in the control group than in the reading disabled group even though their reading level had been

Page 121
Fig. 7.1.
Mean percentage of correct spelling per graphoneme as a function of word
frequency and reading level for reading disabled and control children matched on
reading level.
equated. The group-by-reading-level interaction was significant, too, F(2, 107) = 5.55; p < .001, indicating that the difference
between groups increases with reading level. Two interactions (group by frequency and group by graphoneme), however, were
nonsignificant (Fs < 1).
These results reveal that reading level gives different predictions concerning lexical spelling in normal and disabled readers.
Inspection of Fig. 7.1 suggests that the performance with low-frequency words allows us to predict rather precisely the
performance on frequent words, independently

Page 122
of reading ability, for both normal and reading disabled students. In order to explore this issue, a scatterplot considering the
score obtained on high-frequency words as a function of the score on low-frequency words is shown in the top panel of Fig. 7.2.
The regression equations were basically identical in both groups: slopes reached 1.04 and 0.99, and intercepts 28.3 and 27.6 in
the normal group and the reading disabled group, respectively. The dispersion of the results was slightly larger in the latter
group than in the former group; the Pearson correlation coefficients were 0.60 and 0.78, respectively (ps < 0.01). Finally, an
ANOVA carried out on high-frequency word scores per group, considering low-frequency word scores as covariate, showed that
the group effect was not significant (F < 1).
A parallel analysis considering the reading scores as a function of the performance on low-frequency words was performed in
order to evaluate the predictions about reading differing from group to group. The scatterplot appears in the bottom panel of Fig.
7.2. As expected, the regression equations differed; the intercept was higher in the disabled than in the normal group (26.9 and
14.5, respectively), and the slope was weaker in the former group than in the latter group (0.74 and 0.98, respectively). The
Pearson correlation coefficients were 0.40 and 0.81 in the disabled and the normal readers, respectively (ps < 0.01). The
ANOVA on reading scores per group, considering low-frequency word scores as covariates showed that the group effect was, as
expected, highly significant, F(1,110) = 7.25; p < .01. When the lexical spelling levels were equated, disabled children were
better readers than normal controls.
Conclusion
In order to spell an important proportion of French words, lexical representations containing orthographic specifications are
necessary. Earlier studies (Alegria & Mousty, 1994, 1996) have shown that children at the lowest reading level do not possess
these representations. Children strongly tend to spell inconsistent nondominant graphonemes (i.e., /s/e-i- > "c") using their
dominant transcription (/s/e-i- > "s"). Subsequent studies have compared reading disabled and normal children matched on
reading level. When asked to spell low-frequency and high-frequency words containing inconsistent nondominant graphonemes,
reading disabled children were inferior to normal children, despite their matched reading level (Alegria & Mousty, 1996). The
present study was conducted to extend the comparison of these two groups in lexical spelling.
Previous results suggested that children start spelling with a simplified set of rules, probably acquired through classroom
instructions, and without any lexically based spelling. The present results partially confirm these notions. The results obtained
with inconsistent, nondominant graphonemes

Page 123
Fig. 7.2.
High-frequency word spelling (top graph) and reading (bottom graph)
by low-frequency word spelling, with regression lines for reading disabled
(solid line) and control children (dashed line).

Page 124
are rather poor at the lowest reading level in both groups of children. At this level, they tend to spell using the dominant
graphonemes (/s/- > "s" for instance). Almost 100% of the errors made were of this sort.
Some signs of lexical spelling, however, have been observed at the lowest reading level considered here. Reliable and
substantial frequency effects were already present for the graphoneme / /-> "ain." The reason is probably that the frequent
words containing this graphoneme were indeed the most frequent among those used in the experiment, as indicated by both the
published frequency counts and the subjective frequency evaluations collected among primary school children.
In more general terms, the differences observed between graphonemes can be attributed, at least partially, to differences in the
frequency of the carrier words. In the low-frequency condition, performance was quite homogeneous across graphonemes.
Accordingly, subjective frequency evaluations for these items showed rather small variations between graphonemes. On the
contrary, in the high-frequency condition, such evaluations revealed important differences among graphonemes, and these
differences were reflected in spelling performance. The frequent words containing, for instance, the graphoneme / /-> "ain"
were clearly the most frequent. Therefore, it is not surprising that, for this graphoneme, frequency effects appeared more
precociously and showed a rapid evolution with reading level. The words containing the graphonemes /z/-> "s" and /k/-> "qu''
were the less frequent and, as a matter of fact, produced the poorest spelling performances. An important question for future
research will be to determine whether word frequency alone can fully explain spelling performance. The alternative view
consists in attributing an autonomous role to some graphonemes, possibly in some particular contexts.
The most important fact revealed in the present study was that reading disabled children were poorer at lexical spelling than
normal readers matched on reading level. This suggests that their word identification processes allow them to read but not to
develop an efficient orthographic lexicon. The results of the multivariate analysis are in agreement with this notion. Disabled
readers showed higher reading performance than normal readers paired on spelling level, and this was especially true at the
lowest spelling levels. This finding suggests that more reading exposure is necessary to the disabled readers to achieve the
storage of orthographic word representations. It is interesting to contrast these results with those relating the spelling of
infrequent and frequent words. In this case, disabled and normal readers shared exactly the same regression equation, suggesting
a unique mechanism of spelling in both groups of participants. This does not seem to be the case for reading.
As mentioned previously, the mechanism proposed to explain the development of orthographic representations supposes that the
successful iden-

Page 125
tification of individual words while reading progressively elaborates representations that allow subsequent lexical reading and
spelling (Ehri, 1980; Frith, 1980b; Gough & Juel, 1991; Jorm & Share, 1983; Perfetti, 1991; Stanovich, 1986). According to this
account, phonologically based word identification processes are the dynamic element in the elaboration of lexical
representations of words. The results obtained with normal and disabled readers suggest that reading level, as it was evaluated in
this experiment, gives different predictions in each group concerning the size of the corpus of orthographic representations that
the participant possesses and/or the quality of the items contained in this corpus. Two factors might explain this fact. One
concerns the word-identification processes and the other concerns the storage of orthographic representations.
Let's first consider the notion that disabled and normal readers matched on reading level might differ in their word identification
processes. Some authors have systematically compared reading and spelling abilities in the same children. They have found that
some of them present a homogeneous pattern of success; they are either good readers and good spellers or poor readers and poor
spellers. Some children, however, present a mixed pattern; they are good readers but poor spellers (Bruck & Waters, 1990; Frith,
1980a; Waters, Bruck, & Seidenberg, 1985). A discrepancy between reading and spelling is not surprising. Indeed, spelling
involves recalling, a more difficult activity than reading, which is an act of recognition. Frith quite reasonably noted that reading
can be achieved using partial cues, whereas spelling requires full cues. It has been demonstrated that the mixed children reach
reading levels that are equivalent to those of the good readers-good spellers by using compensatory reading procedures (Bruck
& Waters, 1990; Waters et al., 1985). Their high performance on reading tests was obtained by using sentence context. Several
studies (e.g., Perfetti & Hogaboam, 1975; Perfetti & Roth, 1981; Stanovich, 1980) showed that poor readers' word identification
depended to a greater extent on sentence context. All readers can take advantage of context to predict the nature of the following
items in a sentence, but orthographic word processing in good readers is too fast to allow guessing to be revealed. The basic
feature that characterizes good readers-good spellers is their word processing mechanisms; these mechanisms are complete, (i.e.,
in Frith's terms, based on full cues instead of on partial cues, as is the case for poor readers). On the basis of these data, it is
tempting to speculate that word processing based on full cues is a necessary condition for the elaboration of an orthographic
lexicon that could be used in word spelling.
Another potential explanation for reading disabled students having difficulties developing an orthographic lexicon has been
illustrated by Goulandris and Snowling (1991). According to these authors, surface dyslexia may be due to a nonspecific visual
memory deficit. They have described the case

Page 126
of a successful reader presenting a developmental surface syndrome with severe difficulties in tasks requiring visual recognition
of geometric shapes and reproduction of Greek letters or abstract letter-like sequences. They assume that this impairment has
prevented the participant from developing detailed orthographic representations of words. Their account may, in principle,
explain the use of partial cues in reading. This tendency simply results from the participant's inability to store full cues because
of a general visual memory deficit. This notion contrasts with the influential criticism provided by Vellutino (1979) against
peripheral accounts of dyslexia. However, more recent evidence based mainly on brief presentations and psychophysical
procedures (see Aaron, 1993; Lovegrove, Martin, & Slaghuis, 1986; Lovegrove, Slaghuis, Bowling, Nelson, & Geeves, 1986;
Slaghuis, Lovegrove, & Davidson, 1993) suggested that subtle visual deficits may exist in the reading disabled and that the
crucial variables may not have been clearly identified. In addition, the case reported by Goulandris and Snowling suggests that
peripheral deficits may characterize at least some disabled readers (see the cases reported by Rayner, Murphy, Henderson, &
Pollatsek, 1989, or by Valdois, Grard, Vanault, & Dugas, 1995).
Acknowledgments
The writing of this chapter was supported by the National Fund for Scientific Research (Loterie Nationale, convention
8.4505.92) and by the Ministry of Education of the Belgian French-speaking Community (Concerted Research Action,
convention 91/96-148). Josiane Lechat, Eva Debaix, and Nicole Dedonder participated in data collection. We are grateful to the
staff and children from the schools Etablissement d'Enseignement Primaire Spcial (Court-St-Etienne) and Ecole de la Sainte
Famille (Braine-Lalleud).
Correspondence concerning this chapter should be addressed to Jesus Alegria or Philippe Mousty, Laboratoire de Psychologie
Exprimentale, Universit Libre de Bruxelles, Avenue F.D. Roosevelt, 50, C. P. I. 191, B-1050 Brussels, Belgium. Electronic
mail may be sent to alegria@ulb.ac.be or pmousty@ulb.ac.be
References
Aaron, P. G. (1993). Is there a visual dyslexia? Annals of Dyslexia, 43, 110-124.
Alegria, J., & Mousty, P. (1994). On the development of lexical and non-lexical spelling procedures of French-speaking, normal
and disabled children. In G. D. A. Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp.
211-226). Chichester, NY: Wiley.

Page 127
Alegria J., & Mousty, P. (1996). The development of spelling procedures in French-speaking, normal and reading disabled
children: Effects of frequency and lexicality. Journal of Experimental Child Psychology, 63, 312-338.
Beauvois, M.-F., & Drouesn, J. (1981). Lexical or orthographic dysgraphia. Brain, 104, 21-50.
Boder, E. (1973). Developmental dyslexia: A diagnostic approach based on three atypical reading-spelling patterns.
Developmental Medicine and Child Neurology, 15, 663-687.
Brown, G. D. A., & Ellis, N. C. (1994). Handbook of spelling: Theory, process and intervention. Chichester, NY: Wiley.
Bruck, M., & Waters, G. S. (1990). An analysis of the component spelling and reading skills of good readers-good spellers,
good readers-poor spellers, and poor readers-poor spellers. In T. H. Carr & B. A. Levy (Eds.), Reading and its development (pp.
161-206). San Diego, CA: Academic Press.
Campbell, R., & Butterworth, B. (1985). Phonological dyslexia and dysgraphia in a highly literate subject: A developmental
case with associated deficits of phonemic awareness and processing. Quarterly Journal of Experimental Psychology, 37A, 435-
475.
Content, A., Mousty, P., & Radeau, M. (1990). BRULEX: Une base de donnes lexicales informatise pour le franais crit et
parl [BRULEX: A computerized lexical database for written and spoken French]. L'anne Psychologique, 90, 551-566.
Ehri, L. C. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive processes in spelling (pp. 311-338).
Ehri, L. C. (1991). Learning to read and spell words. In L. Rieben & C. A. Perfetti, Learning to read (pp. 57-63). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Frith, U. (1980a). Unexpected spelling problems. In U. Frith (Ed.), Cognitive processes in spelling (pp. 495-515). London:
Academic Press.
Frith, U. (1980b). Cognitive processes in spelling. London: Academic Press.
Frith U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, J. C. Marshall, & M. Coltheart (Eds.), Surface
Dyslexia (pp. 301-330). London: Routledge & Kegan Paul.
Funnell, E., & Davison, M. (1989). Lexical capture: A developmental disorder of reading and spelling. Quarterly Journal of
Experimental Psychology, 41A, 471-487.
Gleitman, L. R., & Rozin, P. (1977). The structure and acquisition of reading I: Relations between orthographies and the
structure of language. In A. S. Reber & D. L. Scarborough (Eds.), Toward a psychology of reading: The proceedings of the
CUNY conference (pp. 1-53). Hillsdale, NJ: Lawrence Erlbaum Associates.
Gough, P. B., & Juel, C. (1991). The first stages of word recognition. In L. Rieben & C. A. Perfetti, Learning to read (pp. 47-
56). Hillsdale, NJ: Lawrence Erlbaum Associates.
Goulandris, N. K., & Snowling, M. (1991). Visual memory deficits: A plausible cause of developmental dyslexia? Evidence
from a single case study. Cognitive Neuropsychology, 8, 127-154.
Hatfield, F., & Patterson, K. (1983). Phonological spelling. Quarterly Journal of Experimental Psychology, 35A, 451-468.
Jorm, A. F., & Share, D. L. (1983). Phonological recoding and reading acquisition. Applied Psycholinguistics, 4, 103-147.
Klima, E. S. (1972). How alphabets might reflect language. In J. F. Kananagh & I. G. Mattingly (Eds.), Language by ear and by
eye (pp. 57-80). Cambridge, MA: MIT Press.
Lobrot, M. (1973). Lire. Paris: ESF.
Lovegrove, W., Martin, F., & Slaghuis, W. (1986). A theoretical and experimental case for a visual deficit in specific reading
disability. Cognitive Neuropsychology, 3, 225-267.
Lovegrove, W., Slaghuis, W., Bowling, A., Nelson, P., & Geeves, E. (1986). Spatial frequency processing and the prediction of
reading disability: A preliminary investigation. Perception and Psychophysics, 40(6), 440-444.
Morton, J. (1989). An information-processing account of reading acquisition. In A. M. Galaburda (Ed.), From reading to
neurons (pp. 43-68). Cambridge, MA: MIT Press.

Page 128
Perfetti, C. A. (1991). Representations and awareness in the acquisition of reading competence. In L. Rieben & C. A. Perfetti,
Learning to read (pp. 33-44). Hillsdale, NJ: Lawrence Erlbaum Associates.
Perfetti, C. A., & Hogaboam, T. (1975). The relationship between single word decoding and reading comprehension skill.
Journal of Educational Psychology, 67, 461-469.
Perfetti, C. A., & Roth, S. F. (1981). Some of the interactive processes in reading and their role in reading skill. In A. M.
Lesgold & C. A. Perfetti (Eds.), Interactive processes in reading (pp. 269-297). Hillsdale, NJ: Lawrence Erlbaum Associates.
Rayner, K., Murphy, L. A., Henderson, J. M., & Pollatsek, A. (1989). Selective attentional dyslexia. Cognitive
Neuropsychology, 6, 357-378.
Slaghuis, W. L., Lovegrove, W. J., & Davidson, J. A. (1993). Visual and language processing deficits are concurrent in
dyslexia. Cortex, 29, 601-615.
Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual differences in the development of reading
fluency. Reading Research Quarterly, 16, 32-71.
Stanovich, K. E. (1986). Matthew effects in reading: Some consequences of individual differences in the acquisition of literacy.
Reading Research Quarterly, 21, 360-407.
Trsor de la langue franaise [Treasure of French language]. (1971). Paris: Klincksieck.
Valdois, S., Grard, C., Vanault, P., & Dugas, M. (1995). Peripheral developmental dyslexia: A visual attentional account?
Cognitive Neuropsychology, 12(1), 31-67.
Vellutino, F. R. (1979). Dyslexia: Theory and research. Cambridge, MA: MIT Press.
Venezky, R. L. (1970). The structure of English orthography. The Hague: Mouton.
Vronis, J. (1986). Etude quantitative sur le systme graphique et phono-graphique du franais [Quantitative study upon the
graphic and phonographic system in French]. Cahiers de Psychologie Cognitive [European Bulletin of Cognitive Psychology], 6,
501-531.
Vronis, J. (1988). From sound to spelling in French: Simulation on a computer. Cahiers de Psychologie Cognitive [European
Bulletin of Cognitive Psychology], 8, 315-334.
Waters, G. S., Bruck, M., & Seidenberg, M. (1985). Do children use similar processes to read and spell words? Journal of

Page 129
Chapter 8
Learning to Spell in the Classroom
Linda Allal
Skill in spelling, even more than skill in reading or in mathematics, is primarily a product of school learning. One might argue
that all research on spelling acquisition, even when conducted in a developmental perspective, is invariably concerned with
learning shaped in a classroom setting. The focus of the present chapter is on the conditions of learning to spell in the
classroom, particularly during the first 6 years of schooling. It examines alternative instructional approaches, proposed activities
and tools, aims and observed outcomes. The chapter is based on literature reviews and presentations of research conducted both
in English-speaking and in French-speaking countries. Without attempting to cover all curricular trends, it identifies the main
features of the most widely adopted instructional methods.
Over the past 20 years, a large number of publications have presented instructional approaches designed to foster students'
writing competency (e.g., Chiss, Laurent, Meyer, Romian, & Schneuwly, 1987; Dipardo & Freedman, 1988; Englert, Raphael, &
Anderson, 1992; Graves, 1983; Scardamalia & Bereiter, 1986; Turco, 1988). In most cases, priority is given to higher order
processes of text production (planning, revising, sentence generation), whereas lower order processes such as spelling tend to be
neglected. Researchers in the area of language instruction often seem to share the attitudes of teachers and students for whom
spelling is a necessary but altogether disliked component of the school curriculum (Downing, DeStefano, Rich, & Bell, 1984).
As a school subject, spelling can cover varying ranges of content, thus making it difficult to compare research conducted in
different countries.

Page 130
Studies of the acquisition of French spelling generally consider two areas of orthographic knowledge as defined and taught in
school:
1. Lexical or usage aspects of spelling (phoneme-grapheme transcription as well as morphemic and ideographic features of word
construction),
2. Grammatical aspects of spelling (morphemes specifying relationships, particularly agreements, between sentence components
or between components of different sentences).
These aspects of spelling are closely linked to topics studied in other domains of the French language curriculum: grammar, verb
conjugation, and punctuation. Thus, spelling instruction, broadly defined, often extends to all conventions of written language
transcription. Studies of English spelling acquisition sometimes have a similar scope when they investigate mastery of writing
mechanics (a rather inappropriate but widely used term referring to knowledge of spelling, syntax, and punctuation conventions).
However, most English language studies focus exclusively on the spelling of individual words. In addition to obvious
differences between the French and English languages,1 differences in instructional traditions influence the orientations and
methods of spelling research, as is subsequently shown in this chapter.
The chapter is organized in four sections. The first section presents our conceptual framework that distinguishes two categories
of instructional approaches as well as four basic features of teaching-learning activities in the classroom. In the next two
sections devoted to each instructional approach, we analyze research findings on spelling acquisition in English and
1 An example will help to show why French spelling instruction must grapple with a wider range of problems than those
encountered in English. In the following sentences, spelling difficulties that occur in French but not in English are
underlined: (a) The neighbor's daughters sing beautiful Greek songs; (b) Les filles de la voisine chantent de belles
chansons grecques.
In addition to the lexical aspects of spelling that can be quite challenging in both languages (e.g., neighbor, chanson), the French
sentence entails the following orthographic difficulties linked to inflectional morphology: (a) plural inflections on nouns are
usually not marked orally (fille and filles, chanson and chansons are pronounced the same); (b) third-person plural verb
inflections are often not marked orally (chante and chantent are pronounced the same); (c) most determiners and adjectives have
to agree with nouns in number and in gender and the inflections are often not marked orally (same pronunciation for bel, belle,
belles, and for grec, grecs, grecque, grecques); (d) modifiers separating the subject and the verb occur more often in French and
thus increase the likelihood of incorrect verb inflections (voisine chante instead of filles chantent).
Moreover, third-person referential pronoun agreement is often not marked orally in French and increases the probability of
incorrect plural inflections, for example: (a) ''The neighbor's daughters. . . .Having lived here for 20 years, they speak French
fluently." is clearly distinguished from " . . . Having lived here for 20 years, she speaks French fluently." (b) On the other hand,
"Les filles de la voisine. . . .Habitant ici . . . , elles parlent. . . ." cannot be distinguished orally from "Habitant ici . . . , elle
parle . . . "

Page 131
in French. A brief concluding section raises the question of possible links between different approaches to spelling instruction.
Instructional Approaches and Features of Classroom Learning

Instructional approaches for spelling acquisition can be classified in two broad categories, as shown in Fig. 8.1. The first type of
approach aims at specific learning of spelling skills. Although advocates of this approach recognize that spelling problems arise
and must be minimally dealt with in all reading and writing activities in the classroom, they consider it essential to organize a
substantial body of instruction focused specifically on spelling. This approach is based on the assumption that spelling
knowledge acquired in specific instruction will transfer to more complex reading and writing tasks. Specific instruction based on
memorization, various types of exercises, and word and text dictation is intended to develop skills that will be reinvested in
subsequent situations of written communication.
In the second instructional approach, spelling acquisition is integrated in activities of text production or in situations involving
both reading and writing. This approach aims at creating conditions for spelling acquisition within an authentic context of
communication. Writing tasks are both the departure point and the overall framework for studying various aspects of spelling.
Although specific instructional activities, such as exercises, may be used to consolidate spelling skills by providing additional
practice on problems identified in the writing tasks, they are proposed as follow-ups on writing rather than as self-contained
learning situations. The underlying
Fig. 8.1.
Two instructional approaches for spelling acquisition.

Page 132
assumption of the integrated approach is that extensive writing activities will progressively lead to a functional organization of
the student's orthographic knowledge and skill. Although teachers may adopt mixed instructional strategies combining some
aspects of both specific and integrated approaches in classroom practice, it is useful to differentiate the contrasting conceptions
underlying each approach. In order to specify these conceptions, we consider four important features of teaching-learning
processes in the classroom. The following is a brief, general definition of each feature. The usefulness of these features for
comparing different approaches to spelling instruction is shown in subsequent sections of the chapter.
1. The concept of didactical transposition2 offers a fruitful theoretical framework for analyzing instructional content. Introduced
by Chevallard (1985) and developed by Bronckart and Schneuwly (1991) in the field of language instruction, this concept
describes the transformations between successive knowledge states that account for the content structure of disciplines taught in
school. Scientific knowledge and relevant social practices in a given domain constitute the initial reference points from which
the subject matter to be taught in school is selected and specified. The selected topics are then consolidated into an autonomous
configuration as the instructional content of a particular curriculum. We find it useful to extend the sequence to include two
additional transpositions: (a) the transformation of curricular content into subject matter actually taught (or transmitted by
instructional materials), and (b) the transposition of what is taught into knowledge in fact acquired by students.
2. The contextualization of school learning is a long-standing concern (Dewey, 1938/1963) that is receiving new emphasis in
contemporary work on "situated cognition" (Brown, Collins, & Duguid, 1989). The basic assumption is that declarative and
procedural knowledge acquired in school will be more easily activated in out-of-school settings if the conditions of class-room
learning closely approximate the real-world contexts of knowledge utilization. This assumption is grounded in a sociocognitive
conception of learning as "apprenticeship" in a set of cultural practices (Collins, Brown, & Newman, 1989). In contrast, the
highly decontextualized tasks usually employed in school are derived from two interrelated phenomena: (a) the process of
"content reification" demonstrated in research on didactical transposition and (b) the classical conception of learning as an
accumulative process proceeding from the mastery of basic skills to their application in more complex situations (Needels &
Knapp, 1994).
3. Studies by Sweller (1988; Sweller & Chandler, 1994) showed that schema acquisition and automation are affected by the
cognitive load induced by
2 We have made the translations of all quotations from French-language references.

Page 133
two aspects of an instructional activity: (a) its intrinsic properties (in particular, the number of elements to be dealt with and the
degree of interactivity among elements) and (b) the extraneous properties of the instructional format. This research shows that
instructional design that modulates the cognitive load of a task is a powerful means of optimizing learning. In the area of
language learning and utilization, problems of cognitive overload can explain certain types of errors committed by experts who
master a spelling rule but fail to apply it (Fayol & Largy, 1992). Moreover, studies of individual differences in writing
(McCutchen, Covill, Hoyne, & Mildes, 1994) have suggested that increasing fluency of so-called lower level skills (sentence
generation, lexical retrieval) can reduce working memory load and thereby allow writers to invest more cognitive ressources in
higher level processes such as online planning and reviewing.
4. Investigations of school learning show that both cognitive and metacognitive aspects of self-regulation play an important role
in knowledge acquisition and in the use of existing skills to carry out new tasks (Allal & Saada-Robert, 1992). As
conceptualized by Vygotsky (1978), the learner's regulation strategies are developed by a process of internalization, defined as
"internal reconstruction" of socially mediated regulations (p. 56). In the classroom, this mediation can be provided by
interactions with the teacher, or with more advanced peers, or by other forms of instructional scaffolding (interactive material or
tools).
Specific Instructional Approaches to Learning to Spell
Learning to Spell in English
Specific instructional activities for learning English spelling are nearly always focused on individual words, isolated from any
surrounding linguistic context, or in cases where words are inserted in sentences or in a text, the context has a limited effect on
how each word is spelled. A long-standing debate exists, however, concerning the relative effectiveness of two forms of
instruction: (a) memorization of lists of words, or (b) systematic analysis of spelling regularities by a series of appropriate
exercises.
Memorizing Word Lists. A century ago, Rice launched a large-scale survey of spelling ability based on tests administered to
13,000 children in fourth through eighth grades and on interviews with 200 teachers (Venezky, 1980). Although his data showed
no direct relationship between teaching method and learning outcomes, the findings led him to two conclusions regarding
prevailing instruction: neither rote memorization of arbitrary word lists, nor

Page 134
the teaching of spelling rules, accompanied by endless lists of exceptions, could be considered as truly effective. Rice's
recommendations retained memorization of word lists as the basic approach to spelling instruction but proposed several ways of
organizing the lists (e.g., grading words by orthographic differences and by use) and suggested that a limited number of basic
rules be taught (e.g., rules for suffix morphology). Studies by Horn (1919, 1960), conducted over a period of 40 years, showed
that teaching rules of phoneme grapheme correspondences had little positive impact, thus reinforcing the idea that spelling
acquisition must rely essentially on memorization of a large body of words.
In a review article, Graham (1983) presented the principles that can be derived from the numerous studies of spelling instruction
based primarily on word-list memorization. These approaches consider spelling to be a complex but relatively independent skill
that does not develop by transfer from other curricular areas (Allen & Ager, 1965). Spelling acquisition, therefore, requires
specific instructional activities distinct from classroom situations that focus on the communication functions of reading and
writing.
For Graham (1983), the essential instructional objective is acquisition of a ''basic spelling vocabulary" (p. 562) (i.e., a stock of
words that can be automatically retrieved in memory). Memorization of word lists is the primary means of attaining this
objective, but it can be usefully supplemented by instruction in a small number of key rules that apply to a large number of
words. In addition, the student is supposed to acquire two skills dictionary use and proofreading that facilitate use of
orthographic knowledge in writing activities.
The instructional sequence proposed by Graham (1983) for elementary school Grades 1-6 is organized in four main phases:
1. The lists to be studied are composed primarily of words frequently used, or likely to be used, in student writing projects. The
teacher can create lists based on tabulations of words actually used in class or can refer to published lists. Following
recommendations from Horn's (1954) research, words are presented in columns rather than being inserted in sentences or in
paragraphs. Each word is presented as a whole, unsegmented unit, without syllabication or indications (e.g., colors, typography)
drawing attention to so-called "hard spots."
2. An initial test, usually based on transcription of dictated words, allows identification of the words spelled correctly and
incorrectly by each student. The mistakes are corrected by the child rather than by the teacher. Subsequent study efforts focus on
the individual list of misspelled words compiled for each student.
3. Word study is based on a systematic technique for effective memorization. Graham (1983) presents five experimentally
validated techniques,

Page 135
including the Horn (1954) "Method 2" that includes the following steps (p. 565):
1. Pronounce each word carefully.
2. Look carefully at each part of the word as you pronounce it.
3. Say the letters in sequence.
4. Attempt to recall how the word looks, then spell the word.
5. Check this attempt to recall.
6. Write the word.
7. Check this spelling attempt.
8. Repeat the above steps if necessary.
The study technique is presented on a worksheet that the student learns to use with teacher guidance until the steps are
internalized; the student can then proceed at an individual pace without supervision. The effectiveness of the technique is
explained by its reliance on combined visual, auditory, and kinaesthetic representations of word structure.
4. After a study period, the student is retested on the word list, and the words misspelled are incorporated into the next list (or a
later list) to be learned.
The effectiveness of the approach comes from the systematic repetition of the "test-study-test" cycle (Fitzgerald, 1953; Yee,
1969). In addition, Graham (1983) proposed the use of spelling games such as hangman, scrambled words, and spelling bingo,
but he insisted on the fact that they should "only supplement rather than supplant" systematic word memorization (p. 565). He
also advocated tutoring as a means of reinforcing the efficiency of individualized word study.
We can summarize the characteristics of the word memorization approach to spelling as follows: Strong emphasis is placed on
individualized instruction. An initial list is proposed to the entire class, but very rapidly thereafter, different lists are constituted
for each student on the basis of his or her spelling mistakes. It is also a highly structured approach that is "planned, monitored
and modified on the basis of assessment information" (Graham, 1983, p. 563). Student motivation is encouraged by the study of
words likely to be rapidly reused in writing, by the progress resulting from application of the study technique and, to a lesser
extent, by a certain diversification of activities (such as the introduction of games). On the whole, this approach involves many
classical features of mastery learning. It is not, however, incompatible with cooperative learning methods involving team study
and peer tutoring (for example, the CIRC program developed by Slavin, Stevens, & Madden, 1988).

Page 136
Study of word lists is very widespread in elementary schools, but many teachers do not apply the principles that assure
instructional effectiveness. Ineffective practices include lack of individualization (all students study the same lists), badly
designed exercises (students manipulate words without engaging in sufficient spelling practice), and limited content relevance
(too much time spent learning low-frequency words rarely used by students in their writing).
Systematic Study of Spelling Regularities. In contrast to the research by Horn (1954) showing the lack of regularity of English
spelling, studies based on more sophisticated linguistic models have provided a more optimistic view of the possibility of
defining rules that can predict a major portion of phoneme grapheme correspondences by taking into account variations in
position, environment, and stress (Hanna, Hanna, Hodges, & Rudorf, 1966). These studies lend support to the development of
instructional approaches focused on spelling patterns, that is, on regularities of sound letter relationships and on underlying
linguistic rules that account for regularities. For Hodges (1982), "learning to spell involves developing an understanding of the
total framework of English orthography and the interrelationships among phonological, morphemic, and other language factors
which the orthography reflects" (p. 289).
An instructional approach based on systematic study of spelling patterns has been proposed by Henderson (1985) and Templeton
(1991). In this approach, spelling is not considered as a specific skill but as one manifestation of "word knowledge"
progressively acquired through all reading and writing activities conducted in the classroom (Templeton, 1991). It is believed,
however, that in addition to these activities, teachers must provide explicit spelling instruction focused on words studied out of
context. Direct instruction, conducted in small groups or with the entire class, aims at allowing the student to discover, analyze,
and manipulate orthographic patterns and rules. The purpose is to insure systematic consolidation of knowledge acquired
partially and implicitly through reading and writing.
Developmental studies of spelling acquisition (Henderson, 1985; Templeton, 1989, 1991) have provided a foundation for
specifying the progression of subjects to be studied during the elementary school grades. The recommended sequence starts with
basic letter patterns within one-syllable words, goes on to variations of spelling linked to phoneme position, environment and
stress, then to polysyllabic words and patterns at syllable juncture, adds the study of morphemic elements (e.g., prefixes,
suffixes) and, finally, offers more advanced activities dealing with the relationships among meaning, morphology, and
derivation.
Three types of instructional activities are included in this approach (Templeton, 1991). First, it is accepted that a certain amount
of word memoriza-

Page 137
tion must take place, especially at the beginning, so the student can acquire a sufficiently large sight vocabulary to which word
analysis techniques can then be applied. However, in contrast with the memorization approach described previously, words are
grouped in lists to show their regularities, the systematic variations of letter patterns, or derivational relationships.
Second, during the lessons given by the teacher, the students learn to observe, compare, decompose, and construct words.
Importance is given to the discovery of patterns inferred by students on the basis of the manipulations they carry out. Their
activity is, nevertheless, strongly guided by the teacher's interventions and by the pre-established structure of the material
(corpus of examples). Because this material is constructed to facilitate the intended discoveries, it does not necessarily
correspond to the spelling difficulties that are commonly encountered in the students' personal writing.
Third, in order to consolidate the discoveries made during spelling lessons, the students apply the same principles in various
types of individual exercises (fill-in-the-blank, matching, sorting, transformations, construction, or completion of classifications
and analogies, etc.). Lessons and exercises are designed not only to foster students' spelling skill in itself but also "the
construction of abstract knowledge structures that will underlie their advancing word knowledge and afford more efficient
effective reading and more elaborated writing" (Templeton, 1991, p. 198). Some programs of instruction adopt a more eclectic
view. For example, the aforementioned activities of word sorting, analysis, and classification may be carried out on material of
the type usually employed in the classical memorization approach, that is, on an individualized bank of words belonging to the
student's reading and writing vocabulary (Sulzby, 1980).
Learning to Spell in French
Research on the acquisition of French spelling has several long-standing orientations (Jaffr, 1992): theory building based on
findings from linguistics, psycholinguistics, and cognitive psychology; descriptive research using comparative, correlational, or
historic methods; action research aimed at implementing new instructional approaches. There exist, on the other hand, relatively
few experimental or quasi-experimental field studies similar to those conducted in English language countries as a means of
verifying the effects of different instructional conditions on student learning.
Three specific instructional approaches are used in teaching French spelling. For the lexical aspects of spelling and for verb
conjugation, word memorization is often combined with the systematic study of orthographic regularities. These approaches
relied, for more than half a century, on large-scale surveys giving guidelines for sequencing topics according to their difficulty

Page 138
level as defined by average age of acquisition. One can mention, as examples, the Dubois-Buyse word scale, based on a survey
conducted between 1936 and 1939 with nearly 60,000 Belgian students and periodically revised since then (Guion, 1974), and
the conjugation tables established by Roller (1954) on the basis of surveys of Genevan school children. In French language
instruction, the study of orthographic regularities extends over a wider range of content than in English spelling instruction: In
addition to lexical patterns, a sizeable number of rules governing the grammatical aspects of French spelling are analyzed by
various types of inductive and deductive exercises. Moreover, French spelling instruction relies heavily on a device generally
unknown in English language classrooms: la dicte.
A dicte usually entails dictation of a coherent text excerpt (e.g., a paragraph) composed of several sentences. The length and
the complexity of the texts vary: In the beginning elementary grades, only two or three short, basic sentences are dictated,
whereas in the formal competitions for adult specialists, the texts are composed of lengthy, interconnected sentences involving
numerous orthographic complexities. Traditionally, dictation texts are drawn from literary sources, but they are sometimes
invented by teachers so as to limit (or, on the contrary, maximize) the number of difficulties the student will encounter.
Text dictation has been used for more than a century as an instructional exercise and as a means of assessing the knowledge
acquired by students in more restricted activities focused on word memorization or analysis of spelling regularities. Compared to
these latter activities, text dictation offers one undeniable advantage: The student must link together all aspects of spelling
(lexical and grammatical knowledge) in a text-based framework that requires reflection not only on sentence components but
also on relationships between sentences. Text dictation allows treatment of aspects of French spelling that cannot be dealt with in
exercises carried out on isolated words or unconnected sentences. But, beyond this specific instructional advantage, la dicte has
often been endowed with a much more ambitious educational mission: a cultural function of conveying feeling for bel usage as
well as other unverifiable virtues such as contributing to "the child's corporal and mental mastery" (Vial, 1970, p. 85).
In spite of recurrent criticisms of text dictation as a pedagogical tool, it continues to occupy an important place in the Official
Directives of the French school system (Guion, 1974) and is the most frequently declared means of teaching spelling in surveys
of teachers (Michelet, 1984, cited in Chervel & Manesse, 1989). The effectiveness of text dictation on student learning has not,
however, received thorough investigation by well-controlled empirical studies in the classroom. Chervel and Manesse have
combined two interesting lines of research: (a) a meticulously documented historical study of the role of la dicte in French
elementary and secondary

Page 139
schooling, and (b) a survey comparing student performance across the span of a century (3,114 dictation copies produced
between 1873 and 1875 and an equivalent number based on the same text collected in 1987). The survey invalidates the
persistent myth of declining spelling skills among today's schoolchildren, but the historical study is even richer in instructional
insights. We summarize several of the key findings.
Before the introduction of text dictation, around 1830, prevailing instruction consisted primarily of copying and memorizing
isolated words or sentences. Several aspects of this classical approach were similar to modern word memorization methods: The
student worked individually, with a minimal amount of direct teacher guidance, on material that varied from one learner to
another. However, in contrast with modern methods, little concern was given to student comprehension of the content to be
memorized. "One often sees students open randomly any book they come across and then undertake the thankless task of blindly
transcribing words and sentences whose meaning is totally lost on them," reported Inspector Cadet in 1881 (Chervel & Manesse,
1989, p. 126). In this context, it was not surprising that the introduction of oral dictation of texts commented by the teacher
constituted a major instructional innovation, vigorously defended by the school administrators and pedagogues of the period.
Chervel and Manesse (p. 136) explained the progressive adoption of the new method throughout the school system in the
following terms:
/La dicte/ has one merit that ranks it very much above other exercises. It obliges the teacher to break completely with the
"individual mode" of instruction that was prevalent before 1850. It obliges him to direct his class, one might even say, to
"teach to his class." . . . It allows him to use the blackboard, for which a lively campaign promoting its introduction in the
classroom was underway at the same period.
From the mid-1800s on, published collections of dictation texts, graded by difficulty level, fostered increasing use of this
instructional tool by teachers.
The effectiveness of this new approach, which by 1870 had become almost a daily feature of French classrooms, was confirmed
by impressive increases in student spelling skills across the nation, as substantiated by inspectors' reports and by results of
elementary school final examinations. We find it interesting to relate this observation to explanatory variables identified in
contemporary research on teaching. In studies based on the process product paradigm (Brophy & Good, 1986), it has been found
that student-learning outcomes are strongly dependent on two variables (among others): (a) engaged time spent on a given task
and (b) direct instruction carried out with expertise and enthusiasm by the teacher. It is intriguing to consider that the
observations regarding the effectiveness of dictations, as

Page 140
formulated by French inspectors between 1870 and 1880, may constitute an early validation of the role of these two key
variables of process product research.
Today's approaches to French spelling instruction (e.g., Chiss & David, 1992) focus primarily on linguistic and psychological
principles that can guide interpretation of student answers and explain spelling errors. Whatever the task (memorization,
analytical exercises, text dictation), an attempt is made to discover student representations of the orthographic system and to
develop instructional interventions on this basis. The theoretical contributions of the HESO group (Catach, 1978) offer a
framework for the development of error typologies and for the construction of instructional tools (Jaffr, 1992). This perspective
tends to give more importance to the presumed validity of scientific references than to controlled empirical testing of
instructional conditions in the classroom.
Comparison of Approaches for Specific Spelling Instruction
The four features we have chosen to describe teaching learning processes in the classroom allow us to compare the approaches
aimed at specific spelling instruction, that is, word list memorization (approach M), systematic study of spelling regularities and
rules (approach R), and text dictation (approach D).
Didactical Transposition. In the M approach, the scientific knowledge base is codified (spelling specified in the dictionary) and
the instructional content is often prescribed institutionally (word lists defined in curricular materials), but these basic reference
materials are accompanied by little or no theoretical elaboration. In contrast, the R approach refers explicitly to scientific
knowledge in several areas of linguistics (phonology, morphology, etc.) and to even broader cultural or literary concerns (e.g.,
relations between meaning and etymology). The transposition of developmental data guides the progression of instructional
content, and the activities to be carried out (analysis, classification, etc.) are embodied in specific instructional tools (teacher
manuals, textbooks, workbooks). Teachers need sufficient training in the underlying theoretical models to be able to create
classroom activities that go beyond the examples provided in the curricular material and to interpret children's spelling errors
and strategies.
The D approach lies at the crossroads of several tendencies in didactical transposition linked to its pivotal use both as a means of
consolidating prior learning based on M and R methods and as a starting point for subsequent M and R activities. Moreover,
given the pervasiveness of dictations in French spelling instruction, Chervel (1977) formulated a hypothesis suggesting an

Page 141
inversion of the usual direction of didactical transposition (from scientific concepts to teaching content). He believed that in
order to improve the efficiency of instruction in spelling, it was necessary to create a ''school grammar" designed in terms of
orthographic utility. This school grammar has exerted enduring influence on the scientific grammar models developed by
linguists while remaining largely impervious itself to scientific criticisms.
What learning outcomes can be expected at the end of the chain of didactical transposition when employing each instructional
approach? The effectiveness of the M approach depends on the student's capacity to apply the prescribed memorization
techniques. It assures knowledge of language (mastery of spelling usage) without developing knowledge about language. In the
R approach, student learning depends primarily on reasoning and logical analysis. Acquisition of knowledge about language
(e.g., rules governing orthographic patterns) is considered essential for the development of spelling skills embedded in general
knowledge structures. The D approach aims at synthesis of several language competencies and to achieve this goal strongly
advocates acquisition of knowledge about language. Classroom lessons and exercises on grammar, conjugation, and vocabulary
all have a common goal: to foster mastery of spelling conventions and their systematic application in text transcription.
Contextualization. In both the M and R approaches, spelling is learned principally by decontextualized activities (study of
unrelated words or sentences). However, the transfer of orthographic knowledge to more global reading writing activities is
based on two different theoretical principles: in the M approach, transfer is facilitated by the study of words similar to those
used in the student's personal writing, whereas in the R approach, transfer is supposed to result from the elaboration of general
lexical and grammatical knowledge structures. The D approach offers a greater degree of contextualization from a linguistic
viewpoint. The fact that all aspects of spelling are dealt with in a text framework is expected to facilitate the transfer of
knowledge acquired in dictation to the student's personal text production. Transcription of a dictated text remains, nevertheless,
considerably more artificial than the text production tasks required in real-life settings.
Cognitive Load. In the M and R approaches, the decontextualized form of specific spelling activities is justified by an implicit
assumption of reduced cognitive load by isolating the elements to be learned from any surrounding noise produced by other
elements of a sentence or text. The D approach is often considered to provide the opposite situation: The student is confronted
with a complex spelling task in which he must think of all aspects of spelling and their interrelations. The practice of prepared
dictations (by appropriate

Page 142
lessons or exercises) attempts to facilitate student management of the otherwise heavy cognitive load imposed by text dictation.
Regulation of Learning. The M approach attempts, by the repetition of a simple study technique, to assure progressive
internalization of the different operations and, thereby, foster self-regulated management of spelling acquisition. In the R
approach, the regulation of learning comes from student interactions with the teacher and with other students and from the
organization of the instructional material. Expert teacher guidance is required to foster optimal use of the material. The D
approach also relies heavily on teacher guidance, especially by think-aloud modeling of the reasoning to be carried out when
transcribing dictation and when re-reading the dictated text. In some varieties of this approach, interaction among students
working in small groups or dyads allows confrontation of differing viewpoints that can contribute to the regulation of learning.
Integrating Spelling Acquisition in Text Production

Although specific instructional approaches of the M, R, and D types are still widespread in classrooms, there is increasing
interest in approaches that integrate spelling acquisition in situations of text production. These approaches seek to extend to
spelling the general principles that characterize process-oriented writing instruction. As proposed by Bean and Bouffler (1987),
spelling should be studied in the same "functional, social, contextual" (p. 7) perspective as the writing process itself. For
advocates of "whole language" instruction (McKenna, Robinson, & Miller, 1990), it goes without saying that orthographic skills
are best acquired by immersion in global reading and writing activities. Other researchers adopt a more cautious position while
still insisting on the importance of authentic communication situations in language instruction. For example, Jaffr (1986, 1992)
considers that functional text production activities are a necessary but not a sufficient condition for mastering spelling: They
provide the starting point and the overall framework for subsequent analytical activities whose goal is to extend and structure
the reflections provoked by writing.
Research on Integrated Instructional Approaches
Relatively few empirical studies have examined methods of integrating spelling instruction in text production and the learning
outcomes resulting from this instructional approach. We present findings of several recent investigations of English language
acquisition in classrooms using integrated instruc-

Page 143
tion as well as some conceptual contributions of recent work on French spelling instruction.
In a descriptive study of classroom writing practices, Davis, Clark, and Rhodes (1994) have attempted to identify the effects of
different instructional practices on student writing proficiency. Their investigation, carried out in 39 fourth- and fifth-grade
classrooms, is based on three types of data: observations of classroom activities in the area of literacy instruction, interviews
with teachers and school principals, and student writing samples. The data allow classification of classrooms in terms of the time
devoted to extended text production as compared to time spent on exercises dealing with language conventions. Four categories
of classes are established: (a) mostly exercises; (b) mixed, mostly exercises; (c) mixed, mostly extended text; and (d) mostly
extended text.
The results, obtained by hierarchical linear modelling, show several positive effects of instruction emphasizing extended text
production (categories c and d) as compared to instruction focused on exercises (a and b). After adjustment for pretest
differences at the student and classroom levels, statistically significant effects are found on two measures of student writing
proficiency: a measure of the quality of writing content (ideas and text organization) and a measure of writing mechanics
(syntax, punctuation, spelling). On the other hand, no significant differences between classroom categories are found when
mastery of writing conventions is assessed by a standardized multiple-choice battery (Iowa Test of Basic Skills). The effects are
relatively modest but, nevertheless, indicate that emphasis on extended text production is likely to foster students' capacity to
write texts correctly, without diminishing their performance on standardized tests. The stability of the effects for children of
different ethnic origins tends to confirm the general advantage of instruction integrating spelling and other conventions in text
production.
It should be noted, however, that extended text is defined in this investigation by a minimal criterion (i.e., a text consisting of
several lines written with some student autonomy and initiative). It is difficult to know whether the observed effects would be
accentuated (or, on the contrary, reduced) in the case of longer and more complex forms of text production. Furthermore, as
stated by Davis et al. (1994), classroom emphasis on extended text is not a homogeneous category but "a marker for a complex
web of interactional and instructional variables" (p. 565) that the research methods they applied were not able to untangle.
Needels and Knapp (1994) went further toward the proposal of principles that could guide integrated spelling instruction. They
discuss the possible advantages and limitations of a skill-based perspective involving drill and practice on each component skill
of written expression; they also point out the lack of empirical validation of the claims of the whole language perspective.

Page 144
Their own proposals belong to a third perspective: a sociocognitive framework for writing instruction inspired by the work of
Vygotsky (1978). They define six dimensions of classroom instruction based on this approach: (a) integrating component skills
within the context of the writing task, (b) using meaningful and authentic writing tasks, (c) linking writing to the child's
background and experience, (d) encouraging peer interaction, (e) approaching writing as a problem-solving process, and (f)
providing ample opportunities to write extended text.
Writing instruction in 42 fourth- and sixth-grade classes is analyzed on the basis of data from teacher logs, classroom
observations, and interviews with teachers. Each teacher's practices are rated according to their degree of coherence with each of
the six pedagogical variables associated with the sociocognitive approach to writing instruction. Methods of hierarchical
regression analysis are used to test the effects of these variables on two measures of student writing: writing competence and
writing mechanics. The results show that the six pedagogical variables account for more than 40% of the variance in student
writing competence but do not account for the variance in writing mechanics. In other words, when instruction is based on the
sociocognitive model, student texts have better compositional qualities, without any decline in conventional correctness. Of the
six pedagogical variables entering into the regression analysis, the strongest predictor is the relation of writing to student
background and experience. Although the results do not show that an integrated instructional approach leads to improved
spelling acquisition, they can at least reassure teachers who fear that decreased emphasis on traditional exercises will invariably
cause a decline in spelling skills.
Qualitative studies conducted by teams of French researchers and teachers (see issue 75 of the journal Repres, 1988) suggest
several pedagogical strategies that might be used in integrated spelling instruction. As proposed by Jaffr (1986) and Ducancel
(1988), two types of follow-up activities can be carried out after text production: (a) deferred activities in which explicit,
systematic treatment is given to spelling problems encountered during writing, and (b) detached activities aimed at knowledge
conceptualization through analysis and manipulation of well-chosen examples. Although these tasks are similar to those
typically proposed in the R approach, their utilization remains linked to the preceding text production situation or prepares the
way for a new writing situation allowing rapid reinvestment of acquired knowledge. In order to foster orthographic problem
solving during drafting and during revision, analogical reference tools are developed on the basis of common difficulties
encountered by students (Lartigue & Djebbour, 1988). In addition, attempts are made to define the role of student-teacher
interactions and of peer interactions as a means of regulating learning (Bled, 1988; Sandon, 1988).

Page 145
In an ongoing research project in Geneva,3 we are studying the implementation and the effects of an approach integrating
spelling acquisition in text production in second-grade classrooms and sixth-grade classrooms. Our approach draws on previous
research concerning four related topics: relationships between reading and writing in the beginning elementary grades (Rieben,
1993), self-regulation processes in text production (Allal, Michel, & Saada-Robert, 1995), analyses of student spelling errors
and representations (Btrix Khler, 1995), and strategies of spelling instruction (Berset Fougerand, 1993). Our approach is
similar in its general conception to the sociocognitive perspective defined by Needels and Knapp (1994).
The field work is based on a quasi-experimental design allowing comparison of two groups of matched classrooms that apply
contrasting instructional approaches during an entire school year: spelling acquisition integrated in text production versus
specific instruction emphasizing exercises and dictations. The effects of instruction are analyzed by pretest-posttest measures on
three tasks (exercises, text dictation, text completion). In addition, the reinvestment of orthographic knowledge and skill is
studied in a more complex narrative text production situation taking place at the end of the school year.
Data analysis is still in progress, but the principles underlying the integrated spelling approach can be briefly presented. Eight
text production situations have been prepared for each grade level. In each situation, two spelling objectives are targeted. The
choice of the objectives is linked to properties of text structure that provide a functional context for learning: for example, in
sixth grade, determiner-noun-adjective agreement is worked on in the context of descriptive text production, whereas verb
conjugation in the imperfect and past definite tenses is dealt with in narrative text production. Each text is written in an authentic
communication perspective for an audience inside or outside the classroom. Mastery of the targeted spelling objectives is
fostered by the regulations resulting from student interactions during writing and by deferred or detached follow-up activities.
Figure 8.2 presents the three phases of each instructional situation and the principles associated with each phase.
Comparison of Specific and Integrated Instructional Approaches
Approaches integrating spelling acquisition in text production do not yet constitute a well-recognized instructional option
validated by long-term empirical research in the classroom. We attempt, nevertheless, to describe the
3 This study is partly funded by a grant from the Swiss National Scientific Research Fund. The project team is headed by
Linda Allal, Dominique Btrix Khler, Madelon Saada-Robert, Laurence Rieben, and Edith Wegmuller.

Page 146
Phases Principles
1. Before text production
Define the communication setting: Create a functional communication
- aim and audience, situation by defining the rhetorical
- type of text (e.g., description, constraints and the characteristics of
narration, dialogue, cooking recipe). the
intended text.
Plan the content to be dealt with: By preliminary content planning,
collective discussion enumerating decrease the cognitive load required
ideas for
and lexical items likely to be used in the generation of ideas during writing,
writing the text. and thereby facilitate the students'
focus
on sentence transcription, including
spelling.
Present the targeted spelling Establish reference criteria for the
objectives regulations intervening during and
and specify the corresponding after
reference tools. text production.
2. During text production
Implement different means of Foster the development of spelling
regulating writing, based on the skills
student's interactions: that are fully integrated in the
- with the teacher, processes
- with other students, of drafting and text revision.
- with instructional tools. Enhance (across 8 successive
production
situations) the progressive
internalization
of self-regulation strategies.
3. After text production
Carry out various text-based Use deferred and detached follow-up
activities: activities to consolidate students'
analysis, classification and reflection spelling skills; differentiate the tasks
based on excerpts from student texts according to student needs.
and on other supplementary material
(specific exercises if needed).
Fig. 8.2.
An instructional approach integrating spelling acquisition in text production.
overall orientation of these integrated approaches (hereafter abbreviated as I approaches) as compared to that of the specific M,
R, and D approaches (hereafter grouped as S approaches) with respect to the four features of our conceptual framework.
Didactical Transposition. S approaches imply the transposition of specific components of knowledge concerning the
conventions of written language. I approaches entail a more complex process of transposition involving not only orthographic
knowledge but also the multiple levels of semantic, syntactic, and discursive knowledge required for text production. Analysis
of all aspects of this process is beyond the scope of this chapter, but two brief remarks can be made. First, implementation of I
approaches requires teacher education that articulates the different areas of literacy instruction and links them to general
strategies for regulating learning-teaching pro-

Page 147
cesses in the classroom. Second, given the complexity of the contextualized knowledge the student is supposed to acquire, it is
necessary to plan a well-coordinated curriculum sequence covering first grade through sixth grade in order to assure that
functional spelling skills do in fact become an integral part of writing competence.
Contextualization. I approaches are designed to promote contextualized learning as advocated in contemporary research on
situated cognition. The goal is to help students construct spelling knowledge in classroom settings that are close to the out-of-
school conditions in which knowledge is to be activated and used. Rather than the transfer of knowledge from a learning
situation to a context of future use, I approaches aim at learning in context from the outset. In this respect, they differ clearly
from S approaches that promote skill acquisition outside of any meaningful communication setting.
Cognitive Load. By proposing structured tasks involving a limited number of elements, S approaches (at least of the M and R
types) attempt to facilitate the learner's management of cognitive load. Because I approaches are highly contextualized, the
complexity of task elements and relationships tends to accentuate the cognitive load of the operations to be carried out.
Advocates of I approaches believe, however, that learning to manage the cognitive load of complex tasks is a goal that should be
explicitly dealt with in the classroom. Instructional design can attempt to restrain certain factors that contribute to cognitive load
without transforming other essential parameters of a complex task. In our ongoing study, for example, we maintain the
functional communicative purpose of the writing situation, but we have the students plan text content before beginning to write
so that they can subsequently concentrate on sentence generation and spelling. Moreover, by learning to make use of various
tools (reference books, techniques of proofreading), students can progressively acquire techniques for managing the multiple
task elements more efficiently.
Regulation of Learning. Depending on how S approaches are implemented, they can include various types of regulations
resulting from the student's interaction with the teacher, with peers, or with instructional material. I approaches also make use of
these means of regulation but attribute particular importance to the development of self-regulation through the internalization of
strategies and tools initially embedded in a network of social interaction. Instruction is designed to develop not only cognitive
skills in writing and in spelling but also students' metacognitive competence in managing relationships among task components.

Page 148
Toward the Coordination of Specific and Integrated Approaches?
In conclusion, we wish to raise the question of possible ways of coordinating the two types of instructional approaches dealt
with in this chapter. Although the theoretical principles underlying S and I approaches differ, in classroom practice it may well
be useful to seek ways of articulating aspects of each. The effectiveness of S approaches for the acquisition of word knowledge
has been demonstrated in a sizeable number of studies. Although available research provides relatively limited validation of I
approaches, the potential usefulness of these approaches seems obvious when dealing with grammatical aspects of spelling,
particularly inflectional morphology in French, and when attempting to assure that contextualized spelling skills become an
integral part of student writing competence. Coordination of specific and integrated approaches could be sought in two
directions. When applying an integrated approach, it is possible to insert modules of specific instruction at regular intervals and
thereby reinforce the acquisition of lexical knowledge. Conversely, after having applied a sequence of S modules, it is possible
to introduce a situation of integrated instruction designed to consolidate specific skills in a meaningful writing context. Further
work is needed in both of these directions to judge their relevance for classroom practice.
References
Allal, L., Michel, Y., & Saada-Robert, M. (1995). Autorgulation en production textuelle: Observation de quatre lves de 12
ans [Self-regulation in text production: Observations of four twelve-year-old students]. Cahier d'Acquisition et de Pathologie du
Langage, 13, 17-35.
Allal, L., & Saada-Robert, M. (1992). La mtacognition: Cadre conceptuel pour l'tude des rgulations en situation scolaire
[Conceptual framework for studying processes of regulation in school]. Archives de Psychologie, 60, 265-296.
Allen, D., & Ager, J. (1965). A factor-analytic study of the ability to spell. Educational and Psychological Measurement, 25,
153-161.
Bean, W., & Bouffler, C. (1987). Spell by writing. Portsmouth, NH: Heinemann.
Berset Fougerand, B. (1993). Ecrire haute voix: Intgration de l'orthographe dans la production crite [Writing out loud:
Integration of spelling in written production]. In L. Allal, D. Bain, & P. Perrenoud (Eds.), Evaluation formative et didactique du
franais (pp. 171-196). Neuchtel, Switzerland: Delachaux & Niestle.
Btrix Khler, D. (1995). Orthographe en questions [Spelling in questions]. Lausanne, Switzerland: Centre vaudois de
recherches pdagogiques.
Bled, B. (1988). Adapter les stratgies d'enseignement aux stratgies d'apprentissage des lves [Adapting teaching strategies to
students' learning strategies]. Repres, No. 75, 55-65.
Bronckart, J.-P., & Schneuwly, B. (1991). La didactique du franais langue maternelle: L'mergence d'une utopie indispensable
[Didactics of French as mother tongue: Emergence of an indispensable utopia]. Education et Recherche, 13(1), 8-25.
Brophy, J. E., & Good, T. L. (1986). Teacher behavior and student achievement. In M. C. Wittrock (Ed.), Handbook of research
on teaching (pp. 328-375). New York: Macmillan.

Page 149
Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. Educational Researcher, 18(1),
32-42.
Catach, N. (1978). L'orthographe [Spelling]. Paris: Presses Universitaires de France.
Chervel, A. (1977). . . .Et il fallut apprendre crire tous les petits franais: Histoire de la grammaire scolaire [And it was
necessary to teach all the little French school children to write: History of school grammar]. Paris: Payot.
Chervel, A., & Manesse, D. (1989). La dicte: Les Franais et l'orthographe 1873-1987 [The dictation: The French and spelling
1973-1987]. Paris: INRP/Calmann-Lvy.
Chevallard, Y. (1985). La transposition didactique: Du savoir savant au savoir enseign [Didactical transposition: From
scientific knowledge to taught knowledge]. Grenoble, France: La Pense Sauvage.
Chiss, J.-L., & David, J. (1992). La rgle orthographique: Reprsentations, conceptions thoriques et stratgies d'apprentissage
[The spelling rule: Representations, theoretical conceptions and learning strategies]. In M. Fayol & J. P. Jaffr (Eds.),
L'orthographe: Perspectives linguistiques et psycholinguistiques (Special issue). Langue franaise, 95, 80-98.
Chiss, J.-L., Laurent, J.-P., Meyer, J.-C., Romian, H., & Schneuwly, B. (Eds.). (1987). Apprendre/Enseigner produire des
textes crits [Learning/teaching to produce written texts]. Bruxelles: De Boeck.
Collins, A., Brown, J. S., & Newman, S. (1989). Cognitive apprenticeship: Teaching the crafts of reading, writing, and
mathematics. In L.N. Resnick (Ed.), Knowing, learning, and instruction: Essays in honor of Robert Glaser (pp. 453-494).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Davis, A., Clark, M. A., & Rhodes, L. K. (1994). Extended text and the writing proficiency of students in urban elementary
schools. Journal of Educational Psychology, 86, 556-566.
Dewey, J. (1963). Experience and education. London: Collier-Macmillan. (Original work published 1938)
DiPardo, A., & Freedman, S. W. (1988). Peer response groups in the writing classroom: Theoretic foundations and new
directions. Review of Educational Research, 58, 119-149.
Downing, J., DeStefano, J., Rich, G., & Bell, A. (1984). Children's views of spelling. The Elementary School Journal, 85, 185-
198.
Ducancel, G. (1988). Enseigner l'orthographe: De la gestion des marques de surface la connaissance du fonctionnement du
systme graphique [Teaching spelling: From managing surface markers to knowledge of the functioning of the graphical
system]. Repres, 75, 1-5.
Englert, C. S., Raphael, T. E., & Anderson, L. M. (1992). Socially mediated instruction: Improving students' knowledge and
talk about writing. The Elementary School Journal, 92, 411-444.
Fayol, M., & Largy, P. (1992). Une approche cognitive fonctionnelle de l'orthographe grammaticale [A functional cognitive
approach to grammatical spelling]. In M. Fayol & J. P. Jaffr (Eds.), L'orthographe: Perspectives linguistiques et
psycholinguistiques (Special issue). Langue franaise, No. 95, 80-98.
Fitzgerald, J. (1953). The teaching of spelling. Elementary English, 30, 79-84.
Graham, S. (1983). Effective spelling instruction. The Elementary School Journal, 83, 560-567.
Graves, D. H. (1983). Writing: Teachers and children at work. Portsmouth, NH: Heinemann.
Guion, J. (1974). L'institution orthographe: A quoi sert l'orthographe? quoi sert son enseignement?
[The spelling institution: What is spelling's purpose? What is the purpose of spelling instruction?]. Paris: Le Centurion.
Hanna, P. R., Hanna, J. S., Hodges, R. E., & Rudorf, E. (1966). Phoneme-grapheme correspondences as cues to spelling
improvement. Washington, DC: U.S. Government Printing Office.
Henderson, E. H. (1985). Teaching spelling. Boston: Houghton Mifflin.
Hodges, R.E. (1982). Research update on the development of spelling. Language Arts, 59, 284-290.
Horn, E. (1919). Principles of methods in teaching spelling as derived from scientific investigation. In G. Whipple (Ed.),
Eighteenth yearbook, National Society for the Study of Education. Bloomington, IN: Public School Publishing.
Horn, E. (1954). Teaching spelling. Washington, DC: American Educational Research Association.

Page 150
Horn, E. (1960). Spelling. Encyclopedia of educational research. New York: Macmillan.
Jaffr, J. P. (1986). Construire des savoirs sur la langue: Le cas de l'orthographe [Constructing knowledge about language: The
case of spelling] (pp. 56-68). In Communiquer, a s'apprend. Paris: Institut National de Recherche Pdagogique (srie
Rencontres pdagogiques, cahier no. 11).
Jaffr, J. P. (1992). Didactiques de l'orthographe [Didactics of spelling]. Paris: Hachette.
Lartigue, R., & Djebbour, S. (1988). Vers des projets d'enseignement/apprentissage de l'orthographe [Toward projects of
teaching/learning to spell]. Repres, 75, 47-54.
McCutchen, D., Covill, A., Hoyne, S. H., & Mildes, K. (1994). Individual differences in writing: Implications of translating
fluency. Journal of Educational Psychology, 86, 256-266.
McKenna, M. C., Robinson, R. D., & Miller, J. W. (1990). Whole langage: A research agenda for the nineties. Educational
Researcher, 19, 3-6.
Needels, M. C., & Knapp, M. S. (1994). Teaching writing to children who are underserved. Journal of Educational Psychology,
86, 339-349.
Rieben, L. (1993). Production crite en situation de classe et acquisition de connaissances lexicales [Written production in the
classroom and acquisition of lexical knowledge]. In J.-P. Jaffr, L. Sprenger-Charolles, & M. Fayol (Eds.), Lecture-criture:
Acquisition. Paris: Nathan.
Roller, S. (1954). La conjugaison franaise: Essai de pdagogie exprimentale [French conjugation: Contribution of
experimental pedagogy]. Neuchtel, Switzerland: Delachaux & Niestl.
Sandon, J.-M. (1988). Etude des comportements des matres face aux demandes des lves [Study of teacher behavior with
respect to student requests]. Repres, 75, 19-28.
Scardamalia, M., & Bereiter, C. (1986). Research on written composition. In M. C. Wittrock (Ed.), Handbook of research on
teaching (pp. 778-803). New York: Macmillan.
Slavin, R. E., Stevens, R. J., & Madden, N. A. (1988). Accommodating student diversity in reading and writing instruction: A
cooperative learning approach. Remedial and Special Education, 9(1), 60-66.
Sulzby, E. (1980). Word concept development activities. In E. H. Henderson & J. W. Beers (Eds.), Developmental and
cognitive aspects of learning to spell. Newark, DE: International Reading Association.
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12, 257-285.
Sweller, J., & Chandler, P. (1994). Why is some material difficult to learn? Cognition and Instruction, 12, 185-233.
Templeton, S. (1989). Tacit and explicit knowledge of derivational morphology. Reading Psychology, 10, 233-253.
Templeton, S. (1991). Teaching and learning the English spelling system: Reconceptualizing method and purpose. The
Elementary School Journal, 92, 185-201.
Turco, G. (1988). Ecrire et rcrire, au cours lmentaire et moyen [Writing and rewriting in elementary grades]. Rennes,
France: Centre rgional de documentation pdagogique.
Venezky, R. L. (1980). From Webster to Rice to Roosevelt: The formative years for spelling instruction and spelling reform in
the U.S.A. In U. Frith (Ed.), Cognitive processes in spelling (pp. 9-32). London: Academic Press.
Vial, J. (1970). Pdagogie de l'orthographe franaise [Pedagogy of French spelling]. Paris: Presses Universitaires de France.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge, MA: Harvard
University Press.
Yee, A. (1969). Is the phonetic generalization hypothesis in spelling valid? Journal of Experimental Education, 37, 82-91.

Page 151
Chapter 9
Spelling and Grammar The Necsed Move
Terezinha Nunes
University of London
Peter Bryant
University of Oxford
Miriam Bindman
University of London
The Phonological Connection and its Limitations

It is widely accepted that it takes children a long time to learn how to read and spell. Most current hypotheses about children's
reading, such as those of Frith (1985) and of Marsh (Marsh & Desberg, 1983; Marsh, Friedman, Desberg, & Saterdahl, 1981;
Marsh, Friedman, Welch, & Desberg, 1980) asserted that children's knowledge of written language increases right through their
school years and that their approach to it changes quite radically as they become more knowledgeable. Yet, by far, the largest
part of the remarkable body of recent research on children's reading has concentrated on its beginning stages, and very little of it
has dealt with the considerable learning that children still have to do later on.
The reason for this concentration on beginning reading is the striking success of research on one aspect of learning to read and
write. The data that we now have on the early stages of reading are rich and varied in detail, but put together, they lead to one
general and inescapable conclusion, which is that children's phonological awareness and phonological abilities play a crucial
part in learning to read. There is ample evidence now that children have to be able to analyze the sounds that make up words
and

Page 152
syllables in order to begin learning to read, that their awareness of these sounds is a contributory factor in learning to read and
write, that extra teaching in breaking words up into their constituent sounds and connecting these sounds to alphabetic letters
and to sequences of alphabetic letters improves their reading, and that children who are slow to learn to read are often also
particulary insensitive to the sounds in words. Much of the work on the importance of children's phonological awareness, as it is
usually called, is described in other chapters in this book, and it has been amply reviewed elsewhere (Goswami and Bryant,
1990; Perfetti, 1985). We mention it now only to show why the early stages of reading have been given so much prominence in
recent research.
The phonological connection is at its most important when children are first introduced to the alphabet. The reason why children
must analyze the sounds in words is that alphabetic letters represent those sounds. Unless children can break up words and
syllables into their constituent sounds, they will not understand how the alphabet works because it works by representing those
constituent sounds.
That beginning readers learn about this representation and use it in ingenious and often quite original ways is dramatically
demonstrated by the interesting work on invented spelling. This work shows that young children (usually in the 4- to 6-year age
range) do begin to apply letter-sound correspondences when they write words, and that they get progressively better at doing so
as they grow older and more experienced.
Early spellings are characterized by a strong tendency to literal phonetic transcriptions. Examples are easy enough to find. Here
are some from Charles Read's (1986) excellent book on the subject: ''kwyit'' for "quiet," "wrx" for "works," "sowemig" for
"swimming," "wodr" for "water," "halpt" for "helped," "watid" for "waited," "wotid" for "wanted." These spellings are
ingenious, but they are, of course, wrong in the conventional sense. However, there is more than one reason for this departure
from conventionality, and it is extremely important to see what these reasons are.
In some of these examples, the children concerned did not manage to get the phonetic transcriptions completely right. The
missing vowel in "wrx," the wrong vowel in "halpt," and the missing consonant in "wotid" are all typical examples of phonetic
mistakes that children make when applying the alphabetic, letter-sound correspondences that they have so recently learned. We
know that children often make these kinds of phonetic mistakes in the early stages and that this sort of error declines during the
early period. As they read and write more, their use of letter-sound correspondences in the literal sense gets better and better.
There is, however, another kind of error in these invented spellings that no amount of improvement in phonetic transcription will
ever remove. In written English and in many other orthographies, there are conventional

Page 153
spellings for the morphemes which tell us about the grammatical status of the words, and these conventional spellings often do
not accord with letter-sound correspondences. The ending of plural nouns is usually spelled with an "s" despite the fact that this
"s" often represents a /z/ sound. The ending of regular past tense verbs is spelled as "ed," and yet this spelling sequence
sometimes represents /d/, sometimes /t/, and sometimes /id/. Figure 9.1 presents an entirely representative piece of writing by a
7-year-old child that demonstrates quite clearly that she doesn't have any idea about how to write the final morpheme in regular
past tense verbs. Plainly she has not yet learned about the "ed" sequence and nor, as we shall show, have most English children
younger than 8 years of age.
It is easy to see why they have not. These regularities are tied to grammar, not to phonology. Nowadays, we do not pronounce
the ending of any words as /ed/, and our reasons for spelling these endings as "ed" are entirely grammatical. We end the written
word "helped" with an "ed" precisely because it is a regular past tense verb, and we do not end the written word "abrupt" in the
same way precisely because it is not a past tense verb. Surely children must learn about this grammatical connection in order to
be certain that the ending of ''helped'' is spelled in one way and of "abrupt" in another.
These simple, almost obvious, observations clearly lead to a new hypothesis, which is that children first have to conquer the
phonological connection and then the grammatical connection in written language. First they have to learn about the relation
between sounds and alphabetical letters and sequences of letters: Next they must conquer the relation between each word's
grammatical status and the way that it is spelled.
What then is already known about the grammatical connection?
The Grammatical Connection and The Idea about Context

Although the phonological connection has been at the center of recent research on children's reading, the idea that grammar
might be important has also received a fair amount of attention. However, this interest in the possibility of a grammatical
connection was almost always prompted by concerns that are quite different from those that we set out in the previous section.
As far as we can discern, the driving force behind work on the possible importance of children's grammatical knowledge in
learning to read is the notion of the psycholinguistic guessing game. The phrase is Kenneth Goodman's (1982) and it
encapsulates the idea that children use the context of the passage that they are reading in order to guess the meaning of any
difficult words in it. Contextual cues can take several different forms semantic, syntactic, or pictorial, for example and the
possibility that children's

Page 154
Fig. 9.1.
A story by a 7-year-old child that illustrates the depiction of past verb endings,
which is characteristic for that age.

Page 155
sensitivity to syntax might be important has led several researchers to devise tests of syntactic awareness and then to relate
children's performance in these tasks to the progress that they had made in learning to read.
This line of research has scored some notable successes. It has consistently been found that children who score poorly on such
tasks are likely also to make slow progress in learning to read (Rego & Bryant, 1993; Tunmer & Bowey, 1984; Tunmer,
Nesdale, & Wright, 1987), and it is now clear that there really is an interesting connection between children's syntactic
awareness scores and their ability to use the context to help them decipher the meaning of difficult words (Rego & Bryant,
1993).
We should like to make two points about this line of research. One is that there is a great deal of uncertainty about what
syntactic awareness tasks actually measure. For example, one of the tasks used in this research is a sentence anagram task in
which children have to make meaningful sentences out of words spoken to them in a jumbled order, and another is a cloze task
in which children hear a sentence with a missing word and are asked to fill the gap. Both these tasks, as Gombert (1992) and,
subsequently, Bowey (1994) have argued, make semantic as well as syntactic demands, and are not therefore pure tests of
syntactic awareness.
The second point is that there is a considerable difference between the grammatical knowledge that is at stake both in these tasks
and in Goodman's hypothesis and the grammatical knowledge that we ourselves wrote about in the previous section. Sentence
anagram tasks and cloze tasks deal with the structure of the sentence; we, in contrast, are concerned with the spelling of
morphemes. Sentence anagram and cloze tasks measure, for example, children's knowledge that a definite article is usually
followed by a noun or by an adjective and then a noun and never by a verb. In contrast we are speculating, for example, about
the difficulties that children might have in understanding why the last sound in kissed is spelled "ed" and not "t". There is no
conflict between the two approaches, and the hypotheses guiding them might both be right, but they are quite different from
each other. It is time to say what our own hypothesis is.
Syntactic Awareness and the Spelling of Morphemes Our Hypothesis and How to Test it
The hypothesis that we adopted at the start of our research was that there is a definite stage at which children come to grips with
the spelling of morphemes and that this new development is based on their grammatical or, more specifically, on their
morphosyntactic awareness. Children have to understand that some words end in "ed" because they are past tense

Page 156
verbs, that some words begin with "wh" because they are interrogatives, that some words end in "'s" because they are singular
possessives, and others end in ''s''' because they are plural possessives. Just as the earlier learning about letter-sound
relationships depends to a large extent on the children's awareness of phonological distinctions, so this later learning depends on
their sensitivity to grammatical distinctions, a sensitivity that we call morphosyntactic awareness.
Our hypothesis also raises the question of the origins of this form of linguistic awareness that we deem to be so important. It
could be the result of children's informal linguistic experience outside the classroom, or it could be the direct result of being
taught to read and to spell. In this chapter, we present no evidence on this particular question, though we think it an important
one. However, we think that the first step is to establish whether there is a connection between this kind of lingusitic awareness
and children's success in learning the conventional spelling of morphemes, and that is the step that we try to take here.
It is easy to see that any test of this hypothesis has two predictions to test: The first is that there is a developmental shift from
learning about letter-sound correspondences to learning how to represent grammatical distinctions in writing, and the second is
that this later form of learning is based on children's grammatical awareness.
The best test of any hypothesis about a developmental sequence is a longitudinal study because it shows not only that older
children behave differently from younger children but also that such differences are to be found in the development of
individual children: Longitudinal research demonstrates that individual children progress from one way of behaving to another.
Another advantage of longitudinal work is that it can provide a plausible test of the idea that learning how to spell morphemes is
heavily influenced by children's morphosyntactic awareness. If this is true, measures of this sensitivity taken in one session
should be strongly related to the subsequent progress that children make in spelling morphemes.
We devised a 3-year longitudinal study to look at children's spelling of morphemes and also at their morphosyntactic awareness
and to track the relation between the two. The children 363 at the outset of the project were 6, 7, or 8 years old the first time
that we saw them. In this chapter, we describe some of our results in the first year of the project, and so the ages that we are
dealing with are 6, 7, 8, and 9 years.
In the spelling tasks, we concentrated on past tense verbs. Several times during the first 2 years, we asked the children to write
for us the words that are presented in Table 9.1. Our list consisted of two kinds of past tense verbs: regular ones with the "ed"
spelling (e.g., kissed), and irregular ones whose ending is spelled phonetically (e.g., felt). Each of these types was divided into
two those ending in a "t" sound when spoken and those ending

Page 157
TABLE 9.1
The Words in Our Spelling List
Irregular Past Tense Verbs
/d/ sound ending /e/ sound ending
Called Dressed
Covered Kissed
Filled Laughed
Killed Learned
Opened Stopped
Regular Past Tense Verbs
Found Felt
Heard Left
Held Lost
Sold Sent
Told Slept
Nonverbs
Bird Belt
Cold Except
Field Next
Gold Paint
Ground Soft
Interrogatives
How
What
When
Where
Which
Who
Why
in a "d" sound. In addition, we included words that ended with a sound similar to the past tense verbs in our list, but they were
not past tense verbs. This made six categories, and our idea was that there were markedly different patterns in the way that
children spelled these different kinds of words at different ages. At first, they spelled them phonetically, later they introduced the
conventional "ed" spelling, particularly with past tense verbs, and finally they learned that there are some exceptions to the rule
that past verbs end with an "ed'' spelling.
In addition, we included some interrogatives, most of which began with a /w/ sound (the exceptions were "who" and "how") and
in their written form started with the "wh" spelling (the exception is "how") because they

Page 158
are another example of the spelling of a sound depending on the grammatical status of the word that the sound is in. Of course,
it is always possible that children could learn the right spellings for interrogatives by rote because there are so few of these
words. However, if we were to find a particularly strong connection between children's success with these words and with past
tense verbs, we could conclude that the two sets of words are learned in the same way and that the connections are due to both
sets of spellings having a grammatical basis.
Our main syntactic measures were new. Because our hypothesis was that children's spelling of these morphemes depends on
their awareness of differences in parts of speech (past tense verbs versus nonverbs and interrogatives versus noninterrogatives),
we needed tasks that measured this kind of awareness, which we have called morphosyntactic awareness, quite specifically. We
devised two new tasks of our own, and for a third we used a pseudoword technique originally devised by Berko (1958).
The three tasks are presented in Table 9.2. Both our tasks took the form of a series of analogies. In the sentence analogy task,
we spoke a sentence and then spoke exactly the same sentence except that the tense of the verb was changed: The change was
either from present to past tense or from past to present. Then we spoke another sentence in which the tense of the verb was the
same as the first of the original two sentences and asked the child to carry out exactly the same transformation on this new
sentence. We wanted to see if the child could make the same change in tenses as we had. In some of the trials, we used regular
past verbs and in others, irregular ones.
TABLE 9.2
Sample Questions From the Three Morphosyntactic Tasks
The Sentence Analogy Task
1.Tom helps Mary. Tom helped Mary.
Tom sees Mary. _______
2.Jane threw the ball. Jane throws the ball.
Jane kicked the ball. _______
The Word Analogy Task
1.Anger Angry 2. Teacher Taught
Strength _______ Writer _______
2.Walk Walked 4. Cried Cry
Shake _______ Drew _______
The Pseudoword Task
1.This is a person who knows how to snig. He is snigging onto his chair. He did the
same
thing yesterday. What did he do yesterday?
2.Ever since he learned how to do it, this man has been seeping the iron bar into a
knot.
Today he will do the same thing. What will he do today?

Page 159
The word analogy task took a similar form except that we used single words (where we had used whole sentences in the
sentence task), and the range of grammatical transformations was greater. Some involved past verbs, and others involved
different aspects of grammar.
Our third specific test of children's morphosyntactic awareness was a pseudoword task. This was an adaptation of Berko's (1958)
justly famous task, which seems to us to be a direct and simple way of measuring children's awareness of grammatical
distinctions. We gave children pictures and in describing these pictures, we introduced pseudowords. Then we asked children
questions that required them to transform these new nonverbs by adding a morpheme. Two trials dealt with singular to plural
transformations ("This is a picture of a ZUG; here is another picture and there are two of them in it. There are two ____?"). In 5
of the trials, the transformation involved present and past tenses ("This is a picture of a man who knows how to MAB. He is
MABBING along the street. Yesterday he did it too. Yesterday he ____?"). Finally, we tested the children's IQ, and we gave
them standardized tests of reading and spelling.
In this chapter, we confine ourselves to the data that we gathered on the first occasion when we saw the children and on a
following occasion 6 months later. We organize the data from these two sessions around the two questions that we posed in the
previous section: The first of these is the developmental question (Do children first spell phonetically and later learn the
conventional spelling of grammatical morphemes that are not clearly represented phonologically?), and the second is the causal
question (Is the later form of spelling based on children's grammatical awareness?).
The Developmental Question The Stages and the Data

To answer the developmental question, we looked first at the existence of certain kinds of misspellings of the endings of words.
If our idea that children originally spell phonetically and pay no attention to grammatical distinctions is right, there should be a
considerable number of phonetic spellings ("kist" for kissed). We also predicted that when children initially learn that past verb
endings are spelled as "ed," they apply this liberally to irregular past verbs as well as to regular ones; therefore, there should also
be several generalizations ("sleped'' for slept). We also argued that phonological spellings should be more typical of younger
children and generalizations more typical of older children.
We found these two kinds of mistake in profusion in our data (as Table 9.3 shows), and our developmental hypothesis was
bolstered by the fact that although the phonetic transcriptions decline between 7 and 8 years of age, the generalization of the
"ed" sequence to irregular past tense verbs actually increases as children get older.

Page 160
TABLE 9.3
Mean Number (out of 10) of Correct Spellings, and of
Different Kinds of Incorrect-Spellings by the Three Age Groups
Correct Spellings
6-Year-Olds 7-Year-Olds 8-Year-Olds
Regular verbs:
the correct "ed" ending 2.28 3.37 5.68
e.g., killed (3.42) (3.59) (3.34)
Irregular verbs:
the correct "d" or "t" ending 5.98 7.19 7.90
e.g., felt (3.15) (2.55) (2.33)
Nonverbs:
the correct "d" or "t" ending 6.04 7.49 8.25
e.g., bird (3.46) (2.58) (2.66)
Incorrect Spellings
6-Year-Olds 7-Year-Olds 8-Year-Olds
Regular verbs:
phonetic endings 3.29 3.75 2.53
e.g., "dresst" (2.81) (2.86) (2.26)
Irregular verbs:
generalization of "ed" ending 0.50 0.78 0.83
e.g., "sleped" (1.13) (1.29) (1.24)
Nonverbs:
overgeneralization of "ed" ending 0.33 0.57 0.44
e.g., "sofed (0.69) (1.01) (0.81)
Note. The figures in parentheses are standard deviations.
We also frequently found another, much more surprising kind of error, which we called overgeneralization (and which, in truth,
did not surprise us at the time, for we had found it before in some pilot work). This took the form of children sometimes
spelling the endings of nonverbs as "ed" ("sofed" for soft, "necsed" for next and "ground" for ground). This particular mistake
was less common than the other two kinds, but it was frequent enough and widespread enough to merit a great deal of attention.
Why do some children apply the conventional, but unphonetic, spelling for past tense verbs to words which are not verbs at all?
When we tried to answer this question, we used as a starting point the fact that these errors apparently increase and then
decrease with age as Table 9.3 shows. This suggests an intermediate stage somewhere between phonetic spelling and
grammatical spelling.
The idea we arrived at was that there is a time when children who have learned to use letter sound correspondences become
aware that these are sometimes violated and that one such infringement is that some words,

Page 161
which end in the /d/ or /t/ sound, end with an "ed" spelling in their written form. The children realize this, but they do not at first
understand the grammatical significance of the "ed" spelling, so they attach it to nonverbs as well as to past verbs. They are
genuinely at an intermediate stage too late for simple phonological spelling and too early for syntax.
When we included this last idea, our developmental hypothesis took the form of 5 developmental stages, which are summarized
in Table 9.4. The first is a prephonetic stage; we put this stage in our scheme because the first scan of our data showed that
many of the children, particularly the younger ones, failed to represent the endings of the words that they were given in any
consistent or comprehensible way.
The second stage is the phonetic stage, which we have mentioned a great deal already, in which the main characteristic is
spelling all the endings, past tense verbs included, phonetically. Children in this stage either never write the "ed" ending
appropriately or inappropriately, or write it very scarcely. Thus, our criterion for assigning children to this stage is that they
make a reasonable number of phonetic transcriptions at the ends of words but do not resort to the "ed" ending or do so very
infrequently.
The third stage is the intermediate stage in which children begin to use the "ed" ending, thus demonstrating that they are using
more than just phonetic correspondences but failing to see its grammatical basis. Thus, they write soft as "sofed" as well as slept
as "sleped." These children are not consistent in their use of the "ed" spelling. They use it sometimes but very often spell
phonetically instead. Our criterion for this stage is that children write ''ed''s at the end of words but sometimes do so to
nonverbs.
In the fourth stage, the children realize that there is a grammatical basis to the "ed" ending and put it only at the ending of past
tense verbs. However, they generalize this ending to irregular verbs as well. Children in this stage
TABLE 9.4
Characteristics of the Five Developmental Stages
Characteristics of the Children's Typical Spelling
Spelling at This Stage
Stage 1 Unsystematic spelling of word endings
Stage 2 kist, slept, soft
Frequent phonetic transcriptions of endings;
failure to produce conventional spellings of
morphemes
Stage 3 kissed, sleped, sofed
Some "ed" endings, but generalizations and
overgeneralizations: i.e., failure to confine this
sequence to past tense verbs
Stage 4 kissed, sleped, soft
"ed" spellings confined to past verbs, with
generalizations but no overgeneralizations
Stage 5 kissed, slept, soft
"ed" spellings confined to regular past tense
verbs; no generalizations or overgeneralizations

Page 162
are much more consistent in using the "ed" spelling with past tense verbs. They produce far fewer phonetic spellings than
children in the previous stage. This means that our criterion for assigning children to the fourth stage is that children write the
"ed" ending reasonably frequently and restrict this ending to past tense verbs but do so with irregular verbs as well as regular
verbs. We should like to add here that the transition from earlier stages to the fourth stage is the most important change because
this is when the children recognize for the first time that there is a connection between grammar and spelling.
Children in the fifth and final stage write the "ed" ending at the end of regular past verbs only and return to writing the end of
irregular verbs phonetically as they used to in the second stage. These children restrict the "ed" ending to regular past tense
verbs and do not generalize this spelling sequence to either irregular verbs or nonverbs.
In chapter 15, Rieben and Saada-Robert conclude that it is extrememly difficult to assign chidren to developmental stages in
reading. We did not find this to be so, as the results we report demonstrate quite clearly. There are various possible reasons for
this difference between Rieben and Saada-Robert's efforts and our own. One is that they are concerned with hypotheses about
reading strategies such as the logographic and alphabetic strategies, which may well occur at the same time and at a very early
stage in reading. We, on the other hand, are dealing with two forms of knowledge the connection between phonology and
spelling and the connection between grammar and spelling that probably come into play at very different times in the process of
learning to read. Indeed our hypothesis, clearly supported by our results, is that there is a gap of up to 2 years between the time
when children apply letter sound correspondences to their spelling and the time when they grasp the grammatical basis for
various spelling sequences.
Another possible reason for the difference between our relatively optimistic conclusion and Rieben and Saada-Robert's relatively
pessimistic conclusion about stages in learning to read and spell may be their insistence that the stages should be mutually
exclusive, in the sense that a child using a strategy characteristic of one stage should not at the same time adopt another strategy
that characterizes another stage. To some extent, our model demands this to be so and to some extent it does not. It is an
essential part of our hypothesis and of the objective criteria that we adopted that children in Stage 5 should make no
generalizations at all about past tense verbs or nonverbs (the characteristic patterns of Stages 3 and 4) and that children in Stage
4 should make absolutely no generalizations for nonverbs (the characteristic strategy of Stage 3). To that extent, the strategies
that children use in our stages are mutually exclusive. However, our model is more relaxed about inappropriate phonetic
transcriptions, the characteristic

Page 163
pattern of Stage 2. We did not expect children to stop making these mistakes at the time when they began to adopt the "ed"
ending, and they certainly did not. Our model accepts and even promotes the idea that children may continue to overuse letter
sound correspondences because letter sound correspondences continue to be and will always be an essential part of spelling.
Here, our model of children's development is an additive one. We do not think that children abandon letter sound
correspondences in spelling when they reach our Stage 3. That is an absurd idea. We simply argue that children eventually
begin to add to this strategy a new one that is based on the recognition of spelling sequences that are not literal phonetic
transcriptions and that are based on grammar. So, we were prepared to accept inappropriate phonetic transcriptions made by
children in Stages 3, 4, and 5, and we found them, although to a decreasing extent as the children ascended our developmental
ladder.
Our quite precise, developmental hypothesis produced two predictions that could easily be tested in our large-scale longitudinal
study. The first is that it should be possible to assign all, or at least nearly all, of the children quite unambiguously to one of
these 5 stages on the basis of the way that they spelled the words presented in Table 9.1.1 The second is that when children
change from one stage to another over time, the change should be upward on our developmental scale and not downward. When
we assign the children to stages in two different sessions, the first session and a following one 6 months later, the children
should be either at the same stage in both sessions or at a higher stage in the later session. Our developmental hypothesis
demands that individual changes over time should be in one direction but not in the other, and our longitudinal data allows us to
test this prediction in a most stringent manner by plotting each individual child's position in our sequence of stages in the first
session and in the next one 6 months later.
To test the first prediction, we objectively assigned children to stages on the basis of our criteria (given in Table 9.4). In fact, a
computer did it for
1 When we set the criteria for our stages, we decided that in order to be placed in Stages 2 to 5, children had to
demonstrate that they knew the alphabetic rules. Therefore, the requirement was that all the children in these stages had to
be able to spell half of the endings that were phonetic transcriptions (i.e., the endings of the nonverbs and irregular verbs).
If they could not manage that, children were categorized as prephonetic and assigned to Stage 1. Children assigned to
Stage 2, the phonetic stage, made at least five inappropriate phonetic transcriptions of regular past tense verbs (e.g.,
"killd" for killed) and also produced very few (less than three) "ed" spellings. For Stage 3, the stage when children begin
to produce the "ed" spelling without understanding its grammatical basis, children had to write at least three ''ed" endings.
However, at least one of those was a grammatically inappropriate generalization to a nonword. Our joint criterion for
Stages 4 and 5 was still that children should produce at least 3 ''ed" spellings, but these spellings had to be only for past
tense verbs. Children who generalized the spellings to irregular verbs were placed in Stage 4, whereas children who wrote
"ed" endings for only regular past tense verbs were assigned to Stage 5.

Page 164
TABLE 9.5
Mean Ages and Mean Number of Words Read in a
Standardized Single-Word Reading Test (Schonell) in the
First Session by the Children in the Different Stages
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5
at Time 1 at Time 1 at Time 1 at Time 1 at Time 1
N = 58 N = 78 N = 86 N = 52 N = 63
Mean age at Time 1 7y 0m 7y 2m 7y 9m 8y 0m 8y 1m
6y 8m 7y 2m 8y 6m 8y 11m 9y 7m
Mean reading age in
Schonell reading test
us because we wrote the criteria, given in footnote 1, into the program that selected the children for the different groups. In both
sessions, we found that very few children did not meet the criteria for any of the stages. In the first session, 92.8% (337 out of
363) of the children and in the next session, 95% (307 out of 323) of the children were easily assigned a position in our
sequence of stages.2
Our notion that these stages are developmental was supported by the fact that as Table 9.5 shows, there was a direct relationship
between the stage that the children were assigned, according to our scheme, and their ages as well as their scores on a
standardized reading test. The higher the stage, the older, on the whole, the children were and the more advanced at reading they
were.
Our second prediction was also confirmed. Table 9.6 shows that the majority of the children conformed to our developmental
hypothesis. In fact, 87.7% of the children either stayed at the same stage over the two sessions or were at a higher stage in the
second session than in the first. Of the children who were at different stages in the two sessions, many more were at a higher
stage after 6 months than they had been in the first session: 34.7% changed in the direction that was right, according to our
hypothesis, and only 12.3% changed in the wrong direction. The existence of the relatively small number of backsliders is
awkward for us, but it may be due to these children being held back by extraneous problems. At any rate, the marked trend for
children to move up rather than down our developmental ladder is strong support for our developmental hypothesis.
2 We also found that many of the 26 children who could not be classified in our stages tended to spell the endings of
regular verbs that we had designated as ending in /t/ (e.g., learned) as "d" (e.g., "learnd"). It could be argued that these
spellings should be treated as phonetic transcriptions too. When we analyzed all ''d" or "t'' spellings at the ends of words
as phonetic transcriptions, we found that 14 of the 26 excluded children were now classified as being in Stage 2 the
phonetic stage. However, for our main analysis we decided to stick to a strict criterion for phonetic spellings (i.e., only
"t"s for words ending in /t/ sounds and only "d"s for words ending in /d/ sounds).

Page 165
TABLE 9.6
Number of Children in Terms of Their Spelling Stages in Both Sessions
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Total
at Time 2 at Time 2 at Time 2 at Time 2 at Time 2
Stage 1 at Time 1 28 11 8 1 1 49
Stage 2 at Time 1 5 43 10 5 6 69
Stage 3 at Time 1 0 3 30 21 19 73
Stage 4 at Time 1 0 1 11 13 17 42
Stage 5 at Time 1 0 1 3 11 37 52
Total 33 59 62 51 80 285
We should like to mention one other result that provides extra support for our idea that children come to grips with the
grammatical basis of certain spelling sequences during this period. The scores for spelling the "wh" sequence at the beginning of
interrogatives were rather low, but there was a very strong relationship (r = 0.74) between children's success in writing the "wh"
sequence and writing the ''ed" sequence at the end of the right words. This strong relationship held even when differences in age
and in performance in a standardized spelling test were controlled in a fixed-order multiple regression; even after controls for
these variables, the children's "wh'' scores still accounted for 38.3% of the variance in their "ed" scores. This connection
certainly supports the idea that children learn both the "ed" ending and the "wh" opening sequence in the same way, and this
way must have something to do with grammar. It looks as though the same factor or factors might govern children's progress in
learning about both spelling sequences. What can these factors be?
The Causal question Morphosyntactic Awareness and the Development of Spelling

Our idea is that children's sensitivity to the distinction between different parts of speech is the basis for their eventual mastery of
the conventional spelling of grammatical morphemes. We call this sensitivity "morphosyntactic awareness." According to this
idea, there should be a strong relation between the children's performance in our syntactic tasks or at any rate in the three tasks
(see Table 9.2) that specifically tested morphosyntactic awareness and the children's positions in our developmental sequence.
We looked at this relation in two ways. One was to see whether children at different stages in our developmental sequence of
spelling also scored differently in the syntactic tasks. The other a more stringent prediction

Page 166
was to establish whether children's scores in the syntactic tests in the first session predicted whether they would improve their
position in the stage sequence by the time they were tested in the following session.
Table 9.7 gives the scores for the three syntactic tasks that specifically deal with the distinction between different parts of
speech. The table presents the mean scores for the children in the different stages. It shows that in all three cases, there was a
regular relation between the stage the children were at in spelling the verbs and nonverbs and their scores on the syntactic
awareness tests. The higher the stage, the higher was the performance in the syntactic task. It could be argued that this was just
a function of age because there was also a direct relationship between age and the stage assigned to the different children, but
with each of the three tasks, the relationship remained significant when age differences were controlled in an analysis of
covariance. Another result of these analyses of covariance was that in every case, there was a significant difference on posttests
between the children in Stage 3 and those in Stage 4. This is an important result as far as we are concerned because the
transition from Stage 3 to Stage 4 is a particularly important one. According to our scheme, it marks children's realization of the
connection between grammar and spelling.
Strong as these relationships within sessions are, relationships across time are more persuasive when it comes to testing causal
hypotheses. Our specific prediction concerned children who were at the same stage as each other in the first session but
progressed at different rates after. Our prediction was that the children's scores in the syntactic awareness tests in that first
session be related to the amount of progress. Those who moved to a higher stage should have started with higher scores in the
syntactic tasks than those who stayed where they were. In particular, we predicted that those who made the shift from Stage 1,
2, or 3 (pregrammatical) to Stage 4 or 5 (grammatical) should have scored higher on the syntax awareness tasks in Session 1
than those who remained in the pregrammatical stages.
TABLE 9.7
Mean Correct Scores in the Three Morpho Syntactic Awareness Tasks in
Terms of the Children's Spelling Stages in the First Session
Stage 1 at Stage 2 at Stage 3 at Stage 4 at Stage 5 at
Time 1 Time 1 Time 1 Time 1 Time 1
1.67 2.83 3.84 4.65 5.07
Sentence-analogy task (1.75) (2.13) (2.15) (2.22) (2.04)
(out of 8)
1.04 1.41 2.27 2.62 3.08
Word-analogy task (out (1.31) (1.47) (1.55) (1.64) (1.62)
of 8)
3.78 4.41 4.97 5.15 6.30
Pseudoword task (1.82) (1.81) (1.96) (2.49) (2.23)
(out of 10)

Page 167
TABLE 9.8
The Mean Syntactic Awareness Scores of Those Children in the
Pregrammatical Stages (Stages 1, 2, or 3) at Time 1 in Terms of Whether
They Advanced to a Grammatical Stage (Stage 4 or 5) at Time 2
Task Those Who Remained Those Who Advanced
in Pregrammatical to a Grammatical
Stage at Time 2 Stage at Time 2
N = 138 N = 53
2.45 4.02
Sentence-analogy task at Time (2.13) (2.03)
1 (out of 8)
1.48 2.14
Word-analogy task at Time 1 (1.46) (1.63)
(out of 8)
4.23 5.02
Pseudoword task at Time 1 (1.88) (1.82)
(out of 10)
Table 9.8 presents the results of this comparison. It shows that, in all three tests, the children who advanced from a
pregrammatical stage at Time 1 to a grammatical stage at Time 2 had done better in the syntactic awareness task at Time 1 than
those who remained at a pregrammatical level. These differences between those who advanced and those who did not were
significant in each case and remained so in analyses of covariance in which the children's initial stage in the first session was
entered as a covariate. This means that these tests do predict who, later on, will make progress to a higher level of spelling and
who will not.
Conclusions and Suggestions

The evidence that we have presented supports the idea of a developmental sequence in learning about written language that is
heavily influenced by children's sensitivity to spoken language and that probably stretches into late childhood. First, children
have to learn about letter sound correspondences, and this learning is heavily dependent on their phonological awareness, but as
they encounter a wider range of words and spelling patterns, they realize that there is more to spelling than the phonological
correspondences that they have tried to master so far. They begin now to adopt new spelling patterns such as the "ed" sequence
but with scant regard for the grammatical basis of this sequence. Later on, they learn how to apply the "ed" spelling to the
correct grammatical category and, later still, they conquer the exceptions. Their learning of the grammatical connection largely
depends on the strength of their grammatical awareness.
We have concentrated on one highly specific instance of this second kind of learning the understanding of the "ed" spelling
pattern but there are

Page 168
other examples in English that should be studied now. One is the "wh" sequence, which we have mentioned briefly and which
children conquer at roughly the same time as the "ed" pattern and in much the same way. Another is the use of the apostrophe
to signify possessives a striking instance of spelling indicating a grammatical relation. We have also gathered data in two pilot
studies and in our longitudinal project on children's learning the significance of apostrophes. These data still need further
analysis, but we already know that even 12-year-old children are often completely ''at sea" with the grammatical basis for
apostrophes. This supports our claim that learning about the grammatical basis for spelling continues for most of the school
years. Learning to read and to spell English is a long business, and we need to know about the whole of this business and not
just its early years if we are to ensure that every child is literate by the time that he or she leaves school.
We also assert that it takes children learning other orthographies a great deal of time to learn about the grammatical basis of
spelling patterns. In most European languages, as well as English, one can find the phenomenon of the same sound being spelled
in different ways depending on the grammatical or lexical status of the word that the sound is in. In Portuguese, for example,
there are several phonological sequences that have alternative spellings, only one of which will be correct for certain classes of
words. The phonological sequence /ozu/ may be spelled as "oso" or "ozo" and the pronunciation of these two spellings is
indistinguishable. However, the spelling sequence "oso'' is a derivative morpheme that can be attached to nouns to generate
adjectives; adjectives ending in the phonological sequence /ozu/ will only be spelled as "oso." Nunes (1992) observed that first-
grade children (age 7) have a preference for using "z," which is the letter they initially learn to represent this sound. However,
the letter S between two vowels is always pronunced as /z/ in Portuguese, and children discover soon enough that the sound /z/
may be represented by "s" or "z," without much awareness of either the phonological or the morphological contexts that allow
for the discovery of consistency. In spite of the complexity of the spelling rules, it is possible to observe some consistency in
children's progress in spelling of the phonological sequence /ozu/. Children start out using "z" to represent all the /z/ sounds,
including all instances of /ozu/. Then they use "s" also to represent the /z/ sound, but no particular pattern for this use can be
discerned at first. However, in fourth grade, the probability of children spelling the /z/ sound with "s" in pseudoadjectives
formed by adding the suffix to a real noun was greater than that of spelling with "z," although errors were still observed at this
age.
In French, too, grammar plays a crucial role in spelling. Third-person singular and third-person plural verbs have identical
pronunciations for the most part but are spelled very differently ("il aime" and "ils aiment," for

Page 169
example). The same is true of most singular and plural nouns. They usually sound the same, and yet their endings are spelled
quite differently ("maison" and "maisons"). With both nouns and with verbs there is a written, but unspoken, plural ending.
In Chapter 6, Totereau, Thevenin, and Fayol show that French children take several years to come to grips with these different
endings. At first, they simply ignore the unpronounced plural endings in their writing. They write the endings of plural verbs
and nouns as they sound without an "nt" or a "s" ending in just the same way as English children ignore the "ed'' ending and
write kissed as ''kist." Later, they begin to put in the "s" ending for plurals, but they add it to the ends of verbs as well as to the
ends of nouns. Later still they learn the "nt" ending, but instead of just adding it to verbs, they also sometimes put it at the end
of nouns as well. Finally, they learn to add "nt" to third-person plural verbs and "s" to plural nouns. This developmental
sequence is strikingly similar to the sequence that we have described with English children. In both languages, children do not
write the grammatical morpheme if it does not fit with the grapheme phoneme correspondences that govern their spelling at the
time, and in both languages, this initial phase is closely followed by a period in which children actually do include the
conventional spelling for the grammatical morpheme but often place it in grammatically inappropriate words ("sofed" instead of
soft in English; "maisonent" instead of "maisons" in French). Finally, in both languages, children do learn the grammatical basis
for the different spellings.
It is hard to resist the conclusion that this learning must be based on children's knowledge of grammar. The children are
certainly using some grammar; their readiness in Stage 4, in our sequence, to spell irregular past verbs with an "ed" ending is
sufficient evidence for that. They have never seen words similar to "sleped" before, and so their willingness to put "ed" endings
in past verbs, regular or irregular, and not in other words, shows that they have understood the grammatical significance of "ed."
Our data, however, give us more than a mere a priori case for the grammatical connection. The three syntactic tasks that we
devised are strongly related to our developmental sequence, particularly with the difference in this sequence between the stages
that we describe as pregrammatical and grammatical. The three tests also predict children's subsequent progress up this
developmental ladder.
To clinch the causal argument, we need intervention studies; we need to strengthen children's grammatical awareness and see
how this affects their spelling. Inevitably, such studies raise the educational question, too. What should we teach our children
about grammar? We have no ready answer to this question, and we think that the answer must wait for research on how to
increase children's awareness of grammatical distinctions. That should be the necsed move.

Page 170
References
Bowey, J. (1994). Grammatical awareness and learning to read: A critique. In M. H. Assink (Ed.), Literacy acquisition and
social context (pp. 122-149). Hemel Hempstead, England: Harvester Wheatsheaf.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, M. Coltheart, & J. Marshall (Eds.), Surface
dyslexia (pp. 301-330). London: Lawrence Erlbaum Associates.
Gombert, J. E. (1992). Metalinguistic development. Hemel Hempstead, England: Harvester Wheatsheaf.
Goodman, K. (1982). Miscue analysis: Theory and reality in reading. In F. K. Gollasch (Ed.), Language and literacy: The
selected writings of Kenneth S. Goodman (pp. 129-145). Boston: Routledge & Kegan Paul.
Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. London: Lawrence Erlbaum Associates.
Marsh, G., & Desberg, P. (1983). The development of strategies in the acquisition of symbolic skills. In D. R. Rogers & J. A.
Sloboda (Eds.), The acquisition of symbolic skills (pp. 149-154). New York: Plenum.
Marsh, G., Friedman, M. P., Desberg, P., & Saterdahl, K. (1981). Comparison of reading and spelling strategies in normal and
reading disabled children. In M. Friedman, J. P. Das, & N. O'Connor (Eds.), Intelligence and learning (pp. 363-367). New
York: Plenum.
Marsh, G., Friedman, M. P., Welch, V., & Desberg, P. (1980). The development of strategies in spelling. In U. Frith (Ed.),
Cognitive processes in spelling (pp. 212-237). London: Academic Press.
Nunes, T. (1992). Leitura e escrita: Processos e desenvolvimento [Processes and development in reading and spelling]. In E. S.
de Alencar (Ed.), Novas contribuicoes da psicologia aos processos de ensino e aprendizagem [New contributions from
psychology to teaching and learning] (pp. 13-50). Sao Paulo: Cortez Editora.
Perfetti, C. A. (1985). Reading ability. Oxford, England: Oxford University Press.
Read, C. (1986). Children's creative spelling. London: Routledge & Kegan Paul.
Rego, L. L. B., & Bryant, P. E. (1993). The connection between phonological, syntactic, and semantic skills and children's
reading and spelling. European Journal of Psychology of Education, 8, 235-246.
Tunmer, W. E., & Bowey, J. (1984). Metalinguistic awareness and reading acquisition. In W. E. Tunmer, C. Pratt, & M. L.
Herriman (Eds.), Metalinguistic awareness in children: Theory, research, and implications (pp. 121-144). Berlin: Springer-
Verlag.
Tunmer, W. E., Nesdale, A. R., & Wright, A. D. (1987). Syntactic awareness and reading acquisition. British Journal of
Developmental Psychology, 5, 25-34.

Page 171
PART III
THE RELATIONSHIP BETWEEN SPELLING AND READING


Page 173
Chapter 10
Why Spelling is More Difficult than Reading
Anna M. T. Bosman
University of Nijmegen
Guy C. Van Orden
Arizona State University
Thinking is always limited by the quality of our metaphors, and this is no exception.
Robert James Waller (1988, p. 116)
Before you read this chapter, please get a pen and write down the name of the Indian nationalist and spiritual leader who
developed the practice of nonviolent resistance that forced Great Britain to grant independence to India. He is known as
Mahatma ____.
Campbell and Coltheart (1984) found that undergraduates of two London colleges, despite numerous exposures through ads and
newspaper articles following the release of the famous movie "Ghandi,"1 were unable to spell the name of the man who showed
India the way to freedom. The chance that you spelled his name correctly is between 14% and 44%. An example from the
Dutch language (the first author's native tongue) is the notorious misspelling of the word sperzieboon (green bean). The majority
of people, including greengrocers, misspell the word by writing a C instead of a Z
1 Actually, the proper spelling is GANDHI. Campbell and Coltheart (1984) explained the high percentage of "Ghandi"-
misspellings in English-speaking college students in terms of a high summed position-specific bigram frequency of this
pattern as compared to that of the correct spelling pattern "Gandhi." However, a similar test with Dutch-speaking
University students showed that they had the same tendency of spelling Gandhi's name as Ghandi (21 out of 24
misspellings); Dutch summed position-specific bigram frequency, however, favors the Gahndi-pattern (Klnhammer,
1990).

Page 174
(spercieboon). This misspelling is phonetically accurate but is rather unconventional. Thus, to be in a QUANDERY about the
spelling of QUANDARY is not a problem restricted to children. Highly skilled writers often need a dictionary or the spell-
checker of a word processor to verify a word's spelling, but for skilled readers, doubts about how to read a word are largely
restricted to heterophonic homographs English examples are live, tear, and wind; Dutch examples are kantelen
(topple/crenelation), regent (rains/regent), and bedelen (endow/beg for).
This asymmetry between spelling and reading is evident at all levels of literacy. Skilled readers read more words correctly than
they can spell, and children's spelling and reading skills diverge soon after the onset of formal instruction (Mommers, 1987;
Seymour & Porpodas, 1980). It is relatively easy to find children whose reading performance meets the required level but whose
performance on a spelling test is below average, whereas the opposite pattern is much more rare (Frith, 1980). Spelling
problems of dyslexic readers also prove to be more persistent than reading problems (Frith, 1984, 1985; Nicolson & Fawcett,
1994; Thomson, 1984).
These observations raise many interesting questions, including the general question of how spelling and reading are related.
Moderate to high correlations between scores on spelling and reading tests suggest a fairly tight relationship (Malmquist, 1958,
cited in Frith, 1980). This leads some theorists to emphasize the similarities between the skills, and claim that the processes
(Ehri, 1980; Gough, Juel, & Griffith, 1992) or representations (Perfetti, 1992) underlying spelling and reading are, in fact, the
same. Others, however, stress the differences between spelling and reading processes, suggesting two separate mechanisms
(Bryant & Bradley, 1980; Frith & Frith, 1980). Frith (1979) claimed that reading occurs by "eye" and spelling by "ear" to make
the point that spelling is phonologically mediated but reading is not. Read (1981) also argued that reading and spelling are not
symmetrical because children who attempt to write words or stories often cannot read their own "invented" spellings (see also
Treiman, 1993).
The most cited empirical evidence suggesting separate mechanisms for spelling and reading was provided by Bryant and
Bradley (1980). Dyslexic and nondyslexic beginning readers read words they had previously spelled. Generally, both groups
read more words correctly than they were able to spell. In all groups, however, there were some words that children were able to
spell but unable to read (the percentage varied between 3% and 13%). Gough, Juel, and Griffith (1992) replicated the
experiment with a group of nondyslexic beginning readers. On four different occasions, their participants read each word twice
and spelled each word twice. As in the study of Bryant and Bradley, the children sometimes were able to spell words they were
unable to read (on average 10%). Moreover, they sometimes were able to read words on one occasion but not on the other
(10%) and, sometimes, they

Page 175
spelled words incorrectly on one occasion and not on the other (11%). Thus, the inconsistency observed between spelling and
reading also occurred within reading and within spelling, respectively. If the inconsistency between spelling and reading merely
reflects performance variability across occasions, then this inconsistency no longer motivates separate mechanisms.
In this chapter, we eventually present a framework for spelling and reading that does not require independent components to
explain the asymmetry between spelling and reading performance.
Spelling and Phonology
Since Read's publications on children's invented spellings (1971, 1981, 1986), it has generally been acknowledged that
beginning (and skilled) spellers rely heavily on phonology. All theories of spelling acquisition include a dominant role for
phonology (Frith, 1980, 1985; Henderson & Beers, 1980; Marsh, Friedman, Welch, & Desberg, 1980; Read, 1986; Snowling,
1994). Empirical evidence for phonology's role in spelling comes from two sources. First, phonological awareness and spelling
skill are highly correlated (Bruck & Treiman, 1990; Duighuisen & van Bon, 1992; Holligan & Johnston, 1991; Lundberg, Frost,
& Peterson, 1988; Marcel, 1980; Stuart & Masterson, 1992; uit de Haag, 1994; van Bon & Duighuisen, 1995). Second, spelling
errors are predominantly phonetically accurate.
One common definition of phonetically accurate spelling errors is that the error pattern can be pronounced identically to the
intended word using grapheme-to-phoneme correspondence rules (e.g., Bosman, 1994; Bruck, 1988; Holmes & Ng, 1993;
Treiman, 1993). For instance, the spelling pattern KLINTON is phonetically correct as an attempt to spell the name Clinton.
Sounding out KLINTON by applying grapheme-to-phoneme correspondence rules will sound just like an American President's
last name. A similar example in Dutch, sounding out the incorrect graphemic pattern VAN GOCH produces van Gogh a
phonetically correct representation of a famous Dutch painter's last name. The spelling pattern PLINTON, in an attempt to spell
Clinton, is considered phonetically incorrect because no grapheme-phoneme correspondence rule seems to be able to produce
the intended word. This definition of phonetic accuracy has been applied to the spellings of skilled adult spellers (Bosman, 1994;
Campbell & Coltheart, 1984; Gompel, Tromp, de Vries, & Bosman, 1990; Holmes & Ng, 1993; Sloboda, 1980), normal
beginning spellers (Bosman & de Groot, 1991; Bosman, de Groot, & van Leerdam, 1995; Bosman & van Leerdam, 1993;
Brown & Ellis, 1994; Manrique & Signorini, 1994; Porpodas, 1987; Waters, Bruck, & Seidenberg, 1985), children with spelling
(and reading) difficulties (Bruck, 1988; Bruck & Waters, 1988; Frith 1984; Nelson, 1980; Valtin, 1987), prelingual deaf children
and adults (Camp-

Page 176
bell, 1994; Dodd, 1980; Gibson, Shurcliff, & Yonas, 1970), language-disordered children (Cromer, 1980), an adult with an
acquired spelling disorder (Hatfield & Patterson, 1983), and a hyperlexic child whose spelling and reading skills were
extraordinary but who performed below average on intelligence and language tests (Siegel, 1994). In all these cases, the vast
majority of spelling errors were phonetically acceptable.
The occurrence of phonetically correct errors is traditionally explained by a phonological spelling route, that is, phonemes are
recoded into graphemes by means of phoneme-to-grapheme rules. However, there are inherent limits on a phoneme-grapheme
converter. A large number of phonemes map onto more than one grapheme (e.g., the phoneme [i:] has two possible spellings as
in Heap and as in Deep). Also spellers do make errors that do not qualify as phonologically correct by the previous criterion.
These facts seemed to necessitate a second lexical spelling mechanism. The lexical spelling route reads out whole-word
spellings. Non-phonetic errors occur when something goes wrong in the read-out process (Ehri, 1980; Holmes & Ng, 1993;
Kreiner & Gough, 1990). These two routes from lexicon to print in the dual-route model of spelling production (Barry, 1994)
are the mirror images of the two routes in the dual-route model for reading (Coltheart, 1978).
Next, we question whether spelling errors can be reliably classified in two distinct categories. In doing so, we question the
empirical justification for two spelling routes.
Spelling Errors Reassessed

The previous definition of phonetically acceptable errors is based on applying spelling-to-sound rules to a spelling pattern.
However, both Read (1971, 1981, 1986) and Treiman (1993) showed that an evaluation of spelling errors solely based on
spelling-to-sound rules seriously underestimates the phonetic complexity of children's invented spellings and the spellings of
beginning readers/spellers (see also Moats, 1993). We do not address this issue in full; excellent treatments are provided by
Treiman and Read. An illustration by means of examples is sufficient.
Take, for instance, the letter string DIK in an attempt to spell the word dike. According to spelling-to-sound rules, this is a
phonetically inaccurate spelling pattern. However, if we extend the definition and assess phonetic accuracy using the sound-to-
spelling relation in the other direction, this pattern becomes phonetically correct. The phoneme [ai] covaries with the grapheme
I*e, as in stripe, hive, and agile, but also with I as in pint, find, bind, kind, and mind. A speller who spelled DIK on the basis of
the second set of words has shown sophisticated knowledge of a sound-to-spelling relation. The same applies in Dutch. The
letter string AMSTERDAMERS in an attempt

Page 177
to spell the word Amsterdammers (people from Amsterdam), is a phonetically incorrect spelling pattern, given standard spelling-
to-sound relations, but sound-to-spelling rules render it a perfect representation of the word Amsterdammers. The majority of
Dutch first graders make this mistake initially.
Assessing spelling errors on the basis of both spelling-to-sound and sound-to-spelling appears to reveal more of the phonetic
structure in children's (and probably also in adults') spelling attempts than just on the basis of spelling-to-sound rules. Thus, a
more liberal assessment of phonetic accuracy appears reasonable. However, both the restricted and the liberal definition of
phonetic accuracy assume a categorical distinction between phonetically correct and phonetically inaccurate spelling errors. The
following examples show the difficulty (if not impossibility) of precisely specifying these categories. A child's misspellings in
the words KELL (for bell), KALL (for ball), KUG (for bug), and KIT (for bit) appear phonetically inaccurate in terms of
spelling-to-sound rules and sound-to-spelling rules, but are these spelling attempts void of phonetic knowledge? The most likely
interpretation of the error pattern of this child is that she misremembered the physical shape of the letter B. After all, the
(phonetic) structure in the incorrect spelling patterns is identical to the structure in the correct patterns. (These English examples
are made up to illustrate patterns observed by the first author in Dutch children's spelling.).
Another example illustrates why it is often necessary (and revealing) to know what a child wanted to spell to fully appreciate its
phonetic complexity. For example, the letter string CINPYEUTER is an attempt by Read's son, at the age of 6, to spell the word
computer (Read, 1981). To establish that a spelling error was not mediated by phonology, it is necessary to fully recover the
intention of the speller (see also Moats, 1993), whereas establishing that a spelling error was mediated by phonology appears
uncontroversial. This practical impossibility of sorting errors into two distinct categories bears on theoretical issues because this
distinction is the sole basis for proposing two spelling mechanisms.
The previous examples do not imply that meaningful assessment of spelling errors is impossible. A less theoretically driven
evaluation of spelling errors might be how well they represent the sounds of words according to the conventions of the
orthography in question. Spelling errors of less advanced spellers and of disabled spellers are usually less consistent and more
idiosyncratic than those of skilled spellers (Bosman, 1994; Bosman & de Groot, 1991; Bruck, 1988; Bruck & Waters, 1988;
Lennox & Siegel, 1993; Waters, Bruck, & Seidenberg, 1985). But, there is no basis for the assumption that phonology plays a
less important role in less advanced and disabled spellers (cf. Holligan & Johnston, 1991). In fact, it is highly likely that the
phonologic properties of words overwhelm the less able speller (compare the exaggerated phonologic effects in dyslexic readers
compared to normal

Page 178
readers; cf. Van Orden & Goldinger, 1996; Van Orden, Pennington, & Stone, 1990).
To summarize, we believe that spelling errors cannot be reliably classified into two categories, which renders an explanation in
terms of two independent spelling routes superfluous. In the next section, we show how a single process framework, developed
to explain word perception, also explains spelling performance. Word perception models usually do not incorporate spelling
(Coltheart, Curtis, Atkins, & Haller, 1993; McClelland & Rumelhart, 1981; Seidenberg & McClelland, 1989), and spelling
models do not include word perception (Brown & Loosemore, 1994; Houghton, Glasspool, & Shallice, 1994; Olson &
Caramazza, 1994). The following proposal is an attempt to incorporate both in a common framework.
A Dynamic Systems Framework for Spelling and Reading

Our proposal derives from a new framework for word perception (Stone & Van Orden, 1994; Van Orden & Goldinger, 1994;
Van Orden et al., 1990). This framework is rooted in dynamical systems theory, which supplies a new metaphor for cognitive
systems (e.g., Thelen & Smith, 1994). Space requires a severely abridged version of this account; interested readers are referred
to the original papers. Figure 10.1 depicts the macro dynamic of word perception described in Van Orden and Goldinger
(1994).2 Three families of fully interdependent subsymbols3 (nodes) are assumed for a recurrent network model. Dynamics
among nodes are sufficient to mimic performance attendant on word perception. In a model, the presentation of a printed word
activates letter nodes that, in turn, activate phoneme and semantic nodes. Following initial activation, recurrent feedback
dynamics begin among all these node families. The assumed dynamics are similar to those of the Interactive-Activation model
of McClelland and Rumelhart (1981; see Stone & Van Orden, 1994). Behaviorally meaningful structure emerges in positive
feedback loops among the three sets of nodes. The order in which feedback loops cohere is determined by the history of
bidirectional correlations between words' printed forms and their linguistic functions (cf. Varela, Thompson, & Rosch, 1991).
2 We use macro and micro as relative terms. What is macro for word perception could be micro for discourse processing.
Likewise what is micro for word perception could be macro for letter perception (cf. Ziegler & Jacobs, 1995).
3 The term subsymbol does not refer to a lower level symbol in the sense used in representational theory. Subsymbols are
simply pragmatic notations for the purposes of modelling or illustration. They are not to be taken as psychologically real or
as fundamental units of cognition (for an in-depth discussion of this issue, see Van Orden & Goldinger, 1994; Putnam,
1981).

Page 179
Fig. 10.1.
An illustration of the macro dynamic that describes reading and spelling
performance in a recurrent network. The boldness of the arrows indicates
the overall strength (self-consistency) of the relations between the
respective node families.
At the macro level of description, node families differ in overall strength and consistency of relations with other node families,
as illustrated by the relative boldness of arrows in Fig. 10.1. Overall, the relations between letters and phonemes in alphabetic
languages support the most powerful bidirectional correlations. The same letters and phonemes occur together in very many
words. Phoneme-semantic relations and letter-semantic relations are less strongly correlated. Phonemes and semantic features,
and letters and semantic features covary much less systematically. However, phoneme-semantic relations support stronger
correlations than letter-semantic relations. This is true because, essentially, we speak before and more often than we read.
Once in place, this asymmetry is self-perpetuating. Reading strengthens phoneme-semantic connections because phonology
functions in every instance of printed word perception. Thus, even the exceptional condition of a person who reads more than
she speaks supports phoneme-semantic connections that are at least as strong as letter-semantic connections. Also, in principle,
if a coherent positive feedback loop forms between phoneme nodes and semantic nodes, before the feedback loop between letter
nodes and semantic nodes, then printed or spoken discourse may proceed without settling the feedback loop between letter
nodes and semantic nodes. The absence of coherence in the latter feedback loop may preclude strengthening the connections
between letter nodes and semantic nodes (cf. Grossberg & Stone, 1986).
The relative strength of the relations between letter and phoneme nodes illustrates why phonology supplies such powerful
constraints on word perception. Stated differently, it explains why phonologic mediation is fundamental to reading (and
spelling). Strong bidirectional connections between nodes yield powerful feedback loops that cohere very early in a model's
dynamics. Abundant empirical evidence agrees with this claim (Berent &

Page 180
Perfetti, 1995; Bosman & de Groot, 1995, 1996; Carello, Turvey, & Lukatela, 1992; Perfetti, Zhang, & Berent, 1992; Van
Orden et al., 1990). In Fig. 10.2, we zoom in on the micro dynamics of fine-grain, letter-phoneme relations. In Fig. 10.2a, the
presentation of the word hi has activated the letter nodes H1 and I2, which activate the phoneme nodes [h1] and [a12], as well
as competing correlated nodes such as [i] (as in hit), which must be inhibited. Figure 10.2b shows how, in turn, phoneme nodes
feed back activation to letter nodes (illustrated for the phoneme nodes [h1] and [a12]). [a12] activates the correct letter nodes
H1 and I2 as well as competing letter nodes. For example, [a12] activates the letter node Y2 as in my or by. Dynamics from this
point on select a ''correct'' combination through inhibition of competing nodes.
Reliable performance emerges if the overall bidirectional configuration of connections between the letter nodes H1 and I2 and
the phoneme nodes [h1] and [a12] favors coactivation of these four nodes. This advantage grows over time as the "strong grow
stronger," and the "weak grow weaker" (cf. McClelland & Rumelhart, 1981). This is illustrated in Fig. 10.2c, which combines
the flow of activation from letter nodes to phoneme nodes and from phoneme nodes back to letter nodes, as assumed in a
recurrent network.
The key to how one might apply this model to spelling is that all connections are bidirectional, which means that activation
flows in both directions. Thus, activation arising in semantic or phoneme nodes may generate a coherent pattern of activity
across letter nodes. This characteristic of recurrent models has important implications for word perception as well. It predicts
that word perception should be affected by the relative consistency of phoneme-letter relations. In other words, not only does it
matter for visual word perception that a word's spelling may have more than one pronunciation (e.g., Gibbs & Van Orden,
1995), it also matters that a word's pronunciation may have more than one spelling.
Stone, Vanhoy, and Van Orden (1995) tested this hypothesis using a grain size of spelling-to-phonology that was larger than
letters and phonemes, namely, onsets and rhymes (or bodies). The onset of one-syllable words is the initial consonant cluster
(STR in Street), and the rhyme or body is the vowel and the final consonant(s) (EET in Street). Although letter phoneme
correlations are tracked at the grain size of letters and phonemes, the overall pattern of weights will reflect correlational
structure at any larger grain size. It turns out that, for English, correlations at the grain size of onsets and rhymes are highly
predictive (Treiman, Mullennix, Bijeljac-Babic, & Richmond-Welty, 1995).
Stone et al. (in press) used four sets of words in a 2 x 2 design. Bidirectionally consistent words were consistent in both
spelling-to-phonology and phonology-to-spelling. An example is the word lust. Its spelling body _UST is only pronounced one
way in the various words that share this spelling body, and its pronunciation body /_ust/ is only spelled one way in the words
that

Page 181
Fig. 10.2.
A highly simplified illustration of micro dynamics that describe reading and spelling performance for the word Hi.
The indices refer to the positions of the letters or phonemes in the word Hi. A recurrent network model
presented with the written word Hi sends feed forward activation from letter nodes to phoneme nodes (Fig. 2a) and,
in turn, feeds back activation from phoneme nodes to letter nodes (Fig. 2b). Figure 2c illustrates a resonance that
emerges between letter and phoneme nodes corresponding to Hi.

Page 182
share this pronunciation body. A second set of words were bidirectionally inconsistent. For example, the spelling body _EAK in
bleak has multiple pronunciations, as in break and leak, and the pronunciation body /_eak/ has multiple spellings, as in freak
and creek. The third and fourth sets of words were consistent in one direction but inconsistent in the other. The example heap
has a spelling body _EAP that is always pronounced the same, but the pronunciation body /_eap/ can be spelled multiple ways,
as in creep and leap. The contrasting example hull has a pronunciation body /_ull/ that can only be spelled one way, but a
spelling body _ULL that can be pronounced multiple ways, as in dull and pull.
In a lexical-decision task, words that were consistent in both directions yielded faster correct "yes" response times than words
that were inconsistent in either direction. Recently, Patrice Gibbs (personal communication, May 1995) found similar effects in a
naming task. Thus, the fact that a pattern of phonology can be spelled more than one way even affects performance in simple
reading tasks. These empirical findings agree with our suggestion that spelling and reading are fundamentally related.
Returning to Fig. 10.2c, reading the word hi not only activates the phonemes [h1] and [a12], and the letters H1 and I2 but also
all multiple possible alternative spellings of the phoneme [a12]. Thus, reading a word correctly requires that incorrect phoneme
nodes are inhibited, just as spelling a word correctly would require that incorrect letter nodes are inhibited. In the case of
reading, however, the letters are presented to the model, so phoneme-letter inconsistency is less likely to yield an incorrect
pattern of activation across letter nodes. The match between letters presented to the model and pattern of letter activation fed
back from phoneme and semantic nodes actually accelerates the formation of correct feedback loops (as illustrated by the bold
arrows in Fig. 10.2c). In spelling, however, a model must generate this pattern from phonology and semantic activation alone.
There is no environmental support for correct letter nodes. Activation of letter nodes is determined by micro relations between
phonemes and letters and macro relations among phonology, semantics, and spelling.
In English orthography, there are generally more possible spellings for a particular word than possible readings; for example,
the phoneme [i:] has the following possible spellings: Y as in entry, EY as in key, EE as in deep, EA as in leaf, and IE as in
chief. Another example is the 36 possible spellings of the word she (see Simon & Simon, 1973).4 This is the micro basis for the
asymmetry between spelling and reading (Barry, 1994; Stone et al., in press; Waters et al., 1985). Stone et al. (in press)
estimated that 69% of low-frequency English one-syllable words are letter-to-phoneme consistent (at the
4 In the case of she, according to Simon and Simon (1973), /sh/ can be spelled in nine different ways (ti, sh, ci, ssi, si, c,
ch, t, s), and /e/ in four different ways (e, ea, ee, ie).

Page 183
grain size of onsets and rhymes) but that 72% of this sample are phoneme-to-letter inconsistent (at that same grain size).
Another estimate indicated that 72% of all letter-to-phoneme, consistent, one-syllable words were phoneme-to-letter
inconsistent, which suggests that phoneme-to-letter inconsistency is the rule for English. This characteristic is not restricted to
English, however. Alphabetic orthographies such as Dutch, French, German, and Spanish are also more inconsistent in their
phoneme-to-letter relations than in their letter-to-phoneme relations (the exceptions being unpointed Arabic and unpointed
Hebrew; see Berent, Frost, & Perfetti, chap. 11, this volume).
Both spelling and reading are powerfully constrained by the relatively strong correlational structure of the letter-phoneme
dynamic. However, inconsistencies in these relations must be resolved by other sources of constraint. When a model "reads" a
low-frequency, letter-to-phoneme inconsistent word such as pint, the more consistent relation between spelling and phonology
rhymes with mint (Kawamoto & Zemblidge, 1992). Additionally, the letter-phoneme dynamic yields two correct pronunciations
for words such as wind, although it would favor the more regular pronunciation (Kawamoto & Zemblidge). In both these cases,
relatively strong semantic-phoneme relations may supply sufficient constraints for the appropriate phonology (Strain, Patterson,
& Seidenberg, 1995). In the case of wind, semantic constraints may also be due to context. Contextual sources of semantic
activation contribute via the relatively strong connections between semantic and phoneme nodes (cf. Azuma & Van Orden, in
press; Stone & Van Orden, 1989).
Now consider spelling, however. It must resolve the inverted patterns of ambiguity in the phoneme-letter dynamic. When a
model "spells" a low-frequency, phoneme-to-letter inconsistent word, such as heap, the more consistent spelling for /_eep/
would be as in deep. Additionally, the phoneme-letter dynamic yields two correct spellings for homophones such as deer/dear.
In these cases, correct spelling relies on relatively weak semantic-letter relations to supply sufficient activation of the
appropriate letters (as illustrated in Fig. 10.1). Even contextual support is filtered through the weaker connections between
semantic nodes and letter nodes. This relatively weak support for spelling (the semantic-letter relations) compared to the
stronger one for reading (the semantic-phoneme relations) is the macro basis for the asymmetry between spelling and reading.
To summarize: Spelling is more difficult than reading because phoneme-letter relations are more inconsistent than letter-
phoneme relations and because the phoneme-letter inconsistencies must be resolved by the relatively weak semantic-letter
dynamic, whereas, in reading, letter-phoneme inconsistencies are resolved by the stronger semantic-phoneme dynamic. The fact
that we engage less in spelling than in reading enhances this fundamental asymmetry between spelling and reading. Thus, to
achieve a

Page 184
spelling level that is comparable to reading would require building stronger correlations between words' meanings and their
spellings (or a language with fewer inconsistent phoneme-to-letter relations). Finally, the existence of powerful bidirectional
correlations between letter and phoneme nodes explains why both spelling and reading appear to be phonologically mediated. In
particular, this explains the preponderance of phonetically accurate errors in spelling.
The Relation Between Learning to Spell and Learning to Read

The theoretical account presented previously assumes a close relationship between spelling and reading, but it also predicts an
asymmetry between the two skills. Our earlier examples showed that hundreds of readings of the (proper) spelling of a word do
not guarantee correct spelling.
Beyond circumstantial evidence, experiments with beginning readers also show that reading is not always effective for learning
to spell. For example, Dutch children with 7 months of formal spelling and reading instruction made equal numbers of spelling
errors in new words they read 18 times and in new words they read only three times (Bosman & de Groot, 1991). In another
study, Dutch children with 10 months of formal instruction made equal numbers of errors in new words they read six times and
in words they read only twice (Bosman & de Groot, 1992). Only with 10 months of formal reading and spelling instruction and
words presented at least nine times did spelling performance improve through reading (Bosman & de Groot, 1991). Moreover,
spelling performance of adults who read pseudowords once, twice, three times, four times, or eight times, only benefited from
the "eight times" condition (Gompel, Tromp, de Vries, & Bosman, 1990).
In most of the aforementioned studies, spelling performance was measured by means of a dictation test. That is, the
experimenter read the words, and the child was asked to write them down. In one study, however, spelling performance was
assessed in both a dictation task and by means of a two-choice recognition test (Bosman & de Groot, 1992). In the two-choice
recognition task, the child was presented with the word BLAUW (blue), and a phonologically plausible misspelling BLOUW,
and was asked to pick the correct spelling. As in the previous studies, it did not matter whether the child had seen (i.e., read) the
words twice or six times. The recognition scores were the same in both cases, but recognizing the correct spelling was easier
than producing it even after the recognition scores were adjusted for guessing. This finding is interesting. It indicates that
assessment of spelling skill is strongly task dependent. Children (and adults alike) appear as better spellers when their
knowledge of words' spellings is tested

Page 185
in a recognition task. More importantly, it is consistent with our explanation as to why spelling is more difficult than reading.
Recognizing spelling is more similar to reading because the child can see the letters of the correct spelling. Alternatively,
producing a spelling requires the child to generate the correct spelling from phonology and semantics.
In other studies, we compared the effectiveness of reading practice for learning to spell with explicit methods of spelling
practice: copying (writing down the word into a notebook), problem naming (practicing the ambiguous phoneme(s)), visual
dictation (writing down the word after it has been presented for a brief period of time), word composition (making up the word
using letter tiles), and oral spelling (spelling the word aloud). Children benefited most from oral spelling, and adults benefited
most from oral spelling while learning to spell pseudowords. Most importantly, however, all spelling-instruction methods were
superior to just reading as a means to learn the spelling (Bosman & de Groot, 1992, Bosman & van Leerdam, 1993; Gompel,
Tromp, de Vries, & Bosman, 1990; van Daal, van der Leij, & Geertvliet-van der Hart, 1989; van Doorn-van Eijsden, 1984).
Another interesting result from the previous studies is that spelling performance benefited most when the whole word was
practiced. In principle, practicing the spelling of the whole word is redundant in a relatively shallow orthography such as Dutch
because the major part of each word is unambiguous and can be derived by phoneme-to-letter rules. Thus, practicing the
ambiguous part of the word, as in the problem-naming method, should be as effective as practicing the whole word (as in oral
spelling). Nevertheless, the whole-word, oral-spelling condition yielded superior spelling performance for all participants at all
ages. This finding is also interesting with respect to our account of spelling and reading. General knowledge about the
relationships between letters and phonemes is not sufficient for correct spelling. The spelling context in which letters are
embedded, which is unique for all words except homographs (e.g., wind), contributes strongly to knowledge about the spelling
of words. In our account, we illustrated the intricate interplay between letters and phonemes by showing how each letter, each
phoneme, and each semantic node contributed to the emergence of a word's spelling or a word's reading.
Again, reading is not a very effective way to learn the spelling of a word. This raises the complementary question: Is learning to
spell an effective way for learning to read? Again, most orthographies are more consistent in their letter-to-phoneme relation
than in their phoneme-to-letter relation. It is possible that learning about the phoneme-to-letter relation strengthens the letter-to-
phoneme relation. However, studies that have investigated this issue yielded conflicting results, and results that are difficult to
interpret. Uhry and Shepherd (1993) conducted a training study in which beginning readers from Grade 1 were instructed in
phonemic segmentation and spelling. Their

Page 186
findings indicated that the spelling training was beneficial for reading, but the training method left the participants ample time to
practice reading as well because the words remained visible during the training. This confound masks the basis of better reading
performance. The same problem occurs for results from studies by Ehri and Wilce (1987) and Roberts and Ehri (1983).
Bosman, de Groot, and van Leerdam (1995) conducted spelling-training experiments in which reading and spelling were
disentangled. Children and adults practiced the spellings of letter strings they had not seen before, and they were never visually
confronted with the words. First graders were taught the spellings of actual words, whereas the adults learned the spellings of
pseudowords. The spelling training was a form of the oral-spelling instruction method. After the experimenter named the word
(or pseudoword), the participants spelled the word aloud and appropriate feedback was provided. Each word was practiced four
times. The experiment included two control conditions. In one control condition, a set of words was repeated orally after the
experimenter, and the second control condition included to-be-spelled words that were not practiced at all. Later, in a naming
task, orally spelled and control repetition words were read faster than the words that were not practiced in the training, but there
was no effect of oral spelling over simple repetition. A subsequent test showed that both groups learned the correct spelling of
the oral-spelling items but not those of the items in the repetition condition. These results suggest that learning the spelling of
words without visual presentation does not benefit reading performance over mere repetition. This issue is a subject of our
ongoing research.
We conclude this chapter with some general remarks concerning the implications for education.
Some Implications for Spelling Instruction

In this final section, the discussion emphasizes formal spelling instruction of normally developing children in the Netherlands,
but the general implications hold for other languages too. As becomes clear, spelling difficulties should not be viewed as a static
problem; they change over time. Over 80% of Dutch schools use a phonics-based reading-and-spelling method called "Veilig
leren lezen" (Learning to read safely, Caesar, 1979). It is a fairly rigid preprogrammed curriculum with a strict day-to-day and
week-to-week progression. Initially, the child is exposed to monosyllabic words that are predominantly consistent in their letter-
to-phoneme relations. Thus, proper phonemic analysis provides the beginning speller with predictable spelling patterns.
After 4 months, children are presented with words containing inconsistent phoneme-to-letter relations. In first grade, this is
mainly limited to the

Page 187
following phonemes: [ei] can be spelled IJ as in blij (happy), or EI as in geintje (joke); [ou] can be spelled AU as in paus (pope),
or OU as in touw (rope); and [x] can be spelled G as in zeg (say), or CH as in lach (laugh). Dutch first graders primarily face the
problem of spelling ambiguity of words with historically determined spelling patterns (cf. Assink, 1990). The spelling pattern of
these words must be memorized because phoneme-to-letter analysis yields ambiguity in the letter pattern, and there are no rules
to disambiguate these patterns.
At the end of first grade, children are presented with spelling ambiguities that can be solved by applying rules. Spelling plural
forms of nouns generally consists of adding the morpheme -EN at the end of the singular form. For example, the phoneme-to-
letter, consistent singular noun boek [book] becomes boeken, which is a phoneme-to-letter, consistent plural noun. However, a
large proportion of phoneme-to-letter consistent singular nouns, for example, sok [sock], have a plural form that is phoneme-to-
letter inconsistent, namely, sokken (and not soken). Phonemic analysis of sokken is ambiguous with respect to the spelling of the
phoneme [k]. An additional rule is needed to correctly spell the word, namely, if the body of a singular noun consists of a
single-letter vowel and a consonant, then double the final consonant before adding the EN-morpheme (compare the English
spelling rule of doubling the final letter B of rub in rubbing, but not the final K in looking).
Other types of spelling problems arise with increasing reading and spelling experience. The majority of phoneme-to-letter
inconsistent words can be spelled correctly by the application of a set of linguistic rules. But even the most skilled adult spellers
are usually unaware that these rules exist or are unable to apply them because of their rather complex nature (Assink, 1985,
1990; Verhoeven, 1979). In most cases, therefore, children and adults alike memorize the spellings of words with ambiguous
spelling patterns.
The problems for Dutch spellers in first grade begin a neverending progression. Spellers encounter increasingly complex
relations between phonemes and letters (Verhoeven, 1979; see Bailet, 1990, for the English language). The types of problems
that appear in lower grades disappear in higher grades but are replaced with new types of problems. In turn, these problems
disappear, to be replaced yet again by new spelling problems on entering high school. Changing spelling problems poses a
challenge for the provision of detailed directions, but some implications, all related to the assumption of phonologically
mediated spelling, are less specific to a particular type of spelling problem.
For most people spelling acquisition lasts until adulthood, and for some people, learning words' spellings is a life-long
enterprise. Dutch spellers, and probably most non-English speakers face an additional problem. New words, in particular from
English, are continuously being added to their languages, and these loan words are, by nature, sound-to-spelling inconsistent.
The

Page 188
spelling of loan words with highly inconsistent phoneme-to-letter patterns may best be learned by phonologically regularizing
the pronunciation of a word (Brten, 1994; Ormrod & Jenkins, 1989). In fact, corned beef, rawhide, and palmolive are
regularized in standard Dutch and do not retain the English pronunciation (/kor/ /net/ /bief/, /ra/ /we/ /de/, and /pol/ /mo/ /lee/
/ve/, respectively). Pronouncing an inconsistent word as a letter-to-phoneme regular word may also serve as a mnemonic for
deriving its correct spelling. Examples are "private pronunciations" of the word aisle as /ai/, /s/, /le/, or the word Wednesday as
/wed/, /nes/, /day/. This method stems directly from the assumption that spelling and reading are phonologically mediated.
Another issue related to the assumption of phonologically mediated spelling is the role of dialects in education. Dutch spelling is
based on Standard Dutch (Algemeen Beschaafd Nederlands, ABN). Children who do not also speak Standard Dutch cannot take
full advantage of the consistency between phonemes and letters. Various examples from spelling tests show the influence of
children's dialect on their spelling performance (Bosman & van Leerdam, 1993). However, there are also situations in which
children who speak a dialect can use the knowledge of their local dialect to spell otherwise, ambiguous spelling patterns.
Children in De Achterhoek and De Veluwe (middle-eastern part of The Netherlands) generally know the spelling of the
phoneme [ei] because their pronunciation of words spelled with IJ as in mijn (/mien/ meaning `mine') is different from the EI as
in klein, /klein/ meaning `small'), whereas in the rest of the country, the two spelling patterns IJ and EI have identical
pronunciations (/mein/and /klein/).
A final issue concerns the use of spelling tests that present incorrect spellings to students in the form of multiple-choice tests. In
a multiple-choice test, the correct spelling of a word is presented together with one or more incorrect (pseudohomophone)
spelling(s), and students are required to mark the correct spelling, for example, harrass, harass or harras; a Dutch example is:
enigzins, enigszins (somewhat), or enigsins. An important educational question is whether spelling performance is affected by
encountering phonologically acceptable misspellings. We both experienced the erosion of spelling knowledge after conducting
large numbers of experiments using pseudohomophones. These private experiences are corroborated by experimental studies
(see Brown, 1988, for an overview; see Ehri, Gibbs, & Underwood, 1988, for a null effect). It appears that being visually
exposed to incorrect, but phonologically possible spellings may have an adverse effect on extant spelling knowledge. Thus,
despite the convenience of multiple-choice tests, they may not be the best possible way of testing spelling performance (Bradley
& King, 1992).
We conclude that spelling and reading are interdependent, and phonology mediates both of them. Reading, however, is not the
most effective way to learn to spell. Correct spelling seems to require strategies specific to this

Page 189
skill strategies that strengthen or supplement semantic-letter, correlational structure. Apparently, the most effective methods of
instruction mimic actual spelling production. However, which method of instruction is most effective may change over time as
children encounter an increasing variety of complex relations between phonology and spelling. Successful instruction must take
all this and more into account, keeping in mind the unique matrix of proclivity, background, and motivation that each child
brings to this task.
Acknowledgments
We thank Iris Berent and Martin van Leerdam for their comments on a previous version of this article. We also thank Martin for
creating the figures. Preparation of this article was funded by a NATO fellowship N58-92 awarded to Anna M. T. Bosman and a
National Institute of Health FIRST Award CM 5 R29 NS26247-05 awarded to Guy C. Van Orden.
Anna M. T. Bosman, Department of Special Education, University of Nijmegen, P.O. Box 9104, 6500 HE Nijmegen, The
Netherlands, e-mail: a.bosman @ped.kun.nl; and Guy C. Van Orden, Cognitive Systems Group, Department of Psychology,
Arizona State University, Tempe, AZ 85287-1104, USA, e-mail: guy.van.orden@asu.edu
References
Assink, E. M. H. (1985). Assessing spelling strategies for the orthography of Dutch verbs. British Journal of Psychology, 76,
353-363.
Assink, E. M. H. (1990). Learning to spell. In P. Reitsma & L. Verhoeven (Eds.), Acquisition of reading in Dutch (pp. 65-76).
Dordrecht, The Netherlands: Foris Publications.
Azuma, T., & Van Orden, G. C. (in press). Why SAFE is better than FAST: Relatedness between a word's meanings affects
lexical decision performance. Journal of Memory and Language.
Bailet, L. L. (1990). Spelling rule usage among students with learning disabilities and normally achieving students. Journal of
Learning Disabilities, 23, 121-128.
Barry, C. (1994). Spelling routes (or roots or rutes). In G. D. A. Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory,
process and intervention (pp. 27-49). New York: Wiley.
Berent, I., & Perfetti, C. A. (1995). A Rose is a REEZ: The two-cycles model of phonology assembly in reading English.
Psychological Review, 102, 146-184.
Bosman, A. M. T. (1994). Reading and spelling in children and adults: Evidence for a single-route model. Doctoral
dissertation, University of Amsterdam. Dissertatie reeks 1994-2, Faculteit Psychologie, Amsterdam, The Netherlands.
Bosman, A. M. T., & de Groot, A. M. B. (1991). De ontwikkeling van woordbeelden bij beginnende lezers en spellers [The
development of orthographic images in beginning readers and spellers]. Pedagogische Studin, 68, 199-215.
Bosman, A. M. T., & de Groot, A. M. B. (1992). Differential effectiveness of reading and non-reading tasks in learning to spell.
In F. Satow & B. Gatherer (Eds.), Literacy without frontiers. Widnes, Cheshire, UK: United States Kingdom Reading
Association.

Page 190
Bosman, A. M. T., & de Groot, A. M. B. (1995). Evidence for assembled phonology in beginning and fluent readers as assessed
with the first-letter-naming task. Journal of Experimental Child Psychology, 59, 234-259.
Bosman, A. M. T., & de Groot, A. M. B. (1996). Phonologic mediation is fundamental to reading: Evidence from beginning
readers. The Quarterly Journal of Experimental Psychology, 49A, 715-744.
Bosman, A. M. T., de Groot, A. M. B., & van Leerdam, M. (1995). Orthographic knowledge in reading and the nature of the
spelling process. Manuscript submitted for publication.
Bosman, A. M. T., & van Leerdam, M. (1993). Aanvankelijk spellen: de dominantie van de verklankende spelwijze en de
geringe effectiviteit van lezen als spellinginstructie methode. [Beginning spelling: prevalence of the phonologic strategy in
spelling and the limited effect of reading as a spelling-instruction method]. Pedagogische Studin, 70, 28-45.
Bradley, J. M., & King, P. V. (1992). Effects of proofreading on spelling: How reading misspelled and correctly spelled words
affects spelling accuracy. Journal of Reading Behaviour, 24, 413-432.
Brten, I. (1994). Learning to spell. Oslo: Scandinavian University Press.
Brown, A. S. (1988). Encountering misspellings and spelling performance: Why wrong isn't right. Journal of Educational
Brown, G. D. A., & Ellis, N. C. (1994). Issues in spelling research: An overview. In G. D. A. Brown & N. C. Ellis (Eds.),
Handbook of spelling: Theory, process and intervention (pp. 3-25). New York: Wiley.
Brown, G. D. A., & Loosemore, R. P. W. (1994). Computational approaches to normal and impaired spelling. In G. D. A.
Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 319-335). New York: Wiley.
Bruck, M. (1988). The word recognition and spelling of dyslexic children. Reading Research Quarterly, 23, 51-69.
Bruck, M., & Treiman, R. (1990). Phonological awareness and spelling in normal children and dyslexics: The case of the initial
consonant clusters. Journal of Experimental Child Psychology, 50, 156-178.
Bruck, M., & Waters, G. S. (1988). An analysis of the spelling errors of children who differ in their reading and spelling skills.
Applied Psycholinguistics, 9, 77-92.
Bryant, P. E., & Bradley, L. (1980). Why children sometimes write words which they do not read. In U. Frith (Ed.), Cognitive
Caesar, F. B. (1979). Veilig leren lezen: Structuurmethode voor het aanvankelijk leesonderwijs [Learning to read safely:
Structure method for teaching beginning reading]. Tilburg: Zwijsen.
Campbell, R. (1994). Spelling in prelingual deafness. In G. D. A. Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory,
process and intervention (pp. 249-259). New York: Wiley.
Campbell, R., & Coltheart, M. (1984). Gandhi: The nonviolent route to spelling reform? Cognition, 17, 185-192.
Carello, C., Turvey, M. T., & Lukatela, G. (1992). Can theories of word recognition remain stubbornly nonphonological? In R.
Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 211-226). Amsterdam: North-Holland.
Coltheart, M. (1978). Lexical access in simple reading tasks. In G. Underwood (Ed.), Strategies of information processing (pp.
151-216). New York: Academic Press.
Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993). Models of reading aloud: Dual-route and parallel-distributed
approaches. Psychological Review, 100, 589-608.
Cromer, R. F. (1980). Spontaneous spelling by language-disordered children. In U. Frith (Ed.), Cognitive processes in spelling
(pp. 406-421). London: Academic Press.
Dodd, B. (1980). The spelling abilities of profoundly prelingually deaf children. In U. Frith (Ed.), Cognitive processes in
spelling (pp. 423-440). London: Academic Press.
Duighuisen H. C. M., & van Bon, W. H. J. (1992). Spellen is onverwacht gemakkelijker dan segmenteren [Spelling is
unexpectedly easier than segmentation]. Stem-, Spraak-, en Taalpathologie, 1, 1-13.

Page 191
Ehri, L. C. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive processes in spelling (pp. 311-338).
Ehri, L. C., Gibbs, A. L., & Underwood, T. L. (1988). Influence of errors on learning the spellings of English words.
Contemporary Educational Psychology, 13, 236-253.
Ehri, L. C., & Wilce, L. S. (1987). Does learning to spell help beginners learn to read words? Reading Research Quarterly, 22,
47-65.
Frith, U. (1979). Reading by eye and writing by ear. In P. A. Kolers, M. Wrolstad, & H. Bouma (Eds.), Processing of visible
language, vol. 1 (pp. 379-390). New York: Plenum.
Academic Press.
Frith, U. (1984). Specific spelling problems. In R. N. Malatesha & H. A. Whitaker (Eds.), Dyslexia: A global issue (pp. 83-103).
Den Haag, The Netherlands: Martinus Nijhoff.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. E. Patterson, J. C. Marshall, & M. Coltheart (Eds.),
Surface dyslexia (pp. 301-330). Hillsdale, NJ: Lawrence Erlbaum Associates.
Frith, U., & Frith, C. (1980). Relationships between reading and spelling. In J. F. Kavanagh & R. L. Venezky (Eds.),
Orthography, reading, and dyslexia (pp. 287-295). Baltimore: University Park Press.
Gibbs, P., & Van Orden, G. C. (1995). Pathway selection's utility for strategic control in word recognition. Manuscript
submitted for publication.
Gibson, E. J., Shurcliff, A., & Yonas, A. (1970). Utilization of spelling patterns by deaf and hearing subjects. In. H. Levin & J.
P. Williams (Eds.), Basic studies in reading (pp. 57-73). New York: Basic Books.
Gompel, M., Tromp, J., de Vries, R., & Bosman, A. M. T. (1990). De effectiviteit van spelling-strategien bij volwassenen [The
effectiveness of spelling strategies in adults]. Research Paper, Vakgroep Psychonomie, Universiteit van Amsterdam.
Gough, P. B., Juel, C., & Griffith, P. L. (1992). Reading, spelling, and the orthographic cipher. In P. B. Gough, L. C. Ehri, & R.
Treiman (Eds.), Reading acquisition (pp. 35-48). Hillsdale, NJ: Lawrence Erlbaum Associates.
Grossberg, S., & Stone, G. O. (1986). Neural dynamics of word recognition and recall: Priming, learning, and resonance.
Hatfield, F. M., & Patterson, K. E. (1983). Phonological spelling. Quarterly Journal of Experimental Psychology, 35A, 451-468.
Henderson, E. H., & Beers, J. W. (1980). Developmental and cognitive aspects of learning to spell: A reflection of word
knowledge. Newark, DE: International Reading Association.
Holligan, C., & Johnston, R. S. (1991). Spelling errors and phonemic segmentation ability: the nature of the relationship.
Journal of Research in Reading, 14, 21-32.
Holmes, V. M., & Ng, E. (1993). Word-specific knowledge, word-recognition strategies, and spelling ability. Journal of
Memory and Language, 32, 230-257.
Houghton, G., Glasspool, D. W., & Shallice, T. (1994). Spelling and serial recall: Insights from a competitive queuing model. In
G. D. A. Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 365-404). New York: Wiley.
Kawamoto, A., & Zemblidge, J. (1992). Pronunciation of homographs. Journal of Memory and Language, 31, 349-374.
Klnhammer, M. (1990). Ghandi or Gandhi? Positional frequencies in spelling. Honour Thesis, Department of Psychology,
University of Amsterdam, The Netherlands.
Kreiner, D. S., & Gough, P. B. (1990). Two ideas about spelling: Rules and word-specific memory. Journal of Memory and
Language, 29, 103-118.
Lennox, C., & Siegel, L.S. (1993). Visual and phonological spelling errors in subtypes of children with learning disabilities.
Lundberg, I., Frost, J., & Peterson, O. P. (1988). Effects of an extensive program for stimulating phonological awareness skills
in kindergarten. Reading Research Quarterly, 23, 263-284.

Page 192
Manrique, A. M. B., & Signorini, A. (1994). Phonological awareness, spelling and reading abilities in Spanish-speaking
children. British Journal of Educational Psychology, 64, 429-439.
Marcel, T. (1980). Phonological awareness and phonological representation: Investigation of a specific spelling problem. In U.
Frith (Ed.), Cognitive processes in spelling (pp. 373-403). London: Academic Press.
Marsh, G., Friedman, M., Welch, V., & Desberg, P. (1980). The development of strategies in spelling. In U. Frith (Ed.),
McClelland, J. L., & Rumelhart, D. E. (1981). An interactive-activation model of context effects in letter perception, Part 1: An
account of basic findings. Psychological Review, 88, 375-407.
Moats, L. C. (1993). Spelling error interpretation: Beyond the phonetic/dysphonetic dichotomy. Annals of Dyslexia, 43, 174-
185.
Mommers, M. J. C. (1987). An investigation into the relation between word recognition skills, reading comprehension and
spelling skills in the first two years of primary school. Journal of Research in Reading, 10, 122-143.
Nelson, H. E. (1980). Analysis of spelling errors in normal and dyslexic children. In U. Frith (Ed.), Cognitive processes in
spelling (pp. 475-493). London: Academic Press.
Nicolson, R. I., & Fawcett, A. J. (1994). Spelling remediation for dyslexic children: A skills approach. In G. D. A. Brown & N.
C. Ellis (Eds.), Handbook of spelling: Theory, process, and intervention (pp. 505-528). New York: Wiley.
Olson, A., & Caramazza, A. (1994). Representation and connectionist models: The NET spell experience. In G. D. A. Brown &
N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 337-363). New York: Wiley.
Ormrod, J. E., & Jenkins, L. (1989). Study strategies for learning spelling: Correlations with achievement and developmental
changes. Perceptual and Motor Skills, 68, 643-650.
Perfetti, C. A. (1992). The representation problem in reading acquisition. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.),
Perfetti, C. A., Zhang, S., & Berent, I. (1992). Reading in English and Chinese: Evidence for a ''universal'' phonological
principle. In R. Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 227-248). Amsterdam: North-
Holland.
Porpodas, C. D. (1987). The phonological factor in reading and spelling of Greek. In P. G. Aaron & R. M. Joshi (Eds.), Reading
and writing disorders in different orthographic systems (pp. 177-190). Dordrecht, The Netherlands: Kluwer.
Putnam, H. (1981). Reductionism and the nature of psychology. In J. Haugeland (Ed.), Mind design (pp. 205-219). Cambridge,
MA: MIT Press.
Read, C. (1971). Preschool children's knowledge of English phonology. Harvard Educational Review, 41, 1-34.
Read, C. (1981). Writing is not the inverse of reading for young children. In C. H. Frederiksen & J. F. Dominic (Eds.), Writing:
Process, development and communication (pp. 105-118). Hillsdale, NJ: Lawrence Erlbaum Associates.
Roberts, K. T., & Ehri, L. C. (1983). Effects of two types of letter rehearsal on word memory in skilled and less skilled
beginning readers. Contemporary Educational Psychology, 8, 375-390.
Seidenberg, M. S., & McClelland, J. L. (1989). A distributed developmental model of word recognition and naming.
Seymour, P. H. K., & Porpodas, C. D. (1980). Lexical and non-lexical processing of spelling in dyslexia. In U. Frith (Ed.),
Siegel, L. (1994). The modularity of reading and spelling: Evidence from hyperlexia. In G. D. A. Brown & N. C. Ellis (Eds.),
Handbook of spelling: Theory, process and intervention (pp. 227-248). New York: Wiley.

Page 193
Simon, D. P., & Simon, H. A. (1973). Alternative uses of phonemic information in spelling. Review of Educational Research,
43, 115-137.
Sloboda, J. A. (1980). Visual imagery and individual differences in spelling. In U. Frith (Ed.), Cognitive processes in spelling
Snowling, M. J. (1994). Towards a model of spelling acquisition: The development of some component processes. In G. D. A.
Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 111-128). New York: Wiley.
Stone, G. O., & Van Orden, G. C. (1989). Are words represented by nodes? Memory and Cognition, 17, 511-524.
Stone, G. O., & Van Orden, G. C. (1994). Building a resonance framework for word recognition using design and system
principles. Journal of Experimental Psychology: Human Perception and Performance, 20, 1248-1268.
Stone, G. O., Vanhoy, M., & Van Orden, G. C. (in press). Perception is a two-way street: Feed-forward and feedback
phonology in visual word recognition. Journal of Memory and Language.
Strain, E., Patterson, K., & Seidenberg, M. S. (1995). Semantic effects in single-word naming. Journal of Experimental
Psychology: Learning, Memory and Cognition, 21, 1140-1154.
Stuart, M., & Masterson, J. (1992). Patterns of reading and spelling in 10-year-old children related to prereading phonological
abilities. Journal of Experimental Child Psychology, 54, 168-187.
Thelen, E., & Smith, L. B. (1994). A dynamic systems approach to the development of cognition and action. Cambridge, MA:
MIT Press.
Thomson, M. (1984). Developmental dyslexia: Its nature, assessment and remediation. London: Edward Arnold.
Treiman, R., Mullennix, J., Bijeljac-Babic, R., & Richmond-Welty, E.D. (1995). The special role of rimes in the description,
use, and acquisition of English orthography. Journal of Experimental Psychology: General, 124, 107-136.
Uhry, J. K., & Shepherd, M. J. (1993). Segmentation/spelling instruction as part of a first-grade reading program: Effects on
several measures of reading. Reading Research Quarterly, 28, 218-233.
uit de Haag, I. J. C. A. F. (1994). Segmentatie- en spellingfouten van beginnende spellers: de invloed van fonologische factoren
[Segmentation and spelling errors in beginning spellers: The effect of phonologic factors]. Honour Thesis, Department of
Special Education, University of Nijmegen, The Netherlands.
Valtin, R. (1987). Dyslexia in the German language. In P. G. Aaron & R. M. Joshi (Eds.), Reading and writing disorders in
different orthographic systems (pp. 119-135). Dordrecht, The Netherlands: Kluwer.
van Bon, W. H. J., & Duighuisen, H. C. M. (1995). Sometimes spelling is easier than phonemic segmentation. Scandinavian
Journal of Psychology, 36, 82-94.
van Daal, V. H. P., van der Leij, A., & Geertvliet-van der Hart, J. A. (1989). Het effect van type-oefeningen op het leren lezen
en spellen van kinderen met ernstige leesproblemen [Effects of typewriting practice on reading and spelling in children with
severe reading disability]. Pedagogische Studin, 66, 185-1192.
van Doorn-van Eijsden, M. (1984). Leer je spellen door veel te lezen? [Do you learn to spell by reading a lot?]. Tijdschrift voor
Taalbeheersing, 6, 252-263.
Van Orden, G. C., & Goldinger, S. D. (1994). Interdependence of form and function in cognitive systems explains perception of
printed words. Journal of Experimental Psychology: Human Perception and Performance, 20, 1269-1291.
Van Orden, G. C., & Goldinger, S.D. (1996). Phonologic mediation in skilled and dyslexic reading. In C. H. Chase, G.D. Rosen,
& G.F. Sherman (Eds.), Developmental dyslexia: Neural, cognitive, and genetic mechanisms (pp. 185-223). Timonium, MD:
York Press.

Page 194
Van Orden, G. C., Pennington, B. F., & Stone, G. O. (1990). Word identification in reading and the promise of subsymbolic
psycholinguistics. Psychological Review, 97, 488-522.
Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. Cambridge,
MA: MIT Press.
Verhoeven, G. (1979). Verbeelding en werkelijkheid; Spelfouten in de opstellen uit verschillende typen van onderwijs
[Imagination and reality: Spelling errors in essays from different types of education]. Tijdschrift voor Taalbeheersing, 1-2, 146-
163.
Waller, R. J. (1988). Just beyond the firelight. Ames, IA: Iowa State University Press.
Ziegler, J. C., & Jacobs, A. M. (1995). Phonological information provides early sources of constraint in the processing of letter
strings. Journal of Memory and Language, 34, 567-593.

Page 195
Chapter 11
The Inhibition of Polygraphic Consonants in Spelling Hebrew: Evidence for Recurrent Assembly of Spelling and Phonology in
Visual Word Recognition
Iris Berent
Florida Atlantic University
Ram Frost
Hebrew University
Skilled spelling entails two subcomponents. One permits the spelling of new words and nonwords. A second subcomponent
entails the retrieval of idiosyncratic word-specific orthographic knowledge. Spelling models (e.g., Barry, 1994; Jorm, 1983;
Kreiner, 1992; Link & Caramazza, 1994) account for these two abilities by postulating two strategies for assigning an
orthographic representation to a word. One is a productive strategy that assembles a spelling representation by mapping the
word's phonemic constituents onto graphemes. Such a strategy is necessary to account for the spelling of nonwords such as
/blin/. Because /blin/ does not correspond to any English word, its spelling cannot be obtained by consulting the spelling of any
particular English word. Instead, it must be generated "from scratch." Spellers can do so by mapping each of the phonemes of
/blin/ onto a letter. There is evidence that the assembly strategy contributes also to the spelling of familiar words. Supporting the
contribution of the assembly process in spelling are the findings that most spelling errors among good (Barron, 1980; Frith,
1980) as well as poor (Bosman, 1994; Frith 1980) spellers are phonologically plausible, that a child's early spellings are affected
by the phonetic and phonological properties of the word (Read, 1971; Treiman, 1994), and

Page 196
that the performance of skilled spellers in on-line spelling tasks is impaired by polygraphy (Kreiner, 1992). However, the
assembly of phonemes into letters is insufficient to account for skilled spelling in many orthographies in which spelling cannot
be fully predicted from the word's phonological structure. For instance, the assembly of a spelling to the word rain may yield
also the representation rane. To correctly spell the word rain, readers must rely on a second strategy. This strategy consults an
orthographic representation that is specific to this word. It accesses the orthographic lexicon and addresses a stipulated
orthographic representation of particular words. The contribution of word-specific knowledge in spelling is supported by the
finding that the advantage of good versus poor spellers is particularly noticeable for words whose spelling requires word-
specific knowledge (i.e., exception words; Barron, 1980).
Although it is clear that both strategies are necessary for skilled spelling, there are questions regarding their relative contribution.
In particular, these questions concern the role of the assembly mechanism in the spelling of familiar words. Because spellers
possess a stipulated knowledge of the word as a whole, the application of the assembly procedure seems unnecessary for such
words. The usefulness of this productive mechanism is further reduced in orthographies that exhibit unpredictable mapping
functions between phonemes and graphemes. Consider again the word rain. The spelling of the vowel phoneme in this word is
unpredictable because English supports several ways to spell the vowel that are all quite common. Conversely, English spelling
may be unpredictable due to the use of rare, irregular correspondences between graphemes and phonemes (e.g., gauge).
Orthographies that manifest opaque correspondences between graphemes and phonemes are termed deep orthographies (for a
discussion, see Katz & Frost, 1992). In such orthographies, the application of the assembly mechanism is not only unnecessary
but also detrimental because it may result in an erroneous representation.
The primary issue discussed in this chapter concerns the role of the assembly mechanism in a deep orthography among skilled
spellers. Specifically, we are concerned with the following questions: (a) Do skilled spellers assemble a graphemic
representation in spelling familiar words? (b) If an assembled mechanism is activated in spelling, then how can spellers prevent
spelling errors?
The questions regarding the role of assembled information in spelling echoes a parallel debate in the reading literature: Do
skilled readers recognize familiar words by a process of grapheme-phoneme assembly? (for a review, see Berent & Perfetti,
1995; Frost, 1995; Seidenberg & McClelland, 1989; Van Orden, Pennington, & Stone, 1990). On first examination, it seems that
the roles of assembly in the perception (reading) and production (spelling) of written language are quite independent. The
assembly function in

Page 197
spelling maps phonemes to graphemes, whereas, in reading, it converts graphemes to phonemes. The activation of this mapping
function in one direction (e.g., spelling to phonology) does not necessarily implicate its activation in the opposite direction (e.g.
phonology to spelling). Furthermore, the predictability of these mappings may differ in each direction. Some orthographies (e.g.,
French) exhibit predictable grapheme-phoneme correspondences but unpredictable phoneme-grapheme mappings. In other
orthographies (e.g., Arabic, Hebrew), a word's spelling is more predictable than its pronunciation from print.1
What is the relationship between the assembly of spelling and phonology? On one account, the two directions of the assembly
process are independent. Thus, the activation of phonemes from graphemes, for instance, should be unaffected by the
predictability of the correspondence between phonemes and graphemes. To illustrate the predictions of this account, consider
again the word rain. The assembled phonology of this word is perfectly predictable from its spelling. In contrast, the spelling of
rain cannot be predicted from its sound. In reading the word rain, readers may attempt to assemble its phonology. If the
functions of phonology and spelling assembly are independent, then the assembly of phonology in reading rain should not be
affected by the unpredictability of its spelling. However, an alternative account may predict a close relationship between the two
functions of assembly. In this account, assembly is a recurrent process in which graphemes and phonemes are mapped
bidirectionally. Mapping inconsistency in one direction thus constrains the assembly process as a whole (see also Stone &
Vanhoy, 1994; Stone, Vanhoy, & Van Orden, in press). Hence, in spite of the fact that the phonology of rain is completely
predictable from its letters, the assembly of its phonology may, nevertheless, be affected by the unpredictability of its spelling.
As we illustrate in the following sections, Hebrew orthography presents the speller with considerable ambiguity in predicting a
word's spelling from its phonology. This orthography provides fertile grounds for the examination of the dependency between
the two directions of the assembly process. A second goal of this work is to extend the investigation of the relationship between
the assembly of sound and spelling (Stone & Vanhoy; Stone et al., in press) to Hebrew orthography. In this work, we examine
the dependency between the assembly of spelling and phonology: Does the assembly of a
1 We use the term predictability (Berent & Perfetti, 1995) to refer to the degree of certainty in which the output of the
assembly function (e.g., phonemes) can be derived from its input (e.g., graphemes). The unpredictability in mapping could
result either from inconsistency due to the existence of one to many mappings (e.g., the mapping of graphemes to
phonemes in English) or from underspecification of the input (e.g., the absence of vowels in Hebrew; see Frost, 1995).

Page 198
graphemic representation in spelling a word constrain the assembly of phonology from print?
Our investigation of these questions is based on unpointed Hebrew, a script that manifests quite opaque correspondences
between graphemes and phonemes in both directions. We examine whether the process of recognizing isolated Hebrew words is
constrained by the predictability in which Hebrew phonemes are mapped onto graphemes. Specifically, we identify a spelling
mechanism that is designed to protect word spelling from inconsistencies in mapping phonemes and graphemes. Such a
mechanism, we believe, is required only if Hebrew readers assemble (rather than address) the spelling of familiar words. We
then demonstrate that this spelling mechanism is activated in a reading task. Because this task clearly does not require the
assembly of the word's spelling, our findings suggest that the assembly of a word's spelling is quite automatic. Moreover,
evidence for automatic activation of graphemic assembled information in Hebrew, a deep orthography in which the mapping
between phonemes to graphemes is highly unpredictable, indicates that the contribution of spelling assembly is quite general.
We first assess the reliability of an assembly mechanism in obtaining the spelling of Hebrew words by describing some of the
properties of the Hebrew writing system. We then propose a mechanism that permits spellers and readers to benefit from the
assembly of both the word's spelling and its phonology in reading Hebrew, in spite of the fact that the information provided in
the orthography is partial and unreliable. Finally, we provide empirical evidence for an automatic activation of assembled
spelling and phonology in a reading task. We conclude with reviewing some of the implications of our proposal.
The Reliability of the Assembly Mechanism in Reading and Spelling Hebrew: The Effects of Hebrew's Orthographic Structure
Of the properties of Hebrew orthography, we note two features that are particularly important to spelling: its consonantal nature
and its consonantal polygraphy. The consonantal nature of Hebrew orthography is one of the most widely known characteristics
of this writing system. In Hebrew, letters represent mostly consonants, whereas vowels can be optionally superimposed on the
consonants as diacritical marks. The diacritical marks, however, are omitted from most reading materials and are found only in
poetry, children's literature, and religious texts. Thus, in its unpointed form, Hebrew orthography does not convey to the reader
the full phonemic structure of

Page 199
the printed word, and the reader is often faced with phonological ambiguity. In contrast to unpointed orthography, pointed
orthography is a very shallow writing system. Vowel marks convey missing phonemic information, making the printed word
phonemically unequivocal.
The consonantal nature of unpointed script is a property that has obvious implications for reading. In the absence of vowels,
readers of unpointed script are provided with quite impoverished information regarding the phonology of the word.
Consequently, it has been argued that in reading unpointed Hebrew, readers access the word's meaning directly based on its
graphemic representation. On this view, the assembly of the word's phonology from its print provides a relatively weak source
of constraint on reading Hebrew words (Bentin, Bargai, & Katz, 1984; Bentin & Frost, 1987).
The problems posed by the consonantal nature of Hebrew orthography are even more severe when the assembly of the word's
spelling is concerned. Some vowels (mainly /o/, /u/, /i/) may be represented in print not only by points but also by letters. Thus,
in assembling the spelling of Hebrew words, spellers must determine which vowels should be represented as letters and which
ones should be omitted (Shimron, 1993). This decision, however, is completely unpredictable from the phonology of modern
Hebrew.2 Because the representation of vowels in a word's spelling is highly unpredictable, spellers may limit the assembly of a
word's spelling to consonantal information. On this account, the primary task of the speller in assembling a graphemic
representation to Hebrew words is the assembly of the word's consonants. This process, however, is highly unpredictable due to
another characteristic of Hebrew orthography its consonantal polygraphy.
Modern Hebrew contains 18 consonant phonemes. Twelve of these consonants (/b/-/g/-/d/-/h/-/z/-/y/-/l/-/m/-/n/-/p/-/ts/-/r/-/sh/)
are mapped into their letters unequivocally. In contrast, the representation of six consonants (/a//v/-/s/-/x/-/k/-/t/) is quite
unpredictable as they can be represented by two distinct letters (see Fig. 11.1). Thus, the representation of these consonant
phonemes in Hebrew is polygraphic.3 This polygraphy stems from a histori-
2 The distinction between the two forms of vowel spellings, by a marker or by a vowel letter, reflects an ancient
distinction between different forms of vowels that contrast mainly in their length, long vowels being represented by the
letters. This distinction, however, has been lost in modern Hebrew. Thus, modern spellers cannot predict whether a vowel
letter is required based on the phonology of the word. Another possible source of information for determining the
orthographic representation of vowels is an explicit knowledge of the word's morphology. However, such reflective
knowledge may be inaccessible to the process of assembly.
3 Polygraphy is a property that is not unique to Hebrew. For instance, English manifests considerable ambiguity in the
spelling of vowels. In contrast to Hebrew, the ambiguous graphemes tend to be represented by letter combinations rather than
single letters. Consonantal polygraphy in English seems more limited in its scope compared with Hebrew. Similar to vowel
ambiguity, the competing consonant alternatives differ in their complexity. For instance, the phoneme /k/ may be spelled as
either k, c. However, the phonological value of c depends on
(footnote continued on next page)

Page 200
Fig. 11.1.
The polygraphic consonants in Hebrew.
cal phonological difference between these consonants that were present in ancient Hebrew. In fact, some of these phonological
distinctions can be found in some Jewish communities originating from Arab-speaking countries
(footnote continued from previous page)
the following vowel letter. Thus, its mapping is more complex than that of k. The spelling of these consonants may further
be disambiguated by their position in the word and their frequency. The distributions of the two correspondences of /k/
are roughly complementary: The letter k appears in an environment in which c is mapped into /s/ (Venezky, 1970). In
addition, the mapping of /k/ to k is less common than /k/ to c. Further research is required to determine the effect of
consonantal polygraphy in such orthographies.

Page 201
(e.g., Yemen) probably due to the fact that the same phonemic contrasts are distinctive in Arabic. However, most of these
distinctions were lost among the communities originating from central and eastern Europe and they are no longer present among
the majority of Hebrew speakers born in Israel (Morag, 1962).
The presence of consonantal polygraphy poses severe problems in the assembly of a word's spelling. The fact that the number of
graphemes in Hebrew is greater than their corresponding phonemes implies that there is a greater indeterminacy in mapping
Hebrew consonant phonemes into graphemes than in converting graphemes to phonemes. The polygraphy of the consonantal
system in Hebrew is, thus, more problematic in the assembly of a word's spelling from its sound than vice versa. Although
spelling ambiguity can be reduced in some cases by orthographic constraints and by phonological or morphological knowledge,
by no means can such information eliminate the problem of polygraphic ambiguity as a whole. The severity of the ambiguity in
the spelling of consonants is particularly important because consonants in unpointed script specify almost uniquely the
assembled spelling of the word as a whole. Because each consonant in a consonantal script carries a greater amount of
information (compared to a full writing system), an incorrect spelling of the word's consonants is deleterious for the reader's
ability to recognize the word's meaning. Furthermore, the misspelling of consonants may lead to the activation of competing
irrelevant semantic information due to the prevalent homophony in the Hebrew writing system.
In summary, we have identified two problems in Hebrew orthography: the omission of vowels and the consonantal polygraphy.
The absence of vowels is considered detrimental to the assembly of phonemes from graphemes but has little influence on the
assembly of the word's graphemic representation from its phonemes. In contrast, the consonantal polygraphy of Hebrew is
relatively harmless in assembling the word's phonology from the orthography, but it poses serious problems in assembling the
word's spelling. Given the severe ambiguity in predicting a word's spelling by assembly, what is the role of assembly in
generating the spelling of Hebrew words? If assembly does contribute to the spelling process, how can Hebrew spellers avoid
constant misspellings of consonants?
A Unified Model of the Assembly of Spelling and Phonology in Processing Hebrew Words
In this section, we outline a model that provides a unified account of the role of assembled information in spelling and reading
Hebrew words. We have noted that the information specified in the assembly of both spelling

Page 202
and phonology may be partial and limited to exclusively consonantal information. In reading, this is due to the absence of most
vowels from the orthography, whereas in spelling, spellers might be unable to decide which vowels should be represented by a
vowel letter. Thus, the assembly of both spelling and phonology can result in only partial information. Yet, such information
may be nevertheless sufficient to activate the process of assembly. The central premise of our proposal is that the assembly of
the word's spelling from speech, or its phonology from print, can be triggered by impoverished input that specifies exclusively
consonantal information. Underspecified phonological information thus contributes to the recognition of Hebrew words (see also
Shimron, 1993).
The ability of spellers and readers to use such impoverished information may reflect the fact that the distinction between
consonants and vowels respects an essential linguistic distinction in the grammar of Hebrew. One of the most important
properties of Hebrew morphology is the segregation of consonant and vowel morphemes. In Hebrew, as in other Semitic
languages, all verbs and the vast majority of nouns and adjectives are formed by inserting a root morpheme, a string of
consonants containing (in most cases) three phonemes, into a preexisting word pattern that specifies vowels and, sometimes,
also additional consonants. For instance, the words /zemer/ (a song), /zimra/ (singing), and /tizmoret/ (orchestra) are formed by
inserting the root Z-M-R into one of three preexisting morphological patterns (we denote the root consonants as C): CeCeC,
CiCCa, tiCCoCet. Thus, morphologically related words always share the same consonantal root. The consonantal root also plays
an important role in accessing the mental lexicon. It interconnects the lexical entries of morphologically related words (Frost &
Bentin, 1992). The root serves as an interface that mediates the access to all morphologically related words. Supporting this
view is a recent study by Frost, Forster, and Deutsch (in press) demonstrating that the priming of a target by its root consonants
facilitates lexical decision to targets that are derivations of the root. These results suggest that access to a word's meaning is
achieved by the initial activation of consonantal information corresponding to the root.
If access to the lexicon requires the activation of an interface that is purely consonantal, then the absence of vowel information
in the representation assembled by spellers and readers may not deprive them from any information that is necessary for lexical
access. The assembly of the word's phonology from print may provide the readers with the root consonant phonemes and, thus,
with the key to accessing the lexicon via the consonantal interface. As we have noted before, the assembly of consonant
phonology is quite predictable. Hence, in spite of the absence of vowels in the script, the assembly of consonant phonemes
provides an efficient means for accessing words' meanings in the mental lexicon.

Page 203
The assembly of word spelling is more complex, however. Note that in assembling a word's spelling from speech, readers are
faced with uncertainty not only regarding the representation of vowels but also regarding the representation of polygraphic
consonants. A failure to choose the right alternative will not only result in a misspelling but, in fact, may also distort the word's
meaning because it may yield a homophone word whose meaning is quite different. If the assembly of word's spelling indeed
contributes to word recognition, then how should readers prevent spelling errors?
Our model accounts for the prevention of misspellings by postulating a graphemic inhibitory mechanism in Hebrew spelling.
This mechanism protects against spelling errors by suppressing the competing spellings of polygraphic consonants. There are
several possibilities as to the exact implementation of this inhibition. Earlier in the discussion, we described two types of
strategies governing skilled spelling: The assembly mechanism is a general, productive strategy that operates at a sublexical
level. In contrast, the addressed mechanism employs word-specific information at the lexical level. The inhibitory mechanism
could potentially be implemented at each of these levels within the spelling system: the lexical or the sublexical level.
Consider first the lexical account. As readers become familiar with a word, they may develop a top-down inhibitory connection
between the polygraphic consonant in the lexical representation and its competitor grapheme at the prelexical level (see Fig.
11.2). For instance, in the word /kelev/ (spelled klv, meaning ''a dog'') the consonant /k/ is polygraphic because it can be either
mapped to the letter kaf or kuf (in the following discussion, we notate these two letters as k and q, respectively).4 According to
the lexical account, the familiarity with /kelev/'s correct spelling results in the formation of an inhibition between the lexical
representation of /kelev/ and the competitor grapheme, q. The greater the familiarity with /kelev/, the greater the inhibition
becomes and, consequently, the more secure the spelling is. Thus, only real words that are familiar to the speller can inhibit
their graphemic competitors. Nonwords lack a lexical representation, and they cannot inhibit competitor graphemes.
There are two problems with this version of the inhibitory mechanism, however. One is that the encoding of the competition at a
lexical level does not capture the fact that such a competition reflects an important generalization inherent in the writing system.
Indeed, the competition between the graphemes k and q is not specific to the word /kelev/ but is a general problem faced by
Hebrew spellers in the spelling of all orthographic stimuli,
4 For the sake of simplicity, we disregard the polygraphy of the final consonant /v/. The consequences of this polygraphy,
however, are identical to those discussed for the initial consonant, /k/.

Page 204
Fig. 11.2.
A lexical model of polygraphic inhibition. The presentation
of the spoken word /kelev/ results in addressing the
representation of /kelev/ in the phonological lexicon,
which, in turn leads to two simultaneous patterns of
activation. One is the addressing of the orthographic
representation of /kelev/ in the orthographic lexicon. The
other is mapping the consonants of /kelev/ onto graphemes
by a sublexical assembly mechanism. Because the consonant
/k/ is polygraphic, its assembly yields the activation of two
graphemes, k and q. However, according to the lexical
model, the erroneous grapheme q is immediately suppressed
by a top-down activation from the orthographic lexicon
(inhibitory connections are notated as crossed-bold arrows).
This inhibition leaves the speller only with the correct
representation of klv. The spelling computed by the speller
(in the rectangular box) reflects the contribution of these
two sources of information.

Page 205
regardless of their lexical status. A lexical mechanism may protect the spellers from misspellings of familiar words, but it will
not be able to encode the presence of competition, let alone protect against it, for unfamiliar and new words. Obviously,
however, these are precisely the unfamiliar words that are most vulnerable to misspellings.
A more efficient way of preventing misspellings of both new and familiar words is to implement the inhibitory mechanism at
the sublexical level. The competing graphemes may be each connected to a single consonant phoneme. The activation of a
correspondence between the polygraphic consonant phoneme and one of its graphemes results in the inhibition of the connection
between that phoneme and its competitor grapheme (see Fig. 11.3). Thus, in spelling /kelev/, the assembly of the phoneme /k/
results in two simultaneous patterns of activation. One is the activation of the grapheme k and, consequently, the activation of an
inhibitory connection from /k/ to q. The other is the activation of the grapheme q, resulting in an inhibition from /k/ to the
grapheme k. Consequently, the representation assembled to the word /kelev/ lacks an explicit specification of the grapheme
corresponding to the phoneme /k/. The remaining graphemes, however, serve to address the orthographic lexicon, in which the
identity of the missing grapheme could be recovered based on word-specific knowledge of /kelev/. Thus, in spite of its partial
nature, the assembly of a word's spelling can be used by skilled Hebrew readers in retrieving the word's addressed spelling from
the lexicon without resulting in spelling errors. Note, however, that the inhibition of competitor graphemes is predicted,
regardless of the lexical status and familiarity of the orthographic string: Both familiar words and nonwords can inhibit their
competitors. Hence, the implementation of the inhibition at the sublexical level is a general mechanism designed to protect
spellers from misspelling due to the activation of assembled spelling.
Our model further assumes that the assembly of a word's spelling and phonology are automatically activated in reading and
writing Hebrew words. Indeed, it is precisely the automaticity in the assembly of a word's spelling that requires the protection of
the inhibitory mechanism. Importantly, however, if the assembly of both phonology and spelling is automatic, then the presence
of its input may be sufficient to trigger its activation. Thus, the activation between graphemes and phonemes is a two-way street
(Stone & Vanhoy, 1994): The assembly functions of spelling and phonology are expected to recurrently feed each other until the
representation is stabilized above some given threshold. This property of recurrence leads to an important prediction:
Inconsistencies in the process of mapping phonemes to graphemes may affect the conversion of graphemes to phonemes as well.
The prediction of recurrence may be illustrated using an English example. Consider again the word rain. As noted earlier, the
assembly of this word's phonology from its letters is highly predictable. In reading the word rain,

Page 206
Fig. 11.3.
A sublexical model of polygraphic inhibition. As in the lexical
account, the presentation of the spoken word /kelev/ results
in addressing the representation of /kelev/ in the phonological
lexicon, which, in turn leads to two simultaneous patterns of
activation. One is the addressing of the orthographic
representation of /kelev/ in the orthographic lexicon. The other
is mapping the consonants of /kelev/ onto graphemes by a
sublexical assembly mechanism. Because the consonant /k/ is
polygraphic, its assembly yields the activation of two
graphemes, k and q. In contrast to the lexical account, however,
note that the activation of each grapheme results in the
inhibition of its competitor (inhibitory connections are
marked by a bold-cross arrow). This inhibition occurs entirely
at the sublexical level of assembly. Because k and q inhibit each
other, the resulting output of the assembly mechanism actually
lacks a specification for the initial grapheme. However, this output
nevertheless permits addressing the orthographic lexicon and
retrieving the missing information. Thus, as in the lexical
account, the spelling computed by the speller (in the rectangular
box) reflects the contribution of both assembled and addressed
information.

Page 207
readers can thus easily map its letters onto phonemes. Once the phonemes of rain are assembled, their presence should
automatically trigger the spelling assembly mechanism. Because the spelling of rain is unpredictable, its spelling assembly is
expected to be impaired. As a result, the spelling inconsistency of rain should impair its reading, despite the fact that spelling
assembly is clearly unnecessary in a reading task. This prediction was borne in the results of Stone and colleagues (Stone &
Vanhoy, 1994; Stone et al., in press). Their findings demonstrate that lexical decision performance in English is affected by the
consistency of mapping between graphemes and phonemes in each direction. Specifically, lexical decision to words such as rain,
whose phonology is completely predictable, was impaired by the unpredictability of their spelling.
Similar effects of spelling unpredictability are expected in reading Hebrew as well. Specifically, the presence of polygraphic
consonants in print should inhibit the process of phonology assembly in reading. For instance, when presented with the word
klv, readers will assemble the initial grapheme k into the phoneme /k/. Although this mapping function is completely consistent,
the activation of the phoneme /k/ is expected to result in a recurrent activation to the competing letter q, which, in turn, will
reduce the activation of both the correct letter, k, and the phoneme /k/. Consequently, the graphemic ambiguity of the
polygraphic phoneme will impair its reading.
In summary, we propose a model of Hebrew spelling and reading that makes the following claims: (a) Spelling and reading in
Hebrew are obtained by the contribution of both assembled and stipulated information; (b) Consonantal information is sufficient
to activate the processes of assembly: The assembly of consonant phonemes in reading proceeds in the absence of vowels; word
spelling is constrained by the assembly of primarily consonant graphemes, in spite of their polygraphic ambiguity; (c) The
contribution of erroneous graphemic information is prevented by mutual inhibitory connections between a phoneme and its
competing graphemes implemented at the sublexical level; and (d) The activation between graphemes and phonemes is
bidirectional. Thus, inconsistencies in the mapping of phonemes to graphemes affects the assembly of graphemes to phonemes
as well.
Inhibitory Activation of Spelling Competitors: Evidence from Reading

In this section, we review some empirical evidence supporting the predictions of the model outlined earlier. First, if spelling
Hebrew entails a process of graphemic assembly that resolves polygraphic ambiguity by the activation of an inhibitory
mechanism, then we should expect evidence for inhibition among competitor graphemes. Second, if the inhibition operates at a
sub-

Page 208
word level, then it should be obtained regardless of the lexicality of the orthographic stimulus: Both words and nonwords should
exhibit inhibition. Third, if the mapping between graphemes and phonemes is bidirectional, then evidence for inhibition should
be obtained in each direction of the assembly process: The inhibition between competitor graphemes should emerge not only in
generating the word's spelling from sound but also in assembling its phonology from print for the purpose of reading.
Indeed, the evidence supporting our model is obtained in a reading task using the backward-masking technique. In a typical
trial, the subject is presented with a target word, klv (pronounced /kelev/), that is briefly presented in upper case for 15 ms
followed by one of several nonword masks presented in lower case: A pseudohomophone qlv, a graphemic mask, mlv, and a
control mask, rsx (see Fig. 11.4). The mask is presented for 30ms and is followed by a pattern mask. Targets and masks are both
unpointed. The
Fig. 11.4.
An illustration of the three types of masks used for the target /kelev/.

Page 209
critical comparison in this experiment concerns the effect of two nonword masks: The pseudohomophone, qlv, and the
graphemic mask, mlv. Both nonwords share the same number of graphemes in common with the word /kelev/. However, the
pseudohomophone qlv is more similar to the target on the phonological dimension. In addition, the pseudohomophone contains
the competitor grapheme q, whereas the graphemic mask, mlv, does not. How should the presentation of these two masks affect
subjects' ability to recognize the target klv?
Obviously, the introduction of any new visual input following a brief presentation of the target will interrupt its processing and
inhibit its recognition. The recognition of the target under such difficult masking conditions will only be probabilistic. However,
if the dimension distinguishing the two types of nonwords contributes to the processing of the target, then they will
differentially affect target recognition accuracy. In our example, the pseudohomophone differs from the graphemic mask on two
interrelated properties: One is its greater phonological similarity to the target, and the other is the presence of a polygraphic
consonant. These two properties have distinct effects on each of the two processes of assembly, the assembly of spelling and
phonology. The phonological similarity may facilitate the assembly of a phonological representation from graphemes. Indeed,
similar backward-masking studies conducted in English observed a facilitation in target recognition accuracy when followed by
a pseudohomophone compared to the graphemic mask (Perfetti & Bell, 1991; Perfetti, Bell, & Delaney, 1988). In contrast to
these English studies, however, the phonological similarity between the Hebrew pseudohomophones and their targets is due to
the introduction of a graphemic competitor. If Hebrew readers are equipped with an inhibitory spelling mechanism, then the
presence of such a competitor grapheme may inhibit the assembly of spelling from the word's phonology. The examination of
the combined effect of these two factors, phonological facilitation and graphemic inhibition, requires a more precise account of
the interaction between the two processes of assembly and their locus.
We have presented two accounts as to the locus of the inhibition. Consider first the predictions of the lexical account that
attributes the inhibition to a top-down activation from the lexicon. This account assumes that only real word stimuli can inhibit
their competitor graphemes. Thus, a polygraphic consonant in the target can inhibit its competitor in the mask, but the mask's
polygraph cannot inhibit the target because the mask is a non-word. In fact, target recognition under the pseudohomophone
condition should be at least as good as target recognition under the graphemic mask condition in the assembly of either spelling
or phonology. In assembling the target's spelling, the pseudohomophone should be at least as beneficial as the graphemic mask
because these masks are equated in their graphemic similarity to the target. At the level of phonology assembly, the pseudohom-

Page 210
ophone may even facilitate target recognition compared to the graphemic control due to its greater phonological similarity to the
target. Thus, the lexical account predicts that performance with the pseudohomophone mask should be at least comparable to the
graphemic control. If readers assemble a phonological representation in reading, then the pseudohomophone should facilitate
performance compared to the graphemic mask. Note however, that the pseudohomophone should never inhibit target recognition
compared to the graphemic mask.5
In contrast to the lexical account, our model predicts that the pseudohomophone should impair target recognition compared to
the graphemic mask. Because the inhibition originates at the sublexical level, its presence should be independent of the lexicality
of the orthographic input. Specifically, when presented with the pseudohomophone qlv, readers may attempt to recognize it by
converting letters to phonemes. This should result in assembling the consonant phonemes /klv/. Similarly, the graphemic mask
mlv will result in the assembly of the consonant phonemes /mlv/. The critical difference between these masks occurs at the stage
in which recurrence emerges in the system and the phonemes are mapped back onto graphemes. Consider first the effect of the
graphemic mask. After a recurrent activation in which graphemes are mapped onto phonemes and back onto graphemes, the
reader is left with mlv. Although the resulting representation differs from the desired one (klv), it does, nevertheless, provide
some evidence for the right target. Importantly, the irrelevant information provided by the graphemic mask should not prevent
the activation of the target. The state of affairs due to the presentation of the pseudohomophone qlv is quite different. The
mapping of the phoneme /k/ onto a grapheme will result in an inhibition of the correct letter kaf, and consequently, an inhibition
in target recognition. Hence, the sublexical account predicts that the pseudohomophone should inhibit target processing
compared to graphemic control, and this inhibition should be independent of word frequency.
The primary goal of the following experiments is to seek initial evidence for the presence of an inhibitory spelling mechanism in
Hebrew. A second goal of the experiments described here is to investigate the constraints affecting the emergence of the
inhibition phenomena. If inhibition in reading
5 According to the lexical account, the target cannot be inhibited by the pseudohomophone. However, this account does
predict an inhibition to the pseudohomophone by the target. This inhibition should be observed as a decrease in the
facilitory effect of the pseudohomophone. This decrease should be particularly noticeable for high-frequency words.
Because high-frequency words are more likely to inhibit their competitor graphemes, the pseudohomophones of such
targets are also more likely to be lacking a specification of the competitor grapheme, which is critical to assure their
greater similarity to the target at the stage of phonology assembly. Thus, the lexical account may predict that the
advantage of the pseudohomophone compared to the graphemic mask will be reduced for high-frequency words.

Page 211
indicates the assembly of phonology, then it should obviously arise only under conditions that permit the assembly of phonemes
from graphemes. We have pointed out that the consonantal nature of the Hebrew writing system permits the activation of only a
partial phonological code. It is possible that, in the absence of sufficient vowel information, the assembly process will be
blocked, resulting in the elimination of evidence for inhibition. However, our model assumes that the assembly of consonantal
information can proceed regardless of the presence of vowels. If this assumption is correct, then evidence for inhibition should
emerge even in words that do not contain any vowel letters. To examine the generality of the inhibition process in reading, we
compared two types of targets that differ in the amount of vowel letters present in the script. One type of targets explicitly
specifies most of the vowel information by vowel letters (explicit vowel targets), whereas the other type of targets omits most of
the vowel information (implicit vowel targets). If consonantal information is sufficient for the assembly of phonology, then
evidence for the inhibition should be independent of the presence of vowel letters. In the following section, we briefly review
the results of two empirical studies examining these questions. A complete description of these results may be found in Berent
(1993).
The Assembly of Polygraphic Phonemes Results in the Inhibition of their Graphemic Competitors
The primary goal of our first experiment is to demonstrate the existence of an inhibitory mechanism in Hebrew and determine its
locus. An additional question concerns the generality of the inhibition in respect to the orthographic transparency of the target
words and their frequency. To investigate these questions, we examined the effect of three types of masks (pseudohomophone,
graphemic, and control masks) on the recognition of targets that contrast in terms of their frequency (high vs. low) and the
presence of vowels (explicit vs. implicit). The results of this study are represented in Fig. 11.5.
According to the lexical account, performance with the pseudohomophone should be at least as good as with the graphemic
mask. In contrast, the sublexical account predicts inhibition in target recognition by the pseudohomophone. Contrary to the
predictions of the lexical account, the masking of the target by its pseudohomophone resulted in a decrease of 5.3% in its
recognition accuracy compared to its graphemic control. Furthermore, the magnitude of the inhibition caused by the
pseudohomophone was similar across the two levels of word frequency (D = 4.43%, D = 6.6%, for high- and low-frequency
words, respectively). The observation of such a general inhi-

Page 212
Fig. 11.5.
Recognition accuracy as a function of mask type in Experiment 1.
bition of target recognition by nonwords suggests that the locus of the inhibitory mechanism is in the sublexical connections
between phonemes and graphemes.
If the inhibition operates at a sublexical level of spelling assembly, then its presence may help detect the generality of
phonological assembly in reading as well. Specifically, if the conversion of graphemes to phonemes occurs only in scripts in
which the assembly of the word's phonology is highly predictable, then the inhibition should be limited to words containing
vowel information. In contrast, if consonantal information is sufficient for assembly, then the inhibition should be obtained even
when vowel information is absent. Supporting the general contribution of assembly to reading, the size of the inhibition effect
was relatively unaffected by the presence of vowels in the scripts. The pseudohomophone mask resulted in an inhibition of 4.8%
in the recognition of targets containing no vowel letters and 6% for targets in which some vowel information is present.
The results of this experiment suggest the existence of an inhibitory mechanism between phonemes and their competitor
graphemes in Hebrew. The fact that nonword orthographic stimuli can inhibit the recognition of targets containing competitor
graphemes suggests that the inhibitory mechanism reflects generalized knowledge operating at a sublexical level of phoneme-
grapheme connections rather than specific lexical knowledge with the spelling of familiar words. The generality of the inhibition
in respect to

Page 213
the orthographic transparency of the script supports the idea that the assembly of phonemes in reading does not require the
presence of vowels.
Does the Inhibition Depend on the Presence of Vowels?
The results of Experiment 1 suggest that the inhibition emerges even when the script permits the assembly of only a consonantal
representation due to the absence of vowels in the script. The possibility that the assembly of phonology can be triggered by
impoverished information containing exclusively consonantal information is intriguing. However, before accepting this
conclusion, one may wish to rule out the possibility that the failure to observe a significant attenuation in the inhibition for
words containing no vowel information may be simply due to the fact that the contrast between the two categories of targets in
our experiment is insufficient. Specifically, it is possible that explicit-vowel targets did not differ from implicit-vowel targets
because their vowel information was incomplete. For instance, the word ptish (ammer), pronounced /patish/ contains the explicit
vowel /i/. However, the pronunciation of the first vowel is unspecified in the script. It can thus also be read as /patish/, /pitish/,
/potish/, and /putish/. Hence, the failure to observe a stronger inhibition in the presence of vowels may be due to the failure to
provide sufficient vowel information in the script.
To reexamine the effect of vowels on the magnitude of the inhibition, we conducted a second experiment in which two types of
targets were compared: one in which practically all vowel information was present and one in which most vowel information
was absent. A second variable examined in this study was the length of the word. In general, one expects that the deleterious
effect of competitor graphemes should be inversely related to word length because the shorter the word, the greater the relative
contribution of each grapheme. To examine these questions, we compared four types of targets that contrasted in terms of the
presence of vowel letters (implicit vs. explicit) and length (short vs. long) factors.
The results of Experiment 2 replicated the finding of an inhibition in the recognition of target words by their graphemic
competitors. The presentation of the pseudohomophone after the target resulted in a reduction of 10.46% in recognition accuracy
compared to the graphemic control (see Fig. 11.6). However, this experiment provided stronger evidence for the modulation of
the inhibitory effect by target characteristics (see Fig. 11.7). The size of the inhibition for targets containing explicit vowel
letters (D = 12.75) was numerically greater than for targets in which vowel information is largely absent (7.8%). Thus, the size
of the inhibition appears to increase in the presence of explicit vowels. Importantly, however, the data suggest that the presence
of vowels may not be necessary for the emergence of the inhibition because the trend of inhibition was observed for both
explicit and implicit vowel targets. A second factor affecting the size of the inhibition is

Page 214
Fig. 11.6.
Recognition accuracy as a function of mask type in Experiment 2.
Fig. 11.7.
Recognition accuracy as a function of mask type for each of the vowel x
length target types.

Page 215
word length. The inhibition for short words was numerically larger than for longer words. Short words exhibited a significant
inhibition of 14.58%, whereas the inhibition observed for longer words was a nonsignificant 6.24%.
In summary, the present findings provide converging evidence that the activation of phonemes in reading Hebrew results in the
inhibition of competitor graphemes. In addition, these results suggest that the size of the inhibition may be affected by the
presence of vowels and word length. Because the inhibition for shorter words disrupts a greater proportion of the information
required for target recognition, the inhibitory effect of the pseudohomophone is greater for shorter targets. Similarly, the
presence of vowel letters seems to result in a numerical increase in the size of the inhibition effect for both short and long
vowels. However, in accord with the findings of Experiment 1, the performance for short targets in the present study suggests
that the presence of vowel letters may not be necessary for the emergence of the inhibition. Specifically, although the inhibition
observed for the long, unvoweled targets was negligible (D = 2%), short words exhibited a pronounced inhibition effect
regardless of the presence of vowels (D = 16%, D = 13%; for explicit-vowel and implicit-vowel targets, respectively). Further
research is required to examine the generality of the inhibition for purely consonantal targets that are relatively long and for
circumstances in which assembly is strongly discouraged due to the absence of any words specifying vowel information in the
experimental list.
Conclusions
The existence of polygraphic consonants in Hebrew presents spellers with a severe ambiguity in assembling the spelling of
many Hebrew words. To prevent the omission of spelling errors, spellers have developed an inhibitory mechanism whereby the
activation of a polygraphic consonant results in the suppression of its competitor grapheme via an inhibition from the common
phoneme. Our findings demonstrate the activation of such an inhibitory mechanism in the recognition of familiar words. Clearly,
such a mechanism is required only if assembled spelling contributes to the recognition of the word. Thus, the inhibitory
mechanism is prima facie evidence that skilled spellers assemble a graphemic representation in spelling familiar Hebrew words.
In addition, this mechanism provides a definition of skilled spelling in Hebrew. Although the consultation of word-specific
knowledge seems necessary to support skilled spelling in such a deep orthography, this source of knowledge is insufficient for
skilled spelling. The assembly of words' spelling appears to be an important source of constraint even in the recognition of
familiar words. Thus, skilled Hebrew spelling entails recognizing the potential sources of ambiguity in spelling Hebrew words
and automatically preventing the effect of misleading information through inhibition.

Page 216
Supporting the automaticity of spelling assembly in Hebrew is evidence for inhibition that is observed in a task that clearly does
not require the activation of a word's spelling: reading. This fact has several important implications. One concerns the role of
phonology in reading Hebrew. Because the inhibition originates at the level of phoneme-grapheme mapping, the emergence of
an inhibition necessarily indicates the contribution of an assembled phonological representation to reading. Thus, indirectly,
these results also support the role of assembled phonology in recognizing Hebrew words. Furthermore, the fact that evidence for
inhibition emerges even for targets containing no explicit vowels suggests that consonant phonemes can be assembled regardless
of the presence of vowels. Consonant phonology thus constrains reading even in an extremely deep orthography. The question of
whether assembled phonology constrains reading in deep orthographies is subject to much debate in the word recognition
literature (cf. Baluch & Besner, 1991; Frost, 1994). However, existing evidence for the contribution of assembled phonology in
the Hebrew word recognition literature comes from studies examining the effects of the presentation of vowel information on
the ease of target recognition (Bentin & Frost, 1987; Frost, 1994, 1995; Frost, Katz, & Bentin, 1987; Koriat, 1984, 1985; Navon
& Shimron, 1981). These studies generally suggest that the presence of vowel information facilitates target processing. In effect,
these findings amount to the suggestion that the reduction in the orthographic depth of the script facilitates recognition. This
conclusion, however, leaves open the question of whether assembly can take place even for an extremely impoverished script
containing consonantal information alone. Our finding that the inhibition emerges for targets lacking any vowel information
suggests that consonant information is sufficient to trigger the process of assembly (Berent & Perfetti, 1995), and that such a
process constrains the recognition of printed Hebrew words.
The sufficiency of consonantal information for assembly has some interesting implications regarding the structure of graphemic
and phonemic representations. So far we have discussed skilled spelling in Hebrew exclusively from a processing point of view.
We have identified two processing strategies for obtaining the spelling of the word and its phonology, and we have examined
their contribution to the process of spelling and reading Hebrew words. However, in addition to these processing questions, a
theory of skilled spelling and reading requires an account of the structure of the representation used by skilled spellers and
readers as well. Of particular interest is whether such a representation is subject to some universal constraints and whether some
of these constraints are linguistic in nature. Caramazza and Miceli (1990) suggested that graphemic representations distinguish
between consonants and vowels by representing them on two distinct planes. Such a representation is supported by several case
studies of Italian patients exhibiting

Page 217
specific impairment of vowel processing (Caramazza & Miceli, 1990; Cubelli, 1991). The possibility that consonants and
vowels may be two distinct entities in assembled representations is further supported by the distinction in the processing
characteristics of consonants and vowels in assembling a phonological representation of English words (Berent & Perfetti,
1995). This distinction is attributed to the presence of a universal linguistic constraint on reading that assigns consonants and
vowels into two distinct constituents, or planes, in phonological representations. Indeed, a distinction between consonant
processing and vowel processing emerges in several areas related to the processing of linguistic input, such as speech perception
and production (Fowler, 1980, 1983; Fowler & Smith, 1986), language acquisition (Macken, 1979, 1992; Studdert-Kennedy &
Whitney Goodell, 1992), spelling (Caramazza & Miceli, 1990; Cubelli, 1991), and reading (Berent & Perfetti, 1995). If
consonants and vowels are two distinct constituents in the assembly of both spelling and phonology, then one expects that, given
a language in which such a distinction is strongly supported by the morphology and is further represented in the writing system,
readers and spellers might be able to assemble the consonant constituent in the absence of the vowel constituent. This proposal
is compatible with our present results. The emergence of the inhibition for targets containing no vowel letters suggests that such
targets can support the assembly of consonantal phonology and spelling even when the target word lacks specification of the
vowels.
Finally, perhaps the most important implication of our findings concerns the relations between the assembly of phonology and
spelling. The fact that the assembly of phonology is constrained by inconsistencies inherent in the mapping of phonemes to
graphemes indicates that these two mapping functions, graphemes to phonemes and phonemes to graphemes, are not
independent. Our findings suggest that the assembly process is bidirectional: The activation of one end of the mapping functions,
graphemes to phonemes, automatically triggers the other, phonemes to graphemes (see also Stone et al., in press). Each end of
the mapping function is exposed to the effect of inconsistencies in the other. These results suggest close relations between
theories of spelling and reading: The assembly of phonemes and graphemes are two faces of a single coin. These processes are
bidirectional and highly interactive. It seems impossible to account for one of these processes independently of the other.
Acknowledgments
This research was supported by National Science Foundation Grant BNS 8811042 to Charels A. Perfetti. Correspondence should
be addressed to Iris Berent, Department of Psychology, Florida Atlantic University, 777 Glades Rd., Boca-Raton, FL 33431-
0991.

Page 218
References
Baluch, B., & Besner, D. (1991). Strategic use of lexical and nonlexical routines in visual word recognition: Evidence from oral
reading in Persian. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 644-652.
Barron, R. (1980). Visual and phonological strategies in reading and spelling. In U. Frith (Ed.), Cognitive processes in spelling
Barry, C. (1994). Spelling routes (or roots or rutes). In G. D. A. Brown & N. C. Ellis (Eds.), Handbook of Spelling: Theory,
process and intervention (pp. 27-49). Chichester, UK: Wiley.
Bentin, S., Bargai, N., & Katz, L. (1984). Orthographic and phonemic coding for lexical access: Evidence from Hebrew.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 353-368.
Bentin, S., & Frost, R. (1987). Processing lexical ambiguity and visual word recognition in a deep orthography. Memory &
Cognition, 15, 13-23.
Berent, I. (1993). The role of assembled phonology in English visual word recognition: Towards a nonlinear model of assembly.
Unpublished doctoral dissertation, University of Pittsburgh, Pittsburgh.
Berent, I., & Perfetti, C. A. (1995). A rose is a REEZ: The two cycles of phonology assembly in reading English. Psychological
Review, 102, 146-184.
Bosman, A. M. T. (1994). Reading and spelling in children and adults: Evidence for a single-route model. Amsterdam:
Universiteit van Amsterdam.
Cubelli, R. (1991). A selective deficit for writing vowels in acquired dysgraphia. Nature, 353, 258-260.
Fowler, C. A. (1980). Coarticulation and theories of extrinsic timing. Journal of Phonetics, 8, 113-133.
Fowler, C. A. (1983). Converging sources of evidence on spoken and perceived rhythms of speech: Cyclic production of vowels
in monosyllabic stress feet. Journal of Experimental Psychology: General, 112, 386-412.
Fowler, C. A., & Smith, M. R. (1986). Speech perception as ''vector analysis'': An approach to the problems of invariance and
segmentation. In J. S. Perkell & H. Klatt (Eds.), Invariance and variability in speech processes (pp. 123-139). Hillsdale, NJ:
Academic Press.
Frost, R. (1994). Prelexical and postlexcial strategies in reading: Evidence from a deep and a shallow orthography. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 20, 116-129.
Frost, R. (1995). Towards a strong phonological theory of reading: True issues and false trails. Manuscript submitted for
publication.
Frost, R. (1995). Phonological computation and missing vowels: Mapping lexical involvement in reading. Journal of
Frost, R., & Bentin, S. (1992). Reading consonants and guessing vowels: Visual word recognition in Hebrew orthography. In R.
Frost & L. Katz (Eds.), Orthography, Phonology, Morphology, and Meaning. North-Holland, England: Elsevier.
Frost, Forster, & Deutsch (in press). What we can learn from the morphology of Hebrew: A masked priming investigation of
morphological representation. Journal of Experimental Psychology: Learning Memory and Cognition.
Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and orthographical depth: A multilingual
comparison. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13, 104-115.

Page 219
Jorm, A. F. (1983). The psychology of reading and spelling disabilities. London: Routledge & Kegan Paul.
Katz, L., & Frost, R. (1992). The reading process is different for different orthographies: The orthographic depth hypothesis.
Orthography, Phonology and Meaning (pp. 67-83). Amsterdam: North-Holland.
Koriat, A. (1984). Reading without vowels: Lexical access in Hebrew. In H. Bouma & G. Bouwhuis (Eds.), Attention and
performance: Control of language processes (pp. 227-242). Hillsdale, NJ: Lawrence Erlbaum Associates.
Koriat, A. (1985). Lexical access for low- and high-frequency words in Hebrew. Memory & Cognition, 13, 37-44.
Kreiner, D. S. (1992). Reaction time measures of spelling: Testing a two-strategy model of skilled spelling. Journal of
Link, K., & Caramazza, A. (1994). Orthographic structure and the spelling process: A comparison of different codes. In G. D.
A. Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (pp. 261-294). Chichester, UK: Wiley.
Macken, M. (1979). Developmental reorganization of phonology: A hierarchy of basic units of acquisition. Lingua, 49, 11-49.
Macken, M. (1992). Where's phonology? In C. Ferguson, L. Menn, & C. Stoel-Gammon (Eds.), Phonological development:
Models, research, implications. Timonium, MD: York Press.
Morag, S. (1962). The vocalization system of Arabic, Hebrew, and Aramaic: Their phonetic and phonemic principles. Hague:
Mouton.
Navon, D., & Shimron, J. (1981). Does word naming involve grapheme-to-phoneme translation? Evidence from Hebrew.
Journal of Verbal Learning and Verbal Behavior, 20, 97-109.
Perfetti, C. A., & Bell, L. (1991). Phonemic activation during the first 40 ms of word identification: Evidence from backward
masking and priming. Journal of Memory and Language, 30, 473-485.
Perfetti, C. A., Bell, L. C., & Delaney, S. M. (1988). Automatic (prelexical) phonetic activation in silent reading: Evidence from
backward masking. Journal of Memory and Language, 32, 57-68.
Read, C. (1971). Preschool children's knowledge of English phonology. Harvard Education Review, 41, 1-34.
Shimron, J. (1993). The role of vowels in reading: A review of studies of English and Hebrew. Psychological Bulletin, 114, 52-
67.
Stone, G., & Vanhoy, M. (1994, November). Top-down and bottom-down phonology in visual word recognition. Paper
presented at the 35th Annual Meeting of the Psychonomic Society, St. Louis, MO.
Stone, G., Vanhoy, M., & Van Orden, G. (in press). Perception is a two-way street: Feedforward and feedback phonology in
visual word recognition.
Studdert-Kennedy, & Whitney Goodell (1992). Gestures, features and segments in early child speech. In Haskins Laboratories
Status Report on Speech Research (pp. 89-102). New Haven, CT: Haskins Laboratories.
Treiman, R. (1994). Sources of information used by beginning spellers. In G. D. A. Brown & N. C. Ellis (Eds.), Handbook of
spelling: Theory, process, and intervention (pp. 75-91). Chichester, UK: Wiley.
Van Orden, G. C., Pennington, B. F., & Stone, G. O. (1990). Word identification in reading and the promise of subsymbolic
psycholinguistics. Psychological Review, 97, 488-522.
Venezky, R. L. (1970). The structure of English orthography. The Hague: Mouton.


Page 221
Chapter 12
Children's Use of Analogy in Learning to Read and to Spell
Jean Emile Gombert
Universit de Bourgogne
Peter Bryant
Nicola Warrick
The Orthographic Neighborhood Effect in Reading

Knowledge that the reader already has about the pronunciation of words may influence the way written words (or pronounceable
pseudowords) are pronounced when encountered for the first time during the reading process. It is clear that adults and children
often employ analogies that are based on orthographic neighborhoods when trying to decipher new words.
Glushko (1979) found that American adults, when reading pseudowords, did not always follow the grapheme-phoneme
conversion rules; they sometimes used analogies to the pronunciation of irregular words (see Baron, 1977; Brown & Besner,
1987; Masterson, 1985). The same results were obtained in French by Peereman (1991) and in Spanish by Sebastin-Galls
(1991). Taken together, these results suggest that, contrary to what was claimed in some theories (Bentin, Bargai, & Katz, 1984;
Bentin & Frost, 1987; Feldman & Turvey, 1983; Frost, Katz, & Bentin, 1987; Katz & Feldman, 1983), the reading mechanisms
for languages with shallow and deep orthographies are not different.
In fact, the use of analogies appears early during the process of learning to read. In a cross-linguistic study, Goswami, Gombert,
and Fraca de Barrera

Page 222
(in press) compiled two lists of monosyllabic pseudowords and two lists of bisyllabic pseudowords in English, French, and
Spanish. In each language, one list was composed of pseudowords that had many real word neighbors when the first letter was
changed (for instance, in English, voop or ketter) and one list was composed of pseudowords with few orthographic neighbors
(joog or terket).
The participants were children, from Grades 2, 3, and 4, and adults. They had to read the lists aloud quickly and without errors.
Whatever the language, the age and the item length, the pseudowords that had several real words as neighbors were read faster
and with fewer errors than the others. This effect was stronger in English than in French and stronger in French than in Spanish.
This order of difficulty is probably a consequence of the fact that the more irregular the orthography, the more analogies help
the reader. On the whole, this effect is stable across the ages in all three languages. Thus, the use of analogy in reading already
seems to be effective by the second year of learning to read. It is not enough to establish the early use of analogy. It is also
necessary to specify how this employment of lexical knowledge works. In fact, there are two possibilities. The first is that the
reader uses links to long-term memory between certain spelling patterns and the pronunciation of these patterns. The second is
the simple use of some phonological patterns in memory for known words, which facilitates the pronunciation of all items,
including these phonological patterns, whatever the reason for this pronunciation, and does not take into account the
orthography of the written counterparts of these items.
Let us take as an example the ease with which English readers can read the pseudoword ketter. According to the first
hypothesis, they manage to do this because they are already familiar with the letter sequence etter and the sound that it usually
represents because they have seen it frequently in the words better, letter, and wetter. According to the second hypothesis, they
read the word ketter quite easily because they are familiar with pronouncing and hearing the sound [etor *]: It is this familiarity
rather than the familiarity of the letter pattern that helps them to read the pseudoword. To decide between these two possibilities
(which are not incompatible), Goswami, Gombert, and Fraca de Barrera (in press) used lists of pseudowords that had numerous
phonological neighbors (by changing the first phoneme) but few orthographic neighbors. In fact, these items were pseudowords
that were homophones to orthographic neighbors of real words (for instance, real word = coffee; orthographic neighbor O + P+
= loffee; phonological neighbor O - P+ = loffi).
If the first hypothesis is true (orthographic neighborhood effect), orthographic neighbors (in fact orthographic and phonological
neighbors O + P+) are read more easily than phonological-only neighbors (O - P+). This was very much the case in English:
Whatever the age, the O + P + items were read faster

Page 223
(1.14 sec. vs. 1.41 sec.) and more accurately (correct reading = 80% vs. 68%) than the O - P+ items. In French, this
phenomenon also occurred whatever the age, but the effect was not as strong as in English (reading times: 1.00 sec. vs. 1.07
sec.; correct reading = 92% vs. 87%). As for Spanish, the comparison was impossible. In a language with a shallow
orthography, it is difficult to find different spelling patterns for the same pronunciation. The second hypothesis implies that
phonological-only neighbor items (O - P+) are read more easily than pseudowords with few neighbors (either orthographic or
phonological neighbors: O - P-). This second hypothesis was also clearly validated in French (reading times: 1.18 sec. vs. 1.41
sec.; correct reading = 83% vs. 73%). In fact, it seems that in French, there is a simultaneous use of orthographic analogy in
reading aloud and a facilitation of pronunciation through the use of a phonological code when items to be pronounced share
large phonological patterns with known words.
In English, there were the same kinds of results. Whereas the orthographic neighborhood effect was stronger in English than in
French, the effect of the simple phonological neighborhood was smaller. As a matter of fact, in the English group, this effect
was statistically significant only in children and only for the number of accurate readings (for children, whatever the age: 51%
vs. 45%) but not for reading times (1.77 sec. vs. 1.81 sec.). In fact, it seems that, in English more than in French, the
phonological-only neighbor pseudowords are, orthographically speaking, bizarre items. This oddity could be the source of a
negative effect that cancels the positive effect of having a phonological output code for pronouncing these items.
Analogical Processes in Learning to Read

Whatever form the simple phonological neighborhood effect takes, this series of experiments established that, as early as the
second year of practice, the reader takes into account the orthographic neighborhood of the items met during reading. So, we
must determine more precisely when this phenomenon occurs in the learning process. There are two opposing views.
For some researchers, the phenomenon comes late; it occurs only after the beginning reader has already mastered grapheme-
phoneme correspondences. For others, it occurs precociously and might allow beginning readers to read words that they have
never met before, when they are not yet able to decode these words.
Most of the developmental models that treat the acquisition of reading as a sequence of stages agree with the first view. In one
of the most famous, Frith (1985) postulated a development in three successive stages. In the first stage, the logographic stage,
the absolute beginner develops strategies in order to "guess" the oral words corresponding to the visual patterns that he or she
perceives but does not yet know how to read. These strategies consist of

Page 224
taking all the clues that, in the given context, can help with the identification of the word. These could be the advertising logo
that is characteristic of certain words (for instance, Coca Cola), or they can also be some salient visual clues taken from the
word itself, for instance, some particular features (such as the point above i), and some letters or patterns of letters. At the
extreme, these known letter patterns taken as a clue can include the whole word that is recognized "by heart," globally.
Nevertheless, even in such a case, its component letters can't be used in order to read other words.
This pseudoreading, in which the beginner uses all the available clues in order to guess the word, would permit the recognition
of a limited number of words (probably less than 100), not by direct access to the internal lexicon but by using a pictorial
semantic system (Morton, 1990). In this system, first the concept corresponding to the written word is accessed without any
linguistic processing, and then the spoken word corresponding to the concept is activated. Written linguistic information is
processed as nonlinguistic material, and the relations between written configuration and meaning do not depend on general
linguistic rules.
In the second stage, the alphabetic stage, the main phenomenon is the use of phonological mediation. At this level of learning,
the attentional effort is principally devoted to matching written and spoken linguistic forms. This matching requires the use of
the rules of grapheme-phoneme correspondence (GPC) that depends on knowledge of the alphabet, a metaphonological
awareness of the phonemic unit, and a linkage between these two types of knowledge.
Last, in the third stage, the orthographic stage, which will become the stage of automated reading expertise, words are analyzed
in orthographic units without the systematic use of phonological conversion, and this development allows the reading of
irregular words.
From this point of view, the use of analogies does not appear before the alphabetic stage and constitutes a sophistication of the
matching between writing and oral language by using orthographic units larger than the phoneme, in which the reader has noted
the regularity.
This analysis does not fit with some experimental data that have shown a more precocious use of analogies in reading. In fact, it
is likely that the process of using analogies between words in reading has to be considered as a part of beginning reading and not
simply as a special case of alphabetic processes. The hypothesis is that, in order to read words that they have never seen before
in written form, beginning readers use their knowledge of orthographic patterns in words that they can already read. For
instance, a child who can read the word light might read the words night or sight by using, on the one hand, the orthographic
analogy between these words and, on the other hand, his or her own ability to recognize the rime /ait/ in these three words. The
same phenomenon works for clusters of consonants at

Page 225
the beginning of words (the onset) but less (or only later) for units other than the onset or the rime.
Without special training, young children seem unable to break a word into its constituent phonemes, but they are quite good at
producing (Dowker, 1989) and detecting (Bryant, Bradley, MacLean, & Crossland, 1989) rhymes and alliteration. That means
that they can judge that words and syllables, which are clearly different from each other, but, nevertheless, have phonological
segments in common. These common segments in rhyming tasks are usually larger than the phoneme in that they typically
contain more than one phoneme. For example, the rhyming sound in cat and hat is /at/, which is two phonemes. The ability to
detect rhyme is probably a great deal less refined than the ability to detect phonemes.
The developmental relation between judgments about rhyme, alliteration, and phonemes and the acquisition of reading is an
intriguing one. One of the best measures of rhyme detection and of phoneme detection is an oddity task. Children hear three or
four words at a time and are told that all but one rhyme, or that all but one begin with the same phoneme or end with the same
phoneme. It is not surprising that young children can manage the rhyming oddity task quite well long before they learn to read.
It is rather more interesting to note (Kirtley, Bryant, MacLean, & Bradley, 1989) that 5-year-old nonreaders also do quite well
(and well above chance level) when asked to detect the word that begins with a different phoneme from the rest (an alliteration
task). The difficult task for them is to judge which word ends with the same phoneme. Five-year-old children who have not yet
learned to read are at chance level in this task, whereas those who have learned to read do better than this.
Even prereaders detect rhyme and alliteration quite well. Sometimes this even means that they rise above chance level in
phoneme detection tasks, but other phoneme detection tasks flummox them, and they are plainly unable to fully analyze the
phonemes in most words. This leaves us with the question of whether these early rhyme and alliteration skills have anything to
do with reading.
Given as a postulate the idea that the important phonological link in reading is the grapheme-phoneme correspondence, it is
quite reasonable to suggest that rhyme at any rate should be quite trivial as far as reading is concerned. Rhyme is too "gross,"
too "global," as Content, Morais, Alegria, and Bertelson (1982) put it. However, there is another possibility, which is that
children do not learn just about relations between single letters and phonemes. For example, the letter sequence -ight represents,
in a relatively invariant way, a sound that contains more than one phoneme. What is to stop the child learning an association at
this level?
The answer, it seems, is "nothing at all." This certainly is the conclusion to be drawn from some interesting experiments by
Goswami (1986), which

Page 226
showed that beginning readers are often quite eager to associate sounds with letter sequences of this sort. Goswami gave 6-, 7-,
and 8-year-old children a set of words to read that were mostly too difficult for them. Then she told them what one of these
written words (e.g., beak) meant. The result of her telling them how to read this word was that it helped them to read other
rhyming words with similar final spelling patterns, such as peak, leak, and weak, but it had little effect on their ability to read
other words such as bean or bake, which did not share the same final spelling sequence or rhyming sound. Later, Goswami
(1988a) repeated this striking finding with a task in which children had to read passages of prose.
Bruck and Treiman (1992) showed that training to use rime analogy (e.g., pig, big) is more efficient than training to use other
analogies (e.g., pig, pim) in order to help beginning readers to read new words that were presented just after the training session.
If children are ready right at the start of learning to read to infer that two words that end with the same spelling sequence also
share a rhyming sound, it is quite likely that their ability to group words by rhyme should help them to learn to read and to spell.
In fact, there is ample evidence for this connection. Children's early (preschool) sensitivity to rhyme and alliteration is an
extremely powerful predictor of their eventual success in learning to read. It is also the case that training in rhyme helps children
learn to read (Bradley & Bryant, 1983).
These data are important for the question of knowing if the link between phonological skills and reading is continuous or
discontinuous (cf. Bryant, 1993). The first possibility is that the development here is a continuous one (Bryant & Goswami,
1987; Goswami & Bryant, 1990). Children acquire certain phonological skills and presumably become more skilled
phonologically throughout the preschool period, and these early phonological skills form the basis of their learning about letter-
sound relations when they begin to read. The other possibility (Content et al., 1982) is that the phonological routines that they
have to learn about in reading are at a different level and take a completely different form to any phonological analyses that they
have carried out until then. According to this second argument, children acquire new phonological skills, which otherwise would
not be a part of their cognitive repertoire, as a direct result of learning to read.
The view on this continuity/discontinuity depends on the phonological unit that you think is important. If you hold the view that
the crucial unit is the phoneme and the crucial link the grapheme-phoneme correspondence, then you will probably hold to a
discontinuity position. If, on the other hand, you agree that children probably also learn that words that rhyme also have
sequences of letters in common, the idea of continuity is bound to seem the more plausible one. In fact, the evidence for the
importance of rhyme and for its links with later success in reading seems so

Page 227
compelling that some version of the continuity position seems inescapable. Children's sensitivity to rhyme certainly feeds into
the process of learning to read, but the experience of reading probably leads to the acquisition of entirely new skills. Explicit
attention to phonemes is one.
What seems to be well established in this research is the precocious role of analogical processes. The fact that onsets and rimes
play a role of special importance in these analogies still has to be proved. In fact, it is possible that this phenomenon is peculiar
to the English language, where the pronunciation of vowels depends on the subsequent consonants and where the spelling
patterns corresponding to the rimes are often so complicated that it is cognitively less costly to encode them as a block than to
analyze them. Moreover, even in English, Goswami (1991b) failed to reproduce the advantage for the rime in precocious
analogies when children had to read words included in texts rather than isolated words. In the same way, Bruck and Treiman
(1992) showed that the primacy of the rime exists only during a short period. As soon as the day after, training with analogies
other than rimes produces better results as a help in reading new words (see Wise, Olson, & Treiman, 1990). This result does
not invalidate the hypothesis that onset-rime analysis is the first type of syllable analysis by prereaders. On the contrary, the
authors suggest that it is because this type of analysis is very easy for children that they are less likely to remember the training
than when it involved the use of analogies involving other intrasyllabic units, the identification of which was far more difficult
due to its more explicit nature.
Muter, Snowling, and Taylor (1994) showed that, although 6-year-old children, trained on a series of clue words, read more
words that shared spelling patterns with the clue words than control words, this effect was reduced when the clue word was not
exposed during the presentation of the targets.
These data invite a new description of the process of learning to read and of the role of analogies in this learning process (see
Fig. 12.1). At first, beginners who have already met written words in their prelearning activities are able to take advantage of
two forms of knowledge: (a) visual knowledge that allows them to guess a certain number of written words that they have met
repeatedly before; and (b) phonological knowledge gained from the manipulation of oral language, knowledge that notably
includes a sensitivity to the composite nature of syllables, as shown by the precocious sensitivity of the children to rimes and, to
a lesser extent, alliterations.
As with any set of words, the words that are identified (guessed) by a visual procedure include a certain number of orthographic
regularities, which can be noticed by the visual information processing system. These orthographic regularities are, themselves,
in systematic co-occurrence with phonological regularities in the pronunciation of these words. It is this correspondence
between orthographic analogy and phonological analogy that is used by the cognitive system in a self-learning mechanism in
order

Page 228
Fig. 12.1.
The learning to read process.
to allow the reading of some new words that are phonological neighbors of already known words.
At first, there is no reason that the reader must have any awareness of the analogies made between spelling patterns and between
spoken words. In fact, in a recent study, Seymour and Evans (1994) showed that it is later on that the child becomes able to
perform conscious, intrasyllabic analysis. Complementary evidence comes from Cardoso-Martins (1994) who, in comparing
performance on different rhyme categorization tasks in kindergartners and first graders, provided results that suggest that rhyme
detection does not involve attention to segments. She concludes that this kind of

Page 229
intrasyllabic manipulation may best be described in terms of sensitivity to some sort of global phonological similarity.
Nevertheless, the repeated use of analogies in beginning reading implies an increase in the saliency of the orthographic and
phonological configurations they involve and, thereby, improves their availability for conscious access.
The mastery of the alphabetic code has a double origin. First, it is the issue of a progressive refinement of the analogies the
child makes until taking into account the minimal orthographic unit (cf. Goswami, 1993). Second, it is a consequence of
instruction that helps and, for many children, allows this refinement.
The mastery of the alphabetic code implies a conscious awareness of the units in the code. This way of describing reading
acquisition is very different from previous descriptions. Here, the prereading processing, that we call pictorial and that is usually
referred to as logographic (an ambiguous term in our sense), is not optional. It is necessary for the subsequent building of
orthographic knowledge. This orthographic knowledge is the product of the analyses that children apply to the words that they
are able to identify visually, especially given the way that their phonological knowledge of these words is organized in memory.
Thus, such a visual vocabulary is indispensable for allowing entry into reading.
This pictorial processing is destined to disappear. As a matter of fact, learning to read leads to the establishment of analogical,
alphabetic, and orthographic processes that are automatically activated when the skilled reader meets a written word. These
reading processes are the product of the transformation of the prereading processing that no longer exists in competent readers.
It is impossible to recognize stages in this evolution. In fact, the driving force of this evolution is the confrontation with written
words. The consequence is that the evolution is dependent on the frequency of encountering different words. The dcalage
(split) in functioning levels (succession: prereading, analogical reading, alphabetic reading, and orthographic reading, with the
latter three being combined) as a function of the words to be read is the rule during the first months (or perhaps years) of
learning to read.
The view presented here provides a way of explaining interindividual differences. It is enough to consider the possibility of
interindividual differences in phonological or visual skills to explain why one reader develops some processing skills more
rapidly than another.
Analogical Processes in Learning to Spell

The model proposed by Frith (1985) is not just a model of reading development but a model of reading and writing
development. In this model, there is both a parallelism and an evolving functional difference between reading

Page 230
and spelling. The logographic stage occurs with reading before writing, but it lasts longer, so reading is still logographic when
writing is already alphabetic. The children's experience of logographic reading eventually helps them to use the same code more
effectively when they write. In contrast, the nature of writing where letters are written one by one, leads children to use the
alphabetic code, and their experience of spelling words alphabetically has the effect of showing them how to read alphabetically
as well. The alphabetic stage occurs in spelling before reading, but it lasts longer. Spelling is still alphabetic when reading has
become orthographic. Children use the orthographic code first when they read and later when they spell.
Goswami and Bryant (1990) had two concerns about this model. First, they pointed out that there is no evidence that children
spell logographically before they spell alphabetically. Second, they thought the model to be unsatisfactory because it is vague
about the effects of skills, particularly phonological skills, which children acquire before they learn to read and spell. Of course,
one question arises about the role of analogies in spelling acquisition. Goswami and Bryant argued that analogies must be
important in spelling because they are based on phonology, and phonology plays a stronger role in learners' spelling than in their
reading. From this logical argument, these authors derived the hypothesis that, when reading unfamiliar words, children also
make analogies about phonological units when they try to spell a word for the first time.
A study by Campbell (1985) seemed to contradict this hypothesis. Her study used a priming paradigm with children from 9- to
12-years-old. Each child had to write only the pseudowords included in a dictated list in which there were real words and
pseudowords. The primes varied for the participants. For half of the children, each target pseudoword was preceded by a certain
rhyming real word (e.g., [boup] was preceded by soap); for the remaining half, the same pseudowords were preceded by other
real words (e.g., [boup] was preceded by rope). Only the participants who had a reading age of 11 years or more consistently
spelled the pseudowords using the spelling pattern of the prime, confirming that they used similarities of sound to make
analogies about spelling. However, the fact that the younger participants in Campbell's study used analogies infrequently when
faced with pseudowords does not imply that they were totally unable to use analogies in the normal course of spelling.
Goswami (1988b) presented 6-year-olds some written words (e.g., beak) and pronounced those words. Then the children had to
spell other words. Some of these other words sounded similar to the clue word (e.g., peak or bean); others did not (e.g., lake or
bask). The children spelled the former (the analogy words) better than the latter (control words). Moreover, the children
correctly spelled the analogy words that shared a rime with the

Page 231
clue word (beak/peak) more often than the words that shared the initial consonant-vowel segment (beak/bean). Thus, in spelling
as well as in reading, children seem able to use analogies; analogies based on rime seem easier than those not based on the
onset/rime structure of the syllable.
Nation and Hulme (1994) had slightly different results. They also showed that 6-year-old children can make analogies between
a visible clue word and a similar sounding target when spelling. However, in their data, the use of analogy was not any stronger
with words sharing rimes than with words sharing other phonological segments. There was the same rate of analogy use when
the clue and the target shared a rime, an initial CV-segment, or a vowel. Nevertheless, analogies were not made when the clue
and the target shared only common letters. The same pattern of results was obtained when the clue word was not visible (i.e.,
using a priming experimental paradigm). Thus, as with beginning readers, it seems that beginning writers can make analogies in
spelling, but it is not clearly established that the rime unit plays a special role for such analogies.
Moreover, apart from the cases in which children are prereaders, in the experiments on the use of analogies in reading and
spelling, the level of learning is generally defined in a very rough way. All we know is that, for example, the beginners are in
their first year of learning to read and to spell. This lack of precision might obscure some developmental changes during that
first year of learning. Sprenger-Charolles and Casalis (1996) tested the use of analogies in reading and spelling by French
beginning readers 4 and 9 months after starting to learn to read and spell. The results revealed that analogies are used as early as
the first session in reading but only at the second session in spelling. This difference between reading and spelling was also
found in Coenen's (1994) results concerning children learning to read and to spell Dutch (a language with a rather transparent
orthography) that showed that word-specific knowledge is used in reading before spelling. On the whole, these recent results do
not contradict Goswami's (1988b) claim that children use their reading knowledge in order to decide if one analogy is
appropriate in spelling. The directionality of this link implies that analogy processes occur first in reading then in spelling.
This difference between reading and spelling is not surprising. As a matter of fact, there is a big difference between reading and
spelling tasks as far as the memory is concerned. Reading a word implies recognizing the written patterns; spelling a word
implies recalling the orthographic sequence. It is likely that, in recognition, a global apperception is cognitively less costly than
an analysis of the pattern. In contrast, the quality of recall necessary to be able to reproduce what is recalled implies a precise
identification of each successive letter. In such a context, an analytical treatment of the spoken word in an alphabetic manner and
an activation of the successive letters to be written is,

Page 232
cognitively speaking, more economical than the activation of a complicated configuration that has to be maintained in working
memory until its graphic transcription is completed (a long process in beginners). This processing difference might explain why,
usually (i.e., without special teaching), the ability to segment by phonemes is only a consequence of learning to read but is
necessary for spelling (Goswami & Bryant, 1990).
Conclusion
Almost everybody agrees with the idea that analogies are used at some level of learning to read and to spell. The controversy
concerns the timing of this use. In fact, this debate is part of a more general debate on analogical reasoning in children. Whereas
some people taking a Piagetian perspective suggest that this kind of reasoning is late developing (cf. Inhelder & Piaget, 1958),
others think that it appears early (cf. Brown, 1989). Goswami (1991a) resolved this contradiction by proposing an analogical
development that allows early analogical competence and that also postulates the later development of metalogical skills.
Learning to read and to spell is the acquisition of knowledge in a formal domain, and, as in other acquisitions of this sort, it is
likely that analogy plays a central role (cf. Halford, 1992; Suzuki, 1994). The written environment, as with other learning
environments, is a very complex one, and so it is difficult, and perhaps impossible, for the beginner to fully understand the
relevance of the rules the teacher tries to explain. Analogical processes permit the beginner to go beyond his or her own explicit
knowledge in apprehending written material. This kind of functioning is efficient first in reading; because of memory load it
appears in spelling only later when the child is able to produce each graphic unit to be used in the spelling pattern to be written.
The existence of early analogical processes in learning to read does not contradict the idea that other analogical processes can
appear later as a consequence of alphabetic reading experience. The development of orthographic knowledge through alphabetic
reading, and above all spelling, adds to the basis of knowledge available for making analogies. Therefore, it is not astonishing
that analogy use seems to be more frequent in skilled readers than in beginners. The important point is to recognize that the
early occurrence of analogies in learning to read and to spell provides a way of explaining how an information processing
system, which can neither read nor activate written concepts to be graphically reproduced (as that system exists in prereaders),
can be transformed into a reading and spelling machine.
This view implies the affirmation of the universal character of this way of learning to read and to spell. As shown by the results
of Goswami et al.

Page 233
(in press), and as suggested by those of Wimmer and Goswami (1994) and by the analysis of Goswami (1993), what is different,
depending on the degree of transparency of the orthographies, is more the visibility of the analogical processing than its
existence. If you imagine a written system in which every word is a consonant-vowel monosyllabic word, there is no visible
difference between analogical and alphabetic reading. Nevertheless, these two ways of reading remain, psychologically
speaking, different.
References
Baron, J. (1977). What we might know about orthographic rules. In S. Dornic (Ed.), Attention and performance VI (pp. 557-
572). London: Academic Press.
Bentin, S., Bargai, N., & Katz, L. (1984). Orthographic and phonemic coding for lexical access: Evidence from Hebrew.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 353-368.
Bentin, S., & Frost, R. (1987). Processing lexical ambiguity and visual word recognition in a deep orthography. Memory and
Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read: A causal connection. Nature, 301, 419-421.
Brown, A. L. (1989). Analogical learning and transfer: What develops? In S. Vosniadou & A. Ortony (Eds.), Similarity and
analogical reasoning (pp. 369-412). Cambridge, UK: Cambridge University Press.
Brown, P., & Besner, D. (1987). The assembly of phonology in oral reading: A new model. In M. Coltheart (Ed.), Attention and
performance XII: The psychology of reading (pp. 471-489). Hillsdale, NJ: Lawrence Erlbaum Associates.
Bruck, M., & Treiman, R. (1992). Learning to pronounce words: The limitations of analogies. Reading Research Quarterly, 27,
375-388.
Bryant, P. E. (1993). Reading and development. In C. Pratt & A. F. Garton (Eds.), Systems of representation in children:
Development and use (pp. 235-250). New York: Wiley.
Bryant, P. E., Bradley, L. L., MacLean, M., & Crossland, J. (1989). Nursery rhyme, phonological skills and reading. Journal of
Child Language 16, 407-428.
Bryant, P. E., & Goswami, U. (1987). Beyond grapheme-phoneme correspondence. Cahier de Psychologie Cognitive, 7, 439-
443.
Campbell, R. (1985). When children write nonwords to dictation. Journal of Experimental Child Psychology, 40, 133-151.
Cardoso-Martins, C. (1994). Rhyme perception: Global or analytical. Journal of Experimental Child Psychology, 57, 26-41.
Coenen, M. (1994, October). Reading and spelling in Dutch first graders and second graders. Paper presented at the NATO
ASI Conference, Alvor/Algarve, Portugal.
Content, A., Morais, J., Alegria, J., & Bertelson, P. (1982). Accelerating the development of phonetic segmentation skills in
kindergartners. Cahier de Psychologie Cognitive, 2, 259-269.
Dowker, A. (1989). Rhymes and alliterations in poems elicited from young children. Journal of Child Language, 16, 181-202.
Feldman, L. B., & Turvey, M. T. (1983). Word recognition in Serbo-Croatian is phonologically analytic. Journal of
Experimental Psychology: Human Perception and Performance, 9, 288-298.
Surface dyslexia: Cognitive and neuropsychological studies of phonological reading. Hillsdale, NJ: Lawrence Erlbaum
Associates.

Page 234
Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and orthographic depth: A multilingual
comparison. Journal of Experimental Psychology: Human Perception and Performance, 13, 104-115.
Glushko, R. J. (1979). The organization and activation of orthographic knowledge in reading aloud. Journal of Experimental
Psychology: Human Perception and Performance, 5, 674-691.
Goswami, U. C. (1986). Children's use of analogy in learning to read: A developmental study. Journal of Experimental Child
Goswami, U. C. (1988a). Orthographic analogies and reading development. Quarterly Journal of Experimental Psychology, 40A,
239-268.
Goswami, U.C. (1988b). Children's use of analogy in learning to spell. British Journal of Developmental Psychology, 6, 21-33.
Goswami, U.C. (1991a). Analogical reasoning: What develops? A review of research and theory. Child Development, 62, 1-22.
Goswami, U.C. (1991b). Learning about spelling sequences: The role of onset and rimes in analogies in reading. Child
Development, 62, 1110-1123.
Goswami, U.C. (1993). Towards an interactive analogy model of reading development: Decoding vowel graphemes in beginning
reading. Journal of Experimental Child Psychology, 56, 443-475.
Goswami, U.C., & Bryant, P. (1990). Phonological skills and learning to read. Hillsdale, NJ: Lawrence Erlbaum Associates.
Goswami, U.C., Gombert, J. E., Fraca de Barrera, L. (in press). Children's orthographic representations and linguistic
transparency: Nonsense word reading in English, French and Spanish. Applied Psycholinguistics.
Halford, G.S. (1992). Analogical reasoning and conceptual complexity in cognitive development. Human Development, 35, 193-
217.
Inhelder, B., & Piaget, J. (1958). The growth of the logical thinking from children to adolescence. New York: Basic Books.
Katz, L., & Feldman, L. B. (1983). Relation between pronunciation and recognition of printed words in deep and shallow
orthographies. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 157-166.
Kirtley, C., Bryant, P., MacLean, M., & Bradley, L. (1989). Rhyme, rime and the onset of reading. Journal of Experimental
Masterson, J. (1985). On how we read nonwords: Data from different populations. In K. E. Patterson, J. C. Marshall, & M.
Coltheart (Eds.), Surface dyslexia: Neuropsychological and cognitive studies of phonological reading (pp. 289-299). Hillsdale,
NJ: Lawrence Erlbaum Associates.
neurons. Cambridge, MA: MIT Press.
Muter, V., Snowling, M., & Taylor, S. (1994). Orthographic analogies and phonological awareness: Their role and significance
in early reading development. Journal of Child Psychology and Psychiatry, 35, 293-310.
Nation, K., & Hulme, C. (1994, October). The role of analogies in young children's spelling. Paper presented at the NATO ASI
Conference, Alvor/Algarve, Portugal.
Peereman, R. (1991). Phonological assembly in reading: Lexical contribution leads to violation of graphophonological rules.
Memory and Cognition, 19, 568-578.
Sebastin-Galls, N. (1991). Reading by analogy in a shallow orthography. Journal of Experimental Psychology: Human
Perception and Performance, 17, 471-477.
Seymour, P. H. K., & Evans, H. (1994). Levels of phonological awareness and learning to read. Reading and Writing, 6, 221-
250.
Sprenger-Charolles, L., & Casalis, S. (1996). Lire: Lecture et ecriture [Reading: Lecture and writing]. Paris: P.U.F.
Suzuki, H. (1994). The centrality of analogy in knowledge acquisition in instructional contexts. Human Development, 37, 207-
219.

Page 235
Wise, B., Olson, R., & Treiman, R. (1990). Subsyllabic units in computerized reading instruction: Onset-rime versus postvowel
segmentation. Journal of Experimental Child Psychology, 49, 1-19.


Page 237
Chapter 13
Learning to Read and Learning to Spell are One and the Same, Almost
Linnea C. Ehri
City University of New York Graduate School
Over the years, I have conducted several studies investigating how children learn to read words. I designed these studies to
include information about students' memory for spellings of the words they learned to read in the belief that I was studying more
general word learning processes in which treatments would influence spelling as well as reading. Results of these studies
confirmed my belief. The purpose of this chapter is to consider the relationship between reading and spelling. A conceptual
framework distinguishing the important behaviors, processes, and types of knowledge is presented. Similarities and differences
between reading and spelling concepts and phenomena are examined to achieve a clearer understanding of how reading and
spelling are linked during development. Various studies are examined for evidence bearing on these relationships. I build a
partly theoretical, partly empirical case for the claim that learning to read and learning to spell are one and the same, almost.1
1 My presentation does not include a comprehensive review of the literature on this topic. I apologize to researchers who
have published relevant findings that are not cited. The space available precluded such a treatment. The research cited to
support claims is thought to be consistent with other research not cited. I welcome being notified of discrepant findings I
have overlooked. In constructing my case, I have tried to adopt straightforward language and to avoid terms that have
become ambiguous as a result of researchers using them to mean different things (the term orthographic is one of these)
as Wagner and Barker (1994) note. More precise terms replace this term.

Page 238
Reading and Spelling: Conceptual Ambiguities and Overlap
The English language includes two terms, reading and spelling, that are used by researchers and teachers to divide the world.
Researchers classify studies into those focused on reading, those focused on spelling, and those focused on relations between
reading and spelling. Teachers plan lessons to teach reading during one period and different lessons to teach spelling during
another period of the school day. All of this presumes that there are two separate things that are being studied or taught.
It is important to consider to what extent our language reflects reality. Examination of the term spelling reveals that it is
ambiguous in ways that muddy the distinction between reading and spelling. The first meaning of spelling that probably jumps
to mind is the act of spelling a word by writing it. However, spelling can also function as a noun to refer to the product that is
written, the word's spelling, consisting of a sequence of letters. Spellings of words are the targets not only of spelling behavior
but also of reading behavior. Talk about spellings of words for reading blurs the separation between reading and spelling.
Another factor blurring the distinction is that there are two types of spelling acts. Being able to spell may refer to being able to
write out the spellings of words correctly. However, it may also mean being able to recognize whether words are spelled
correctly or incorrectly as the words are being read. Spelling-recognition tasks as well as spelling-production tasks are
commonly used to measure spelling ability. For example, in a task designed by Olson, Kliegl, Davidson, and Foltz (1985) and
used by many researchers, participants are required to read several pairs of identically pronounced words such as RAIN and
RANE and decide which is the correct spelling. Categorizing this as a spelling task rather than a reading task is arguable.
Cleverly, Olson et al. have avoided the issue by labeling this task orthographic coding.
Even the act of spelling words by writing them is not pure. Most spellers write out words and then read the words to verify their
correctness. To the extent that spellers do this when they spell, reading as well as spelling contribute to the final spelling
product.
The upshot of these considerations is to raise doubt that reading and spelling are different. People read spellings of words.
People spell spellings of words. People read the spellings they have spelled. The lack of a clear distinction between these terms
raises the possibility that we have been misled by our language and that reading and spelling are more similar than we
recognize.

Page 239
Reading and Spelling: Basic Concepts and Processes
Although the ambiguity and overlap in the meanings of these terms might appear to be hopeless, some basic distinctions can be
salvaged. Words have spellings, that is, prescribed sequences of letters. Spellings are the targets of two basic literacy acts
involving words: (a) reading spellings in order to determine their pronunciations and meanings; (b) spelling spellings by writing
them. A third literacy act can be derived from the first. The act of reading spellings includes noticing whether they are spelled
correctly. That readers do this spontaneously in the course of reading words was revealed in a study by McConkie and Zola
(1981). They recorded readers' eye movements and documented their sensitivity to subtly misspelled words embedded in the
text they were reading. For example, they found that readers' normal eye-movement pattern was disrupted when they saw
''fracture'' misspelled as FRAOTURE. I refer to this as the act of recognizing misspellings. Tasks requiring students to recognize
misspellings of words include multiple-choice spelling recognition tests, proofreading tasks, and dictation tasks in which a
spelling just written is checked to verify its correctness.
How are reading and spelling phenomena similar and different? In order to draw comparisons, we need to lay out a conceptual
framework that distinguishes and incorporates the various behaviors, processes, and knowledge sources that appear to be central.
Constituents of the framework are introduced in this section, incorporated into a table, and then discussed more fully.
1. Familiarity of spellings. In learning to read and spell, students encounter words whose spellings are familiar because the
forms have been processed on previous occasions and remembered. They also encounter words whose spellings are unfamiliar
because they have not been seen and remembered.
2. Types of knowledge about spellings. In learning to read and spell, students acquire two types of knowledge that they hold in
memory and use to read and spell words: (a) knowledge about the general spelling system, including such regularities as how
graphemes correspond to phonemes across many words; (b) knowledge about the spellings of specific words that have become
sufficiently familiar to be stored in memory.
3. Types of literacy acts involving words. As indicated previously, three types of literacy acts can be distinguished: (a) reading
words to identify their pronunciations and meanings, (b) producing spellings of words, and (c) recognizing whether words are
correctly spelled or misspelled.

Page 240
4. Ways to process words. There are at least three ways to process words in order to read or spell them: by memory, by
invention, by analogy: (a) Words can be processed by accessing information stored in memory from previous experiences with
those specific words. Familiar words are processed in this way; (b) words can be processed by invention, that is, by assembling
pronunciations from letters to decode the words or by generating letters that are plausible symbols of pronunciations to spell
words; (c) words can be processed by analogizing to known words. Either invention or analogy may be used to process
unfamiliar words.2 The three ways of processing words may be used to either read words, produce spellings of words, or
recognize misspellings.
5. Levels of development. To describe development, researchers have distinguished several developmental levels, referred to as
stages or phases. These levels have been proposed to account for word-reading development and word-spelling development
separately. However, the labels and the underlying processes turn out to be very similar. Both sets of labels refer to the
predominate type of knowledge about the alphabetic system that students use to remember the forms of familiar words and to
invent readings or spellings of unfamiliar words. Level 1 (prealphabetic) applies to prereaders who operate with nonalphabetic
information because they know little about the alphabetic system. Level 2 (partial alphabetic) applies to novice beginners who
operate with rudimentary knowledge of some letter-sound relations. Level 3 (full alphabetic) applies to students who possess
more complete knowledge involving grapheme-phoneme units and how these units form words. Level 4 (consolidated
alphabetic) applies to more advanced students who have knowledge of letter patterns as well as grapheme-phoneme units.3
The various concepts previously distinguished are incorporated into Table 13.1 to lay out the processes, knowledge sources, and
behaviors com-
2 There is another way to read unfamiliar words besides the two ways considered here. When unfamiliar words are
encountered in meaningful contexts, readers may predict the words by using contextual cues, or partial graphophonic cues,
or a combination of both. For example, in the context of discussing various types of dinosaurs in a book, readers might
predict that an unfamiliar word known in speech but not in print is "brontosaurus" based on the topic of the text, an
adjacent picture, seeing b at the beginning of the word, or a combination of these cues. This form of invention used in
reading words in text appears to have no parallel in spelling words, so it is not considered here.
3 Conceptualizations of these levels differ among researchers. Some differences involve issues about where boundaries
should be drawn, what the levels should be called, and what phenomena should be included at each level. Some researchers
distinguish additional levels beyond Level 4. My purpose here is to focus on basic distinctions that expose similarities
between the levels proposed for reading and those proposed for spelling and to overlook more fine-grained differences.

Page 241
TABLE 13.1
Processes That Speakers of English Use to Read Words, to Spell Words, and to Recognize Misspellings
of Words That Are Familiar in Print and Words That Are Unfamiliar in Print
Reading Processes Spelling Processes Misspelling Recognition Processes
Reading Familiar Words: Spelling Familiar Words: Recognizing Misspellings of Familiar
Words
1. By MEMORY: Access representation in 1. By MEMORY: Access representation in 1. By MEMORY: Verify match or
lexical lexical detect mismatch
memory on seeing written form; retrieve memory on hearing word; retrieve sequence between spelling in view and
pronunciation-meaning amalgam. of representation of
letters. target word held in lexical memory.
Reading Unfamiliar Words: Spelling Unfamiliar Words: Recognizing Misspellings of Unfamiliar
Words
1. By DECODING: Generate plausible 1. By INVENTION: Generate plausible 1. By DECODING: Apply knowledge
pronunciation spelling by of the
by sounding out and blending (assembling) stretching out pronunciation and alphabetic system to verify match or
units. segmenting it detect
into units. mismatch between letters in spelling
and
sounds in pronunciation of target word.
Knowledge sources possibly used: Knowledge sources possibly used:
Phonological awareness (blending) Phonological awareness (segmentation)
Grapheme-phoneme units Phoneme-grapheme units
Letter patterns, consolidated units Letter patterns, consolidated units
Morphographs (roots, affixes) Morphographs (roots, affixes)
Partial memory for correct letters
2. By ANALOGY: Access familiar word 2. By ANALOGY: Access familiar word 2. By ANALOGY: Verify match or
with analogous with analogous detect mismatch
spelling pattern in lexical memory; blend pronunciation in lexical memory; retrieve between spelling of word in view and
analogous part with new parts. relevant letters; generate remaining letters. representation
of analogous word held in lexical
memory.
Developmental Levels (Phases, Stages)
Reading Words Spelling Words
1. Prealphabetic, logographic 1. Precommunicative
2. Partial alphabetic 2. Semiphonetic, letter-name
3. Full alphabetic 3. Phonetic, phonemic
4. Consolidated alphabetic, orthographic 4. Transitional, morphemic, within word pattern

Page 242
prising reading and spelling domains to reveal their similarities. These processes are now considered in more detail.
Familiar Versus Unfamiliar Spellings of Words
The basic unit of print in English is the word. Whereas smaller units of print, principally graphemes symbolizing phonemes, may
vary in different contexts, the spellings of specific words remain constant. The major task in learning to read and spell is to
become sufficiently familiar with the spellings of words so that information about their letters is retained in memory and enables
them to be read or spelled easily. Beginners must feel overwhelmed by the number of written words that are not familiar,
whereas highly literate adults may feel surprised if they come across words they have never seen before unless the word is a
new vocabulary item.
Findings of several studies show that learners do remember the written forms of specific words. Ehri (1980) found that
beginners stored the spellings of specific words they learned to read in memory rather than alternative, phonetically equivalent
spellings. Reitsma (1983) showed that first graders needed as few as four practices to retain letter information about specific
words in memory. Ehri and Saltmarsh (1995) found that advanced beginners who had practiced reading words several times
retained fairly complete letter information when tested 3 days later.
I have portrayed word familiarity as a bipolar concept in order to distinguish words that readers have seen before and retained in
memory from words they have not. Among beginning readers, this is an important distinction. However, as students gain more
experience with a greater number of words, it becomes important to distinguish degrees of familiarity among the words that are
familiar. The many studies documenting strong word frequency effects on reading and spelling performances verify the
importance of conceptualizing word familiarity as a continuum.
Sources of Knowledge About Word Spellings Held in Memory
At the heart of reading and spelling is knowledge about the spellings of words. Two types of knowledge have been
distinguished: People possess knowledge about the general alphabetic system, that is, spelling regularities that recur across
many different words, including grapheme-phoneme and phoneme-grapheme units and various types of spelling patterns; people
also possess knowledge about the spellings of specific words held in memory as a result of their experiences reading or writing
those particular words (i.e., the domain of familiar words referred to earlier).

Page 243
Knowledge of the Alphabetic System. Knowledge about the alphabetic system and how it works may include a variety of
capabilities and types of information: the names of letters, how to group letters into functional units, called graphemes, that
symbolize phonemes (smallest units in speech), which graphemes typically symbolize which phonemes, how to segment words
into phonemes so that they correspond to graphemes in spellings, how to blend phonemes symbolized by graphemes so that
recognizable words result, and how to group letters into larger units comprising spelling patterns and morphographs that
symbolize syllabic units including common rime stems, root words, and affixes.
Excluded from this are memorized rules that students can state verbatim but have little idea how to apply. My notion of
knowledge of the system involves working knowledge that students actually apply to read and spell. I would expect working
knowledge to operate more implicitly, "in vivo," than knowledge involving verbalized rules. Learning a rule may be the first
step, but if it is not internalized as part of working knowledge, then it remains outside the system.
Beginners are usually taught grapheme-phoneme correspondences explicitly when they first learn to read and spell. Many of
these associations are learned easily if students already know the names of letters. This is because all letter names but W include
relevant sounds, for example /t/ in "tee," /k/ in "kay," /c/ in "aich" symbolized by CH, /ay/ in ''wie'' symbolized by Y as in
SKY.4 Henderson (1981), Read (1971, 1975), and Treiman (1993) among others have shown that beginners make use of this
knowledge in their attempts to spell words. We have shown that beginners use this knowledge in remembering how to read
words as well (Ehri & Wilce, 1985; Scott & Ehri, 1989).
Although letter names take care of 25 associations, there are many more to be learned that are not found in names. These may
be learned through explicit instruction, or they may be acquired in the course of reading and writing specific words and
recognizing regularities that recur across different words. Gough and his colleagues (Gough & Hillinger, 1980; Gough, Juel, &
Griffith, 1992) referred to this as knowledge of the cipher. Venezky (1970) has identified many of the grapheme-phoneme
regularities that comprise the cipher.
Whereas grapheme-phoneme relations are used for reading, phoneme-grapheme relations are used for spelling. According to
Cronnell (1978), correspondences for reading are not completely isomorphic with correspondences for spelling. There are about
40 distinctive phonemes in English, but
4 To represent phonemes, symbols are placed between slash marks, for example, /p/. To represent spellings of words,
capital letters are used. To represent pronunciations of words, quotation marks are used.

Page 244
there are about 70 letters or letter combinations to symbolize phonemes. This makes pronouncing spellings easier than writing
spellings.
Whereas beginners may utilize graphemes and phonemes to read and spell, once students gain more experience with words, they
recognize the presence of letter patterns that recur across words. Graphemes and phonemes in common patterns become
consolidated into multiletter units that are used to read and spell (Ehri, 1995). Becker, Dixon, and Anderson-Inman (1980)
analyzed English words to detect morphographs, defined as irreducible units of meaning in written English, including affixes and
root words. The most common morphographs, all suffixes, are: -ED, -ING, -Y, -ATE, -ER, -ION, -LY. A type of spelling
pattern found in monosyllabic words consists of common rime stems, for example, -ENT, -IDE, -ICK (Fry, 1994; Treiman,
1985). According to Wylie and Durrell (1970), a list of 37 rimes can be used to generate nearly 500 primary-grade words.
Knowing how to read and spell consolidated units such as these eases the task of processing longer multi-syllabic words by
reducing the number of separate units to be processed.
From this discussion, it should be apparent that my concept of systematic alphabetic knowledge is broad, certainly broader than
that proposed by Venezky (1970) or Hanna, Hanna, Hodges, and Rudorf (1966). It includes regularities that others might
consider irregularities. For example, when the spelling of a sound is ambiguous in having multiple alternatives (e.g., /k/ spelled
C, CK, K, or CH), this is still considered systematic knowledge. If spelling patterns that include silent letters recur in several
words, this pattern is systematic, for example, COULD, SHOULD, WOULD, or TALK, CHALK, WALK, BALK.
Distinguishing spellings of words in terms of their origins, Anglo-Saxon versus Latin versus Greek, and learning patterns
associated with each origin, constitutes systematic knowledge (Henry, 1988, 1990). Of course, I am talking about potential
systematic knowledge here. These sources of regularity and others not mentioned are all available for learners to incorporate
into their working knowledge of the system. Whether they actually do is another matter.
Word-Specific Knowledge. Word-specific knowledge consists of information about the spellings of individual words. This
information is derived from reading and spelling experiences and is held in memory. Word-specific knowledge is constructed
primarily out of students' knowledge of the general alphabetic system (Ehri, 1980, 1986, 1992). Knowledge of the system
functions as a mnemonic tool, enabling students to retain letter-specific information about individual words in memory. In
English, specific word learning is necessary because variable spellings are possible. For example, "telephone" might be spelled
TELEPHONE, or TELLAFOAN, or TELUFOWN. To the extent that learners see TELEPHONE and process its grapheme-
phoneme relations, they remember this spelling and not the alternatives. Of course, any of these

Page 245
alternatives is much easier to remember than spellings that lie outside the system, for example, spelling "telephone" as
KOMIKEH. This illustrates how knowledge of the system is central for remembering the spellings of specific words.
It is important to note that this view of word-specific knowledge differs from other views. Kreiner and Gough (1990)
conceptualized word-specific memory as something that operates independently of knowledge of the system. According to their
view, word-specific memory may involve visual configurations of words or serial lists of letters but it does not involve rules and
regularities. This appears to be quite different from the view that knowledge of the system is the primary stuff used to build
word-specific memory.
Word-specific memory arises when students read and write words and retain information about the spellings in memory. First,
we characterize the representations that are thought to be established in memory for reading words, then we consider
representations for writing words.
According to our theory and research findings (Ehri, 1980, 1984, 1987, 1991, 1992, 1994; Ehri & Wilce, 1979, 1980, 1982b,
1985, 1987a, 1987b) and a similar theory by Perfetti (1992), the representations formed in memory from experiences reading
words are alphabetic and consist of graphemes bonded to phonemes. Elsewhere, I have called this sight-word learning because
sight of the word activates its identity in memory (Ehri, 1992). The learning process works as follows: When readers see and
pronounce words, their knowledge of the alphabetic system is activated and computes connections between graphemes in the
spellings and phonemes detected in the pronunciations of the words. Repetition of this process a few times bonds the spelling of
the word to its pronunciation and meaning in memory, forming an amalgam.
Figure 13.1 illustrates the connections that might be formed to remember how to read several different words. Capital letters
designate spellings, single letters and letters grouped together indicate graphemes, symbols between slashes are phonetic
symbols indicating phonemes, and lines linking graphemes to phonemes indicate connections. Notice that some letters are
grouped to form digraphs or trigraphs that depict a functional grapheme symbolizing a phoneme.5 Notice that when the vocalic
phonemes /l/ and /r/ occur alone in an unstressed syllable, they may be treated as one phoneme, as Treiman (1993) has found.
Alternatively, learners may be taught to break the vocalic sound into two phonemes, each connected to a separate graph-
5 It is important to distinguish letters, which are discrete visual units in print, from graphemes, which are functional units
symbolizing phonemes. The only way to distinguish a grapheme in print is to know the phoneme that it symbolizes.
Although many graphemes are single letters, there are also several multiletter graphemes.

Page 246
Fig. 13.1.
Connections formed in memory between graphemes and phonemes
or between consolidated graphemes and syllabic units to remember
how to read specific words.

Page 247
eme. This approach is more consistent with the principle that every syllable has to have a vowel.
This connection forming process is applied to words that are spelled irregularly as well. As illustrated in Fig. 13.1, the majority
of the letters in irregular words are consistent with grapheme-phoneme conventions (i.e., all but the letters marked with an
asterisk in SIGN, ISLAND, SWORD, and LISTEN). To remember letters that do not symbolize sounds, readers may flag them
as silent in memory (Ehri & Wilce, 1982a), or they may create special spelling pronunciations that include the silent letters, for
example, remembering LISTEN as "lis-ten."
If readers' knowledge of the alphabetic system includes commonly recurring combinations of graphemes and phonemes that
have been consolidated, then connections might be formed out of these larger units. In Fig. 13.1, this is illustrated with the
words CHEST, and also INTERESTING that can be seen to have four consolidated units formed out of nine grapheme-phoneme
units. Notice how many fewer connections must be formed to bond a word in memory when letter patterns symbolizing syllabic
units are known by readers.
To summarize, connecting spellings to the pronunciations of words in this way secures representations of the spellings in
memory to support accurate, rapid word reading. As evident in these examples, the key to establishing high-quality
representations of specific words is having general knowledge of the alphabetic system.
Let us contrast what must be done to remember how to read specific words with what must be done to remember how to spell
them. When students remember how to read a familiar word, they access essentially one "response" from memory, a
pronunciation-meaning amalgam. However, when they remember how to spell a familiar word, they must access several
"responses" from memory consisting of individual letters written in the proper sequence. Clearly, more information is needed to
produce a correct spelling than a correct reading.
What is the nature of the representations in memory that enable students to write out all the letters in words correctly? Results
of our research indicate that the spelling-pronunciation-meaning amalgams formed in memory to read words are also useful for
spelling words. In several studies, we have taught beginners to read specific words under various circumstances and then have
asked students to spell the words. In most cases, transfer from reading to spelling was evident. (These studies are discussed more
fully later.) However, learning to read words did not enable most participants to spell the words perfectly. Typically, students
were able to spell a greater proportion of the letters correctly than entire words correctly: 70%-80% of the letters versus 30%-
40% of the words. It may be that perfect spelling requires more than the amalgams formed from reading practice, at least among
first and second graders.

Page 248
As mentioned previously, it is easier to read words accurately in English than to spell them. Failure to remember one or two
letters dooms a perfect spelling but not necessarily an accurate reading. For example, not knowing the spellings of schwa
vowels6 represented by # in EL#PH#NT thwarts spelling but not reading, particularly if the word has been read before. Of
course, accurate readings of words without full letter knowledge become problematic for words having similarly spelled
neighbors, for example, ST#CK may be STICK, STOCK, STACK, or STUCK.
What kinds of letters make spellings especially hard to remember? According to our theory, graphemes that do not follow the
conventional system in symbolizing phonemes should be harder to store in representations than graphemes conforming to the
system. Also, phonemes having many graphemic options should be a bigger burden on memory than phonemes having only a
couple of options. In addition, graphemes that have no correlates in sound, for example, doubled letters and silent letters, should
elude memory. Likewise, spelling patterns that recur in few other words and are not built out of conventional graphemes and
phonemes should cause problems.
Let us examine some words that have parts known to be difficult to spell. The following words include schwa vowels that I have
underlined:
LETTUCE PIGEON EXCELLENT

TERRACE VENGEANCE CHOCOLATE
TENNIS BARGAIN LIMOUSINE
Notice how variable the letters are. Schwa might be spelled with any vowel letter or pairs of letters, hence making it harder for
students to remember the specific letters. Kreiner and Gough (1990) have shown that spellers make more errors on schwa
vowels than on unambiguously pronounced vowels.
The following words are listed as spelling demons by Fry, Polk, and Fountoukidis (1984):
LIEUTENANT SERGEANT RECEIPT AISLE

UNNECESSARY ACCOMMODATE MUSCLE YACHT
CONSCIENTIOUS NOTICEABLE PNEUMONIA VACUUM
What makes these words difficult to spell? According to our theory, students remember best those letters that conform to their
knowledge of the alphabetic system, and they have the hardest time remembering letters that fall outside of the system.
Inspection of these words reveals that all contain
6 Schwa vowels are the vowel sounds that very often occur in unstressed syllables, for example, the final phoneme in
"sofa."

Page 249
problem letters, including nonconventional graphemes, doubled letters, silent letters, and uncommon spelling patterns, which I
have underlined. In addition, many contain schwa vowels.
Brown (1988) performed a study in which college students were exposed to misspellings of several words. This was done by
having them either read plausible misspellings or intentionally write their own misspellings. Comparison of their performances
on spelling tests with the same words before and after their experience with misspellings revealed that their ability to spell the
words correctly deteriorated to some extent, more so than among participants who did not experience misspellings.
Brown (1988) suggested that the students' representations of some words in memory may include more than one representation.
However, inspection of the words he used reveals that most contained problem letters and would qualify as spelling demons.
Another interpretation is that the students' initially correct representations of spellings were susceptible to deterioration because
some of the letters were not secured well in memory because they fell outside the students' knowledge of the system. As a
result, the exercise of computing grapheme-phoneme or phoneme-grapheme connections to either read or write plausible
misspellings of the words established the misspellings as compelling competitors, perhaps even replacing correct forms in
memory.
This explanation receives some support from another finding in Brown's (1988) study. He found that deterioration was much
greater when students were exposed to commonly occurring misspellings of the words than when they intentionally created
misspellings of the words themselves. It is likely that common misspellings are more plausible and, hence, more seductive than
individually produced misspellings.
Given that some spellings are fragile, how can they be better secured in memory? Drake and Ehri (1984) investigated one
possible way. We showed spellers how to create special spelling pronunciations in which sounds were assigned to letters not
pronounced distinctively in the normal spoken forms of words, for example, "chocolate" pronounced "choc" - "o" - "late." We
found that spellers were able to remember nondistinctively pronounced letters better if they practiced spellings this way than if
they practiced dictionary pronunciations that did not make the letters distinctive in speech. However, this strategy did not
improve their memory for doubled letters, not surprisingly because pronouncing does not distinguish the doubled letters in
memory.
Because of the difficulty remembering all of the letters in many English words, those who are good at it may employ special
mnemonic strategies, or they may simply be blessed with good memories for letter details (Simon, 1976; Sloboda, 1980).
Possible strategies include letter name rehearsal, for

Page 250
example, "em-eye-ess-ess-eye-ess-ess-eye-pee-pee-eye" for MISSISSIPPI. Mnemonic rules might be attached to specific
words, for example, RECEIVE ("i" before "e" except after "c"). As previously mentioned, special spelling pronunciations might
be created. Good spellers might also retain motor routines in memory for writing or typing words they have practiced. How
useful such strategies are needs further study.
Of course, most spellers do not practice writing all the words that they become able to spell, particularly as their word-specific
memory grows very large. According to our theory, if students are able to apply their knowledge of the system to connect and
retain in memory all the letters in a specific word they read, they may not need to practice writing the word to spell it
accurately. With the appropriate systematic knowledge, they may even be able to read and remember some spelling demons. For
example, take the words "noticeable," "vengeance," and "pigeon" that create difficulties for students who process medial E as an
extra, inexplicable letter. However, students who recognize that E functions as a marker softening the preceding consonant (C or
G)7 should remember E when they learn to read these words and may be able to spell the words correctly without any writing
practice.
Waters, Bruck, and Malus-Abramowitz (1988) compared students' ability to spell words that exhibited different kinds of spelling
regularities. The children were in third through sixth grades. The hardest words to spell were those having letters that fell outside
the system, labeled strange words such as AISLE and YACHT. Less difficult were words having letters that could be justified if
students knew about affixes and root words, for example, SIGN (related to SIGNAL) and SHORTAGE. Easier than these were
words that could be spelled in alternative, equally legitimate ways. For example, DETAIL might also be spelled DETALE. The
easiest to spell were completely regular words with few alternative legal spellings. In fact, good spellers made few errors writing
these words. These findings support our hypothesis.
In sum, the word-specific knowledge retained in memory from reading words contributes to students' ability to spell words.
However, the knowledge needed to spell at least some words perfectly may exceed that which is typically stored in memory
when words are read, particularly when the words include letters that fall outside the speller's knowledge of the alphabetic
system. Spellers may need to invoke special strategies in order to perfect their spelling ability in these cases. Of course, if
spellers are taught about additional ways that the system is regular, this should reduce the number of words requiring special
strategies to remember.
7 For those who are unfamiliar with this regularity of the writing system, notice how the sound of C and G becomes hard
(/k/, /g/) or soft (/s/, /j/) systematically in the following words depending on whether E follows the consonant: manic
versus malice; pack versus pace; code versus cede; bag versus badge; sing versus singe; game versus gem (Venezky,
1970).

Page 251
Three Ways to Read and Spell Words: By Memory, By Invention, By Analogy
Having discussed the two types of word knowledge that students hold in memory, we can examine how these sources are used
to process words. There are at least three distinct ways to read and spell words: by memory, by invention, and by analogy.
Which process is used depends on whether the written forms of the words are familiar or unfamiliar. Familiar words are read or
spelled by accessing representations in memory. Unfamiliar words are read or spelled by invention or by analogy. Various
strategies offered by third graders to explain how they spelled particular words illustrate the three different approaches
(Varnhagan, 1995): (a) "I remembered the word;" (b) "I sounded it out." "Can't put another S so put ES at the end;" and (c)
"Just like `old' but with a C." ''I borrowed the GHT from `light.'''
By Memory. When information about the written forms of specific words is available in memory, this information is accessed to
read or spell the words. We have already discussed the memory processes involved in reading familiar words and in producing
correct spellings.
By Invention. Children learning to read and spell typically encounter many words that are not familiar and, hence, require attack
strategies. One attack strategy for reading words involves decoding the word, that is, applying knowledge about spelling-sound
units to assemble a plausible pronunciation that is recognized as a real word. In English, because the spelling system is variable,
some flexibility is required in the use of this strategy to derive a recognizable word.
The attack strategy for inventing spellings of words involves stretching out their pronunciations, detecting sound units, and
applying knowledge of the alphabetic system to generate plausible letter sequences for the sounds. As indicated previously, the
likelihood that spellers will generate correct sequences of letters in English is limited because the system is quite variable.
The alphabetic knowledge used to invent is similar in the case of reading and spelling. The knowledge sources are listed in
Table 13.1. Reading requires blending skill to assemble a unified pronunciation from the separately decoded parts; spelling
requires segmentation skill to pull apart and distinguish the phonemes to be spelled with graphemes. Both segmenting and
blending are phonological awareness skills that many beginners do not acquire easily without explicit instruction (Liberman &
Shankweiler, 1979; Lundberg, Frost, & Peterson, 1988). Beginners rely primarily on grapheme-phoneme or phoneme-grapheme
units (GPUs). More advanced readers and spellers may use their knowledge of spelling patterns, consolidated GPUs,

Page 252
and morphographs in their inventions (Bowey & Hansen, 1994; Laxon, Coltheart, & Keating, 1988; Templeton & Bear, 1993;
Treiman, Goswami, & Bruck, 1990; Venezky & Johnson, 1973; Waters et al., 1988).
In spelling unfamiliar words, students might also be influenced by partial information they hold in memory about specific words
if they remember some but not all of the letters. For example, Simon and Simon (1973) found that the spellings of unfamiliar
words generated by fourth graders were more accurate than spellings generated by a computer applying a spelling algorithm.
They attribute this to students' partial word-specific knowledge. An example makes this difference apparent: "efficiency" spelled
EFITIONSY by the computer versus EFFECIENCY by the students.
By Analogy. Another way to read or spell unfamiliar words is by analogy to familiar words already represented in memory. For
example, students might encounter the new word "manic." Those viewing MANIC and attempting to read it or those hearing
"manic" and attempting to spell it might recognize that it rhymes with a word whose spelling they already know, PANIC.
In order to process words by analogy, students must have information about an analogous familiar word stored in memory in
enough letter detail to perform a comparison, they must recognize the similarity, and they must access and adapt their
knowledge of the familiar word in reading or spelling the unfamiliar word. Of course, in English, this does not guarantee a
correct response, particularly in spelling words where a number of phonemically equivalent options exist, for example, BREAD
versus RED versus SAID, FINE versus SIGN, RATE versus BAIT versus FREIGHT.
Goswami (1986, 1988a, 1988b) has studied the process of analogizing and has found that even beginners can use their
knowledge of familiar words to read or spell unfamiliar words. Ehri and Robbins (1992) have found, however, that beginners
need some decoding skill to read by analogy. That is, they need sufficient knowledge of the alphabetic system to store full
spellings of specific words in memory so that they can detect letter similarities between known and new words. In addition, they
need blending skill to put the relevant parts of known and new words together.
Campbell (1983, 1985) was able to demonstrate spelling by analogy in adults and older children. Students' spellings of
nonwords were influenced by the spellings of real words they heard. For example, if the participants heard "brain," they were
more apt to spell /pren/ as PRAIN whereas if they heard crane, they were more apt to spell the nonword as PRANE. Snowling
(1994) observed this same effect in children as young as 8 years of age. Success in reading or spelling words by analogy
requires remembering related words in sufficient letter detail to perform the transfer. Snowling and

Page 253
Bowey and Hansen (1994) found that analogizing was more common in older students than in beginners.
Levels of Development
Various developmental schemes have been proposed to portray how reading and spelling processes change as students acquire
word-reading skills and word-spelling skills (Beers & Henderson, 1977; Ehri, 1986, 1991, 1994, 1995; Frith, 1985; Gentry,
1982; Henderson, 1981; Morris & Perney, 1984; Templeton & Bear, 1992). The levels within these schemes have been referred
to as stages in some cases and phases in other cases. Although the labels for levels differ somewhat across schemes, the
underlying distinctions between less and more mature levels are similar.8 I have labeled these levels with reference to students'
knowledge of the alphabetic system because this is the key capability that distinguishes among the levels and underlies
development. Each level characterizes the approach that predominates at that level. Development at later levels is viewed as
contingent on development at earlier levels except at Levels 1 and 2.
Prealphabetic Level. At the earliest level, referred to as prealphabetic, logographic, or precommunicative, children know little
about the alphabetic system as it represents speech. In remembering how to read words, they store salient visual features, for
example, the two eyes in LOOK (Gough, Juel, & Griffith, 1992). This explains how they read familiar labels and signs in the
environment. For example, they read McDonalds by remembering its golden arches rather than any letters in its spelling
(Masonheimer, Drum, & Ehri, 1984). At this level, children have difficulty remembering how to read most words because the
connections formed in memory are unsystematic or arbitrary. Their only strategy for reading unfamiliar words is guessing.
In writing words, children may produce scribbles that superficially resemble cursive writing but lack any letter details (Harste,
Woodward, & Burke, 1984). If letters are included, they are not selected because they correspond to any sounds in the words
being written. As a result, children cannot read what they have written. Because children's alphabetic reading and writing skills
are limited or nonexistent, their achievements here probably have little impact on progress at the next level (Byrne, 1992).
8 My levels do not correspond to those proposed by Frith (1985) in all respects. She regards Levels 1 and 2 as part of her
logographic stage. However, results of several studies (Ehri & Wilce, 1985; Rack, Hulme, Snowling, & Wight, 1994)
indicate that novice readers and spellers who possess rudimentary alphabetic knowledge learn to read and spell words
much differently from prereaders who lack this knowledge and rely on nonalphabetic information.

Page 254
Partial Alphabetic Level. Once children learn the names or sounds of alphabet letters, they become capable of reading and
writing words at the next level of development, referred to as partial alphabetic or semiphonetic. Students' knowledge of the
alphabetic system is rudimentary and incomplete at this level. It includes sounds found in the names of letters, including long
vowels. However, other vowels and consonants are less well known.
To remember how to read words they have read before, they use their knowledge of letter names or sounds to form connections
between salient letters seen in words and sounds detected in pronunciations, for example, remembering how to read BEAVER
by connecting B and R to the names of these letters heard in the pronunciation, or remembering STOP by connecting initial and
final letters and sounds. We have called the use of partial-letter cues phonetic-cue reading (Ehri & Wilce, 1985, 1987a, 1987b;
Scott & Ehri, 1989). Because only some of the letters are remembered, words with the same letters may be confused, for
example, BEAVER and BROTHER, or STOP and SKIP.
The only way that children can read words at this level is by remembering them or by guessing them based on context cues.
They are unable to invent pronunciations of unfamiliar words by decoding letters because they lack sufficient knowledge of the
graphophonic system, particularly vowels. In attempting to read nonwords, they may guess real words having some of the same
letters (e.g., KUG guessed as "king").
Reading and spelling words by analogy is not really possible for students with partial alphabetic knowledge if the analogues are
not in view and must be accessed from memory. This is because their memory for specific words lacks sufficient letter detail.
Ehri and Robbins (1992) found that novice readers often mistook new words for the familiar words they were analogous to, for
example, misreading SAVE as "cave" because the new word SAVE contained the letter cues -AVE used to remember the
known word CAVE.
Correct spellings of words are very difficult for children to remember at this level because they lack sufficient knowledge of the
alphabetic system (Mason, 1980). Most of the letters they see lie outside their knowledge of the system and, hence, appear to be
arbitrary.
To invent spellings, students detect and spell only some of the sounds in words, mainly the more salient consonants and vowels
whose names are heard in the words. In writing the word BEAVER, they might detect consonant letter names or sounds at the
boundaries of syllables and write BR or BVR. A letter-name strategy predominates in their writing. This yields some peculiar
spellings, such as HKN for "chicken" with H selected because /c/ is heard in its name "aich." Their spellings are partial rather
than complete representations of the sounds in words for at least two reasons: They have difficulty detecting and segmenting
words into phonemes, and they do not know how to represent all the sounds with letters, particularly vowel

Page 255
sounds. (For fuller treatment of this level of spelling development, see Henderson, 1981; Read, 1971, 1975; Treiman, 1993.)
Full Alphabetic Level. To operate at this level, students need to be able to segment words into constituent phonemes. Also, they
need to know conventional grapheme-phoneme units, particularly how vowels are symbolized with letters. With this knowledge,
they become able to take full advantage of the alphabetic system for reading and spelling words. They can process spellings of
words completely to remember how to read them when letters are consistent with their grapheme-phoneme knowledge. They are
able to generate plausible pronunciations of unfamiliar written words by applying decoding skill. This level of development is
referred to as full alphabetic, or phonetic, or phonemic.
Children operate mainly by applying their knowledge of grapheme-phoneme correspondences in reading, and phoneme-
grapheme correspondences in spelling. The spellings they invent are more complete than those generated at the semiphonetic
level. In fact, in stretching out the sounds in words to spell them, children may find extra sounds not symbolized in conventional
spellings, for example, spelling "blouses" as BALAOSIS (Ehri, 1986). At this level, children understand the basic system that
underlies and accounts for the presence of many letters in conventional spellings, so they find it easier to remember how to read
and write words they have practiced than children at the partial alphabetic level. Also, they have stored words in memory in
sufficient letter detail so that reading new words by analogy to familiar words is possible.
Consolidated Alphabetic Level. As children practice reading and writing conventional spellings of words, they learn about the
structure of larger units in words consisting of letter sequences that recur across several words. These units may involve
spellings of syllables, or parts of syllables, or affixes appearing at the beginnings or endings of words, for example, -ING, -ED,
-IMP, -ABLE, and -TION. Whereas at the previous level such units were analyzed into smaller graphophonic units, at this level
they become consolidated into larger units. In addition, students may internalize other kinds of regularities, for example, letter-
doubling rules (e.g., LATER vs. LATTER, WADED vs. WADDED) and other vowel-marking rules (Templeton & Bear, 1992;
Venezky, 1970). This developmental level is referred to as consolidated alphabetic, transitional, within-word pattern, or
morphemic.
Operating with chunks of letters makes it easier for students to decode and invent spellings of longer multisyllabic words. Also,
they have an easier time forming connections between letters and sounds to retain the spellings of specific words in memory.
This is because their knowledge of the system is more extensive and also because some of this knowledge is redundant.

Page 256
For example, if they know -EST and -OUND as consolidated units symbolizing rimes, they also know the constituent grapheme-
phoneme units. As a result, these units may be double-secured in the representations of specific words.
Evidence for Reading-spelling Relationships

Our description of developmental levels shows that knowledge of the alphabetic system differs at each level, and these
differences are reflected in both the reading and spelling performances of students. Now let us turn to empirical findings to seek
further evidence for the claim that reading and spelling are closely related, perhaps almost the same.
Correlation Coefficients. How closely associated are word-reading performances and word-spelling performances when
measured by word-reading tests and word-spelling tests? Are good readers almost always good spellers and poor readers almost
always poor spellers? Are the relationships stronger or weaker at some points of development than at other points? To find out,
we can extract correlational findings from various studies in which students of various ages were asked to read and spell words
(Ehri & Wilce, 1982a; Greenberg, Ehri, & Perin, in press; Griffith, 1987, 1991; Jorm, 1981; Juel, Griffith, & Gough, 1986). The
reading tasks required students to read a list of English words. The spelling production tasks required students to write English
words to dictation. The misspelling recognition tasks required students to distinguish correct from incorrect spellings of English
words. In some studies, the same words were used across tasks; in other studies, different words were used.
From the correlations in Table 13.2, it is apparent that reading and spelling performances were highly related in these studies.
To realize the strength of these relations, we can note that most of the correlations are above r = .70, which brings them close to
reliability values expected of tests measuring one capability before being corrected for the unreliability of individual measures.
Such high correlations indicate that very similar processes are measured in these tasks. This is true among younger as well as
older students.9
9 In the study by Griffith (1991), first graders were tested as well. I did not include their correlation because mean
performances on the recognition and productions tests were low, with smaller standard deviations than for third graders,
suggesting that the tests may have lacked sensitivity at this level.

Page 257
TABLE 13.2
Correlations Between Reading Words, Producing
Correct Spellings of Words, and Recognizing Misspellings
of Words at Various Grade Levels Across Different Studies
Read Read Spell
Studies Spell Recognize Recognize
Misspellings Misspellings
Juel, Griffith, & Gough (1986). Different words were read, spelled, and
recognized.
First Graders .84 .74 .76
Second Graders .77 .69 .68
Griffith (1987). Same words were read and spelled.
First Graders .83
Third Graders .84
Greenberg, Ehri, & Perin (in press). Different words were read and spelled.
Third-Fifth Graders .86
Jorm (1981). Same words were read and spelled.
Fourth-Sixth Graders .85
Griffith (1991). Different words were spelled and recognized.
Third Graders .80
Ehri & Wilce (1982). Different words were spelled and recognized.
Seventh Graders .77
College Students .78
Effects of Reading on Spelling. According to our theory, students retain word-specific information in memory when they learn
to read words, and this information is available to support spelling performance. Transfer from reading to spelling was evident in
a study with second graders (Ehri, 1980). Participants practiced reading one or another phonemically equivalent spelling for each
of eight made-up words until they could read them perfectly. Examples of the spellings are: WHEOPLE versus WEEPEL,
BISTION versus BISCHUN, GHIRP versus GURP (both members pronounced identically). After a 4-minute delay, students
wrote from memory the spellings that they had read. A majority, 69%, were recalled perfectly, indicating that substantial
transfer from word reading to word spelling occurred, despite alternative ways to spell the words.
Even when students misspelled the words, they restricted their letter choices to those they had seen in the words rather than
phonemically equivalent alternatives. For example, all six students who misspelled WHEOPLE began it with WH, not WE-,
whereas all four students who misspelled WEEPEL began it with WE-, not WH-. All four students who misspelled BISCHUN
included CH, whereas none of the four students who misspelled BISTION included CH. All four students who misspelled
GHIRP used I to spell the vowel even though it has no distinctive sound, whereas I was not included

Page 258
in misspellings of GURP. This provides further evidence that word-specific knowledge retained from reading experiences
influenced students' spellings.
Results of another study also revealed the impact of reading words on students' memory for their spellings (Ehri & Wilce, 1986).
In this study, we used words containing medial flaps that are pronounced more like /d/ in American English but might be spelled
with either D or T, for example, HUDDLE, MODIFY, and PEDIGREE versus METEOR, GLITTER, and ATTIC. Second
graders were exposed to 12 words, half with flaps spelled D and half with flaps spelled T. Half of the subjects practiced reading
the words; the other half heard and practiced saying the words but never saw spellings. Participants practiced the words on one
day and then wrote spellings on the next day.
We expected that students who read the words would connect graphemes to phonemes and would remember the flap in each
word as /d/ or /t/ according to its spelling, whereas students who only listened to the words would spell the flap phonetically as
/d/ in most of the words. This was what we found. Participants who read the words spelled 84% of the flaps accurately, whereas
controls spelled only 64%. (If students spelled all the flaps with D, their accuracy would be 50%.)
Whereas students in the WEEPEL study (Ehri, 1980) spelled words shortly after they read them, students in this study (Ehri &
Wilce, 1986) spelled words on a different day. Despite the delay, students' spellings still reflected memory for word-specific
information, indicating that memory was long term. We found that the participants' memory for complete spellings of the words
in the second study was weak, only 31%, probably because the words contained problem letters such as double consonants (see
earlier examples of words). The fact that word-specific effects were evident in spellings shows that if students have partial letter
information about specific words in memory, they do not ignore this knowledge and invent a spelling; rather, they access the
letters they remember and invent the part they do not remember, as suggested in Table 13.1.
In another study (Ehri & Roberts, 1979), we found that the way students learned to read a set of words influenced their memory
for the words' spellings. First graders practiced reading 16 words either in written sentence contexts or in isolation on lists. They
learned to read the words perfectly. In a spelling test given a day later, students who had read the words in isolation produced
significantly more letters correctly than students who had read the words in context: 68% versus 64%. These findings show that
variations in reading experiences may impact spelling performance.
The focus of transfer effects from reading to spelling in the foregoing studies involved specific words. In other studies, students'
knowledge of the alphabetic system was manipulated. In a short-term laboratory study with kindergartners, half of the students
were taught to decode CVC (consonant-

Page 259
vowel-consonant), CCVC, and CVCC words, and half practiced letter-sound relations in isolation (Ehri & Wilce, 1987a). On
posttests, the students who received decoding training not only learned to read a set of words more effectively but also spelled
them more phonetically and more accurately. Differences between the groups were sizeable.
Foorman, Francis, Novy, and Liberman (1991) reported similar findings in a classroom based study. Students who were taught
to sound out and blend during their year in first grade exhibited superior spelling ability at the end of the year compared to
students who were not taught explicit decoding but did receive other forms of letter-sound training.
To summarize, results of these studies confirm that reading impacts spelling in beginners. Findings indicate that readers retain
word-specific information in memory that they use to spell the words. The way students practice reading words can influence
their ability to spell the words. Instruction that improves students' general knowledge of the alphabetic system through reading
benefits their spelling ability as well.
Effects of Spelling on Reading. Not only effects of reading on spelling but also effects of spelling on reading have been studied
in beginners. Morris and Perney (1984) tested first graders' ability to invent spellings of words before they had received any
formal reading instruction. Most children knew all the letters of the alphabet but they were able to spell few words correctly,
only 9%. Their spelling inventions were scored to reflect their level of development, with phonetic, conventional letter choices
receiving more points than partial nonconventional choices. Results revealed a surprisingly high correlation, r = .68, between
spellings invented at the beginning of the school year, and reading achievement scores at the end of the school year. The
correlation rose to r = .82 between spellings invented midyear and year-end reading scores. The likely explanation is that
invented spellings reflect children's knowledge of the spelling system that determines how quickly they get off the ground and
make progress in learning to read.
The most important systematic knowledge that children need to have when they begin learning to read and spell is phonemic
segmention skill and knowledge of letters. Share, Jorm, Maclean and Matthews (1984) among others have shown that these are
the best predictors of reading achievement measured 1 and 2 years later, better even than measures of intelligence.
We performed a short-term experiment with beginners to examine the effects of spelling training on word reading (Ehri &
Wilce, 1987b). In this study, we manipulated learners' knowledge of the general alphabetic system rather than word-specific
knowledge. The students were kindergartners selected because they had limited ability to read words and could not decode.
They were at the partial alphabetic level of develoment. Experimental par-

Page 260
ticipants were taught to segment phonemes and symbolize them with graphemes in CVC, CCVC, and CVCC words and
nonwords. Control participants practiced isolated phoneme-grapheme associations. Then the participants were given several
trials to learn to read a set of 12 similarly spelled words. Comparison of performances revealed that spelling-trained participants
learned to read significantly more words than control participants. Our explanation is that spelling instruction improved the
participants' working knowledge of the alphabetic system. With this knowledge, they were able to form more complete
grapheme-phoneme connections to remember how to read the words than control participants.
In our study, spelling instruction improved their ability to learn to read a set of words with practice, but it did not improve their
ability to decode unfamiliar words because it did not include lessons in how to assemble and blend graphemes into phonemes.
This same absence of transfer from spelling instruction to decoding was found in a study by Hohn and Ehri (1983).
However, Uhry and Shepherd (1993) conducted a spelling instruction study in first-grade classrooms and observed effects on
decoding as well as word learning. Participants were first graders with above average IQs. Spelling training lasted much longer,
6.5 months, and included having children not only segment and invent spellings of words but also read the words they had just
spelled. Control participants practiced reading the same words but did not segment or spell them. In their classrooms, both
groups received reading instruction during the training period. On posttests, spelling-trained participants outperformed controls
in nonword decoding as well as word reading, spelling, segmenting, and oral blending. These findings suggest that spelling
instruction can facilitate decoding ability as well as specific-word reading if it is structured to improve blending ability.
In a classroom-based study by Clarke (1988), two different forms of spelling instruction were examined for their impact on
reading over the course of first grade. One form involved having first graders invent spellings of words in their classroom
writing assignments. The other involved requiring children to use conventional spellings in their writings. At the end of the year,
test performances revealed an effect of spelling treatment on word reading, with children in the invented spelling condition
outperforming children in the traditional spelling condition. However, this difference was detected only among children who
began first grade with the least knowledge about letters, reading and spelling, not among children who began school with more
knowledge. Our explanation is that invented spelling experiences helped novices learn more about the alphabetic system than
conventional spelling experiences. Requiring novices to use conventional spellings when they could not comprehend the letter
structure of the spellings limited their progress. In contrast, more advanced students understood enough about the system to
benefit equally from inventions and conventional forms.

Page 261
In sum, from the analyses of processes explaining results of these studies, we infer that the reason why spelling helps reading is
that spelling instruction helps to cultivate students' knowledge of the alphabetic system which benefits processes used in reading.
This conclusion is supported by results from other studies as well (Ball & Blachman, 1991; Bradley and Bryant, 1985; Byrne &
Fielding-Barnsley, 1989, 1990; Lundberg, Frost, & Peterson, 1988).
In one other study, we examined transfer from spelling to reading that involved specific-word knowledge rather than systematic
knowledge. We manipulated second graders' ability to spell specific words and looked for an impact on their reading behavior
with the words (Roberts & Ehri, 1983). Although one group learned to spell the words significantly better than the other group
(i.e., 88% vs. 83% of the letters correct), the groups did not differ either in their ability to read the words accurately, which was
close to perfect, or in their speed to read the words. It may be that deliberate practice rehearsing letters in spellings does not
affect reading, or perhaps our groups were too similar in their knowledge of spellings to observe a differential effect on reading.
Whether word-specific knowledge transfers from spelling to reading awaits further study.
Are Reading and Spelling Independent or Merely Unstable Among Novices? When children first acquire letter knowledge and
move into reading and writing, their knowledge of the system is just beginning to form. As discussed earlier, beginners invent
forms that are not conventional, and they do this by resorting to various strategies that may draw their attention to one or another
aspect of sound that they link to a letter. In stretching out the sounds in a word, they may detect extra vocalic elements that they
spell, for example, the O in SOWEMEG (swimming). In carefully analyzing their own speech, they may detect affrication or
voicing onset, and this prompts an unconventional spelling, for example, HRK (truck), or SGAK (skate; Read, 1971; Treiman,
1993). Novices are likely to shift in what they attend to and what they ignore as they invent spellings, as there is much that
might be noticed and little that they know for sure about how the system works. As a result, it is not surprising to find
inconsistency in the features that they represent in their writings.
Likewise, in analyzing novices' word reading performances, we have observed substantial variability. When students at the
partial alphabetic phase of development practice remembering how to read a set of words, they read them quite inconsistently
and mistake one word for another when the two words have letters in common (Ehri & Wilce, 1985, 1987a, 1987b).
Bradley and Bryant (1979, 1985) have observed disparities in beginners' word reading and word writing and have suggested that
the two processes operate independently during this early period of development. They examined whether beginners' readings
and spellings of words were the same or

Page 262
different on two different occasions. Results showed that children misread some words that they were able to spell correctly on a
different day. This was interpreted as evidence for independence. However, Gough et al. (1992) obtained support for a different
interpretation.
Gough et al. (1992) conducted a similar study in which students performed one or another combination of tasks. The same set of
words was either read and reread, spelled and respelled, or read and spelled on two different days. They found a general
inconsistency in beginners' performances that was not limited to instances where words were read and spelled; rather,
inconsistency prevailed to the same extent across all the combinations of tasks, ranging from 8% to 11%. Importantly, students
dealt with the majority of the words consistently (72% to 82%) across the task combinations.
These findings challenge Bradley and Bryant's (1979) conclusion that beginners use a different knowledge source to read and
spell. Consistency, not inconsistency, predominates even at this beginning level. Recall that in Morris and Perney's (1984) study,
spellings invented by novices were highly predictive of later reading. I suggest that even at an early point in development,
despite some tendency to be inconsistent, knowledge of the alphabetic system is the major factor imposing consistency on
reading and spelling performances.
Reading-Spelling Relationships in Disabled Readers and Spellers. So far, we have discussed processes in normally developing
readers and spellers. What about disabled readers and spellers? If students are poor in reading, are they also poor in spelling?
According to the theory and evidence presented so far, reading and spelling are manifestations of the same underlying
knowledge sources. If problems in reading are apparent, we expect to see problems in spelling as well.
There are at least two investigations of disabled reading and spelling in which correlation coefficients between reading and
spelling performances have been calculated for disabled students separately from normally developing students. Results support
expectations in showing that reading and spelling performances are still strongly correlated among disabled students. However,
the correlations are not as high as they are with normal readers, indicating that the underlying processes may be less
interconnected and interdependent in disabled readers and spellers.
In an early study by Guthrie (1973), 19 normal second-grade readers and 19 older disabled readers matched in reading age to the
normals were given a four-choice spelling recognition task in which they circled the correct spellings of nonwords they heard.
Also they read several short-vowel words and nonwords. Correlations between the spelling task and the other two tasks were:

Page 263
Normal readers: r = .84 (words); r = .91 (nonwords) and Disabled readers: r = .68 (words); r = .60 (nonwords).
All the correlations were positive and strong, but those among normal readers were substantially higher than those among
disabled readers.
In a study by Greenberg et al. (in press), 72 normal readers in third to fifth grades were compared to 72 adults matched for
reading age. The adults were enrolled in adult literacy programs. Greenberg et al. gave tasks to measure spelling production,
word reading (WR), nonword reading (NR), and phoneme deletion (PD). Correlations between spelling production and the other
measures were:
Normal readers: r = .86 (WR), .62 (NR), .62 (PD) and
Disabled readers: r = .57 (WR), .41 (NR), .38 (PD).
As in the other study, all the correlations were positive and significantly greater than zero, but normal readers' values were
substantially higher than those for disabled readers. The difference was not attributable to differences in the size of the standard
deviations in the two groups.
One explanation for the lower correlations is that progress is impaired when reading and spelling processes do not become
sufficiently integrated as they develop. Various studies confirm that disabled readers possess weaker knowledge of the
alphabetic system than normal readers matched for reading age (Greenberg et al., in press; Rack, Snowling, & Olson, 1992).
Because of this, knowledge of the alphabetic system may not contribute as strongly to the establishment of word-specific
representations in lexical memory among disabled readers as it does among normal readers. Disabled readers may use context to
guess words rather than read the words by computing grapheme-phoneme connections (Stanovich, 1980), hence limiting the
opportunity for reading to impact spelling.
According to our view, the high correlations between reading and spelling performances indicate that the two abilities are so
closely related as to look like the same ability, almost. However, in opposition to this is research by Frith (1980) that
distinguished a small minority of individuals above age 12 who possessed adequate reading skill but were severely disabled in
spelling skill. Such students appear to contradict our view.
In a recent replication and extension of Frith's (1980) study, Newman, Fields, and Wright (1993) compared 14-year-old spelling-
disabled students to reading/spelling disabled (RSD) students and normal readers. In a retrospective analysis, they found that the
spelling-disabled students had exhibited not only spelling difficulties but also reading difficulties during their earlier years. Their
history was similar to that of the RSD group. The main difference between the spelling-disabled group (SD) and the other two
groups was that the SD group revealed superior verbal IQ scores.

Page 264
Bruck and Waters (1990) compared spelling-disabled students to RSD and normal-reading students on various tasks and found
that spelling-disabled students' oral language skills and reading comprehension were comparable to that of normal students but
their word decoding skills were worse and more similar to that of the RSD students. Also, spelling-disabled students read text
more slowly than normal students. On sentence completion tasks, performance of spelling disabled students indicated greater
reliance on context for reading words than normal readers, revealing the use of compensatory strategies in their text reading.
These findings indicate that strong verbal ability enabled the spelling-disabled students to overcome earlier reading difficulties
through the use of compensatory processes. Of importance for our purpose is the fact that spelling disability did not emerge in
the absence of reading disability. Moreover, when a close look was taken at students' word reading, specific weaknesses were
evident, revealing that spelling disabled students are not ''good readers'' at the level of words. Thus, results do not challenge the
close relationship between word reading and word spelling.
Conclusion: Are Reading and Spelling one and the Same, Almost?
I have built a partly theoretical, partly empirical case for the claim that reading words and spelling words are one and the same,
almost. I conclude that they are one and the same because they depend on the same knowledge sources in memory: knowledge
about the alphabetic system and knowledge about the spellings of specific words. I conclude that reading and spelling are not
quite the same because the response performed to read words differs from the response performed to spell words. The act of
reading involves one response, that of pronouncing a word. In contrast, the act of spelling involves multiple responses, that of
writing several letters in the correct sequence. More information is needed in memory to spell words accurately than to read
words.
According to my theory, the underlying knowledge sources may fully support both reading and spelling responses for many
words that conform to students' knowledge of the system, provided students have read those words enough times in a way that
activates connection-forming processes to secure letters in memory. Underlying knowledge sources provide less support for
reading and spelling responses when specific words include letters that lie outside the students' knowledge of the system. This is
because letters that are not systematic are harder to remember, and even if remembered are weakly secured and hence unstable
in the representations.
It is important to consider how my theory and evidence relate to the issue of literacy instruction, because there is room for
misunderstanding

Page 265
here. I have shown that reading and spelling processes are very closely related in skilled readers and spellers. However, this
should not be construed to mean that we can leave the acquisition of spelling skill to the work of reading instruction and
practice. This is because the memory requirements for spelling English words accurately exceed the memory requirements for
reading words accurately. Moreover, acquiring knowledge of the alphabetic system is central to development and should fall
within the province of spelling instruction. It is unlikely that as students learn to read more and more words, they also
necessarily acquire more advanced levels of alphabetic knowledge. In addition, teaching students to read without also teaching
them to spell may result in reading and spelling skills that are less closely related, a condition characterizing poor readers and
spellers. It is clear that students need explicit spelling instruction as well as explicit reading instruction. According to my theory,
the key to effective instruction is integration, that is, fostering close articulation among reading and writing knowledge sources
and processes so that their acquisition is mutually facilitative and reciprocal. There is no getting around it by subordinating
spelling to the job of computer spell-checkers. Poor spellers do not develop into skilled readers. Spelling instruction must
remain an important goal of teachers and schools.
References
Ball, W., & Blachman, B. (1991). Does phoneme segmentation training in kindergarten make a difference in early word
recognition and developmental spelling? Reading Research Quarterly, 26, 49-66.
Becker, W., Dixon, R., & Anderson-Inman, L. (1980). Morphographic and root word analysis of 26,000 high frequency words.
Eugene, OR: University of Oregon College of Education.
Beers, J., & Henderson, E. (1977). A study of developing orthographic concepts among first graders. Research in the teaching
of English, 2, 133-148.
Bowey, J., & Hansen, J. (1994). The development of orthographic rimes as units of word recognition. Journal of Experimental
Bradley, L., & Bryant, P. (1979). The independence of reading and spelling in backward and normal readers. Developmental
Medicine and Child Neurology, 21, 504-514.
Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading and spelling. Ann Arbor, MI: University of Michigan Press.
Brown, A. (1988). Encountering misspellings and spelling performance: Why wrong isn't right. Journal of Educational
Bruck, M., & Waters, G. (1990). An analysis of the component spelling and reading skills of good readers-good spellers, good
readers-poor spellers, and poor readers-poor spellers. In T. Carr & B. Levy (Eds.), Reading and its development (pp. 161-206).
San Diego, CA: Academic Press.
Byrne, B. (1992). Studies in the acquisition procedure for reading: Rationale, hypotheses and data. In P. Gough, L. Ehri, & R.

Page 266
Byrne, B., & Fielding-Barnsley, R. (1989). Phonemic awareness and letter knowledge in the child's acquisition of the alphabetic
principle. Journal of Educational Psychology, 81, 313-321.
Byrne, B., & Fielding-Barnsley, R. (1990). Acquiring the alphabetic principle: A case for teaching recognition of phoneme
identity. Journal of Educational Psychology, 82, 805-812.
Campbell, R. (1985). When children write nonwords to dictation. Journal of Experimental Child Psychology, 40, 133-151.
Clarke, L. (1988). Invented spelling versus traditional spelling in first-graders' writings: Effects on learning to spell and read.
Research in the Teaching of English, 22, 281-309.
Cronnell, B. (1978). Phonics for reading versus phonics for spelling. Reading Teacher, 32, 337-340.
Drake, D., & Ehri, L. (1984). Spelling acquisition: Effects of pronouncing words on memory for their spellings. Cognition and
Instruction, 1, 297-320.
Ehri, L. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive processes in spelling (pp. 311-338).
Ehri, L. (1984). How orthography alters spoken language competencies in children learning to read and spell. In J. Downing &
R. Valtin (Eds.), Language awareness and learning to read (pp. 119-147). New York: Springer-Verlag.
Ehri, L. (1986). Sources of difficulty in learning to spell and read. In M. Wolraich & D. Routh (Eds.), Advances in
developmental and behavioral pediatrics (pp. 121-195). Greenwich, CT: Jai.
Ehri, L. (1987). Learning to read and spell words. Journal of Reading Behavior, 19, 5-31.
Ehri, L. (1991). Development of the ability to read words. In R. Barr, M. Kamil, P. Mosenthal, & P. Pearson (Eds.), Handbook
of reading research (Vol. II, pp. 383-417). New York: Longman.
Ehri, L. (1992). Reconceptualizing the development of sight-word reading and its relationship to recoding. In P. Gough, L. Ehri,
& R. Treiman (Eds.), Reading acquisition (pp. 107-143). Hillsdale, NJ: Lawrence Erlbaum Associates.
Ehri, L. (1994). Development of the ability to read words: Update. In R. Ruddell, M. Ruddell, & H. Singer (Eds.), Theoretical
models and processes of reading (4th ed., pp. 323-358). Newark, DE: International Reading Association.
Ehri, L. (1995). Phases of development in learning to read words by sight. Journal of Research in Reading, 18, 116-125.
Ehri, L., & Robbins, C. (1992). Beginners need some decoding skill to read words by analogy. Reading Research Quarterly, 27,
12-26.
Ehri, L., & Roberts, K. (1979). Do beginners learn printed words better in contexts or in isolation? Child Development, 50, 675-
685.
Ehri, L., & Saltmarsh, J. (1995). Beginning readers outperform older disabled readers in learning to read words by sight.
Reading and Writing: An Interdisciplinary Journal, 7, 295-326.
Ehri, L., & Wilce, L. (1979). The mnemonic value of orthography among beginning readers. Journal of Educational
Ehri, L., & Wilce, L. (1980). The influence of orthography on readers' conceptualization of the phonemic structure of words.
Ehri, L., & Wilce, L. (1982a). Recognition of spellings printed in lower and mixed case: Evidence for orthographic images.
Journal of Reading Behavior, 14, 219-230.
Ehri, L., & Wilce, L. (1982b). The salience of silent letters in children's memory for word spellings. Memory and Cognition, 10,
155-166.
Ehri, L., & Wilce, L. (1985). Movement into reading: Is the first stage of printed word learning visual or phonetic? Reading
Research Quarterly, 20, 163-179.
Ehri, L., & Wilce, L. (1986). The influence of spellings on speech: Are alveolar flaps /d/ or /t/? In D. Yaden & S. Templeton
(Eds.), Metalinguistic awareness and beginning literacy (pp. 101-114). Portsmouth, NH: Heinemann.
Ehri, L., & Wilce, L. (1987a). Cipher versus cue reading: An experiment in decoding acquisition. Journal of Educational

Page 267
Ehri, L., & Wilce, L. (1987b). Does learning to spell help beginners learn to read words? Reading Research Quarterly, 22, 47-
65.
Foorman, B., Francis, D., Novy, D., & Liberman, D. (1991). How letter-sound instruction mediates progress in first-grade
reading and spelling. Journal of Educational Psychology, 83, 456-469.
Academic Press.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, J. Marshall, & M. Coltheart (Eds.), Surface
dyslexia: Neuropsychological and cognitive studies of phonological reading (pp. 301-330). London: Lawrence Erlbaum
Associates.
Fry, E. (1994). Phonics patterns: Onset and rhyme word lists. Laguna Beach, CA: Laguna Beach Educational Books.
Fry, E., Polk, J., & Fountoukidis, D. (1984). The reading teacher's book of lists. Englewood Cliffs, NJ: Prentice-Hall.
Gentry, J. (1982). An analysis of developmental spelling in GNYS AT WRK. Reading Teacher, 36, 192-200.
Goswami, U. (1986). Children's use of analogy in learning to read: A developmental study. Journal of Experimental Child
Goswami, U. (1988a). Children's use of analogy in learning to spell. British Journal of Developmental Psychology, 6, 21-33.
Goswami, U. (1988b). Orthographic analogies and reading development. Quarterly Journal of Experimental Psychology, 40,
239-268.
Gough, P., & Hillinger, M. (1980). Learning to read: An unnatural act. Bulletin of the Orton Society, 30, 180-196.
Gough, P., Juel, C., & Griffith, P. (1992). Reading, spelling and the orthographic cipher. In P. Gough, L. Ehri, & R. Treiman
(Eds.), Reading acquisition (pp. 35-48). Hillsdale, NJ: Lawrence Erlbaum Associates.
Greenberg, D., Ehri, L., & Perin, D. (in press). Are word reading processes the same or different in adult literacy students and
3rd-5th graders matched for reading level? Journal of Educational Psychology.
Griffith, P. (1987). The role of phonological and lexical information in word recognition and spelling. Unpublished doctoral
dissertation, University of Texas, Austin.
Griffith, P. (1991). Phonemic awareness helps first graders invent spellings and third graders remember correct spellings.
Journal of Reading Behavior, 23, 215-233.
Guthrie, J. (1973). Models of reading and reading disability. Journal of Educational Psychology, 65, 9-18.
Hanna, P., Hanna, J., Hodges, R., & Rudorf, E. (1966). Phoneme-grapheme correspondences as cues to spelling improvement.
Washington, DC: U.S. Government Printing Office.
Harste, J., Woodward, V., & Burke, C. (1984). Language stories and literacy lessons. Portsmouth, NH: Heinemann.
Henderson, E. (1981). Learning to read and spell: The child's knowledge of words. DeKalb, IL: Northern Illinois University
Press.
Henry, M. (1988). Beyond phonics: Integrated decoding and spelling instruction based on word origin and structure. Annals of
Dyslexia, 38, 258-275.
Henry, M. (1990). Words. Los Gatos, CA: Lex Press.
Hohn, W., & Ehri, L. (1983). Do alphabet letters help prereaders acquire phonemic segmentation skill? Journal of Educational
Jorm, A. (1981). Children with reading and spelling retardation: Functioning of whole-word and correspondence-rule
mechanisms. Journal of Child Psychology and Psychiatry, 22, 171-178.
Juel, C., Griffith, P., & Gough, P. (1986). Acquisition of literacy: A longitudinal study of children in first and second grade.
Journal of Educational Psychology, 78, 243-255.
Kreiner, D., & Gough, P. (1990). Two ideas about spelling: Rules and word-specific memory. Journal of Memory and
Language, 29, 103-118.

Page 268
Laxon, V., Coltheart, V., & Keating, C. (1988). Children find friendly words friendly too: Words with many orthographic
neighbours are easier to read and spell. British Journal of Educational Psychology, 58, 103-119.
Liberman, I., & Shankweiler, D. (1979). Speech, the alphabet, and teaching to read. In L. Resnick & P. Weaver (Eds.), Theory
and practice of early reading (vol. 2, pp. 109-132). Hillsdale, NJ: Lawrence Erlbaum Associates.
Lundberg, I., Frost, J., & Peterson, O. (1988). Effects of an extensive program for stimulating phonological awareness in
preschool children. Reading Research Quarterly, 23, 263-284.
Mason, J. (1980). When do children begin to read: An exploration of 4-year-old children's letter and word-reading
competencies. Reading Research Quarterly, 15, 203-227.
Masonheimer, P., Drum, P., & Ehri, L. (1984). Does environmental print identification lead children into word reading? Journal
of Reading Behavior, 16, 257-272.
McConkie, G., & Zola, D. (1981). Language constraints and the functional stimulus in reading. In A. Lesgold & C. Perfetti
(Eds.), Interactive processes in reading (pp. 155-175). Hillsdale, NJ: Lawrence Erlbaum Associates.
Morris, D., & Perney, J. (1984). Developmental spelling as a predictor of first-grade reading achievement. Elementary School
Journal, 84, 441-457.
Newman, S., Fields, H., & Wright, S. (1993). A developmental study of specific spelling disability. British Journal of
Educational Psychology, 63, 287-296.
Olson, R., Kliegl, R., Davidson, B., & Foltz, G. (1985). Individual and developmental differences in reading disability. In G.
MacKinnon & T. Waller (Eds.), Reading research: Advances in theory and practice (vol. 4, pp. 1-64). New York: Academic
Press.
Perfetti, C. (1992). The representation problem in reading acquisition. In P. Gough, L. Ehri, & R. Treiman (Eds.), Reading
acquisition (pp. 107-143). Hillsdale, NJ: Lawrence Erlbaum Associates.
Rack, J., Hulme, C., Snowling, M., & Wightman, J. (1994). The role of phonology in young children learning to read words:
The direct-mapping hypothesis. Journal of Experimental Child Psychology, 57, 42-71.
Rack, J., Snowling, M., & Olson, R. (1992). The nonword reading deficit in developmental dyslexia: A review. Reading
Read, C. (1975). Children's categorization of speech sounds in English. Urbana, IL: National Council of Teachers of English,
Research Report No. 17.
Reitsma, P. (1983). Printed word learning in beginning readers. Journal of Experimental Child Psychology, 75, 321-339.
Roberts, K., & Ehri, L. (1983). Effects of two types of letter rehearsal on skilled and less skilled beginning readers' word
memory. Contemporary Educational Psychology, 8, 375-390.
Scott, J., & Ehri, L. (1989). Sight-word reading in prereaders: Use of logographic versus alphabetic access routes. Journal of
Reading Behavior, 22, 149-166.
Share, D., Jorm, A., Maclean, R., & Matthews, R. (1984). Sources of individual diferences in reading achievement. Journal of
Educational Psychology, 76, 1309-1324.
Simon, D. (1976). Spelling: A task analysis. Instructional Science, 5, 277-302.
Simon, D., & Simon, H. (1973). Alternative uses of phonemic information in spelling. Review of Educational Research, 43, 115-
137.
Sloboda, J. (1980). Visual imagery and individual differences in spelling. In U. Frith (Ed.), Cognitive processes in spelling (pp.
231-248). London: Academic Press.
Snowling, M. (1994). Towards a model of spelling acquisition: The development of some component skills. In Handbook of
spelling: Theory, process, and intervention (pp. 111-128). Chichester: Wiley.
Stanovich, K. (1980). Toward an interactive-compensatory model of individual differences in the development of reading
fluency. Reading Research Quarterly, 16, 32-71.

Page 269
Templeton, S., & Bear, D. (Eds.). (1992). Development of orthographic knowledge and the foundations of literacy: A memorial
festschrift for Edmund H. Henderson. Hillsdale, NJ: Lawrence Erlbaum Associates.
Treiman, R. (1985). Onsets and rimes as units of spoken syllables: Evidence from children. Journal of Experimental Child
Treiman, R. (1993). Beginning to spell. New York: Oxford University Press.
Treiman, R., Goswami, U., & Bruck, M. (1990). Not all nonwords are alike: Implications for reading development and theory.
Memory and Cognition, 18, 559-567.
Uhry, J., & Shepherd, J. (1993). Segmentation/spelling instruction as a part of a first-grade reading program: Effects on several
measures of reading. Reading Research Quarterly, 28, 218-233.
Varnhagan, C. (1995). Children's spelling strategies. In V. W. Berninger (Ed.), The varieties of orthographic knowledge II:
Relationships to phonology, reading, and writing (pp. 251-290). Dordrecht, The Netherlands: Kluwer.
Venezky, R. (1970). The structure of English orthography. The Hague: Mouton.
Venezky, R., & Johnson, D. (1973). Development of two-letter sound patterns in Grades 1 through 3. Journal of Educational
Wagner, R., & Barker, T. (1994). The development of orthographic-processing ability. In V. Berninger (Ed.), The varieties of
orthographic knowledge I: Theoretical and developmental issues (pp. 243-276). Dordrecht, The Netherlands: Kluwer.
Waters, G., Bruck, M., & Malus-Abramowitz, M. (1988). The role of linguistic and visual information in spelling: A
developmental study. Journal of Experimental Child Psychology, 45, 400-421.
Woodcock Reading Mastery Tests-Revised. (1987). Circle Pines, MN: American Guidance Service.
Wylie, R., & Durrell, D. (1970). Teaching vowels through phonograms. Elementary English, 47, 787-791.


Page 271
Chapter 14
Interactions in the Development of Reading and Spelling: Stages, Strategies, and Exchange of Knowledge1
Nick Ellis
University of Wales
Introduction
A number of models of the development of spelling have been progressively refined over the last 15 years (Ehri, 1986; Gentry,
1978, 1982; Henderson & Beers, 1980; Morris, 1983). These share the following commonalities: (a) They are based on analyses
of spelling errors when children attempt to spell novel words (invented spellings); (b) they are stage theories, proposing that
qualitatively different cognitive processes are involved in children's spelling at different points in development and that there is a
characteristic progression from stage to stage; (c) they emphasize that phonological awareness plays a crucial role in children's
early spelling but, also, that children eventually acquire orthographic descriptions of words. These models have been developed
in parallel with cognitive developmental stage theories of reading acquisition (see Marsh, Friedman, Welch, & Desberg, 1980,
1981) that are also based on error analysis, which also hold that there are very different strategies of information processing used
in reading at different stages of its development, and which also emphasize the links between phonological awareness and
reading development. Although synchronous, paradoxically these theoretical developments concerning reading and spelling were
essentially independent until Ehri (1979, 1984) showed that the improvements
1 Parts of this chapter appeared in G. D. A. Brown & N. C. Ellis (1994a) and are reprinted with permission of the
publisher.

Page 272
in phonological awareness, which are based on the acquisition of alphabetic reading, are themselves a consequence of learning
how sound segments in words are spelled conventionally. Hence, Frith (1985) and Ehri (1986) proposed models of literacy
development where reading and writing mutually influence and grow from each other.
The purpose of this chapter is to briefly summarize the modal aspects of stage theories of spelling development, reading
development, and Frith's integrative model, and then to see how well these theories have fared in empirical tests from several
longitudinal studies of development that have been prosecuted in recent years.
Stages of Spelling Development

The idea that children's misspellings reflect a developing sense of phonetic properties in words was pioneered by Read (1971,
1975, 1986) who found evidence that young inventive spellers used a system of grouping sounds together according to shared
phonetic features. Thus, they might represent a particular vowel sound in their spelling by substituting a letter whose name
shared a salient phonetic feature with the sound. Read's exhaustive studies of invented spellings demonstrated that children use
processes of both speech production and perception to group sounds together and that these categorizations may not coincide
with the classification system used by adults:
We now value spellings for what they can tell us about psycholinguistic processes. Standard spellings are of less interest,
not because they represent successful instruction, but because they do not indicate how a child arrived at them. . . Some
nonstandard spellings represent a more advanced conception of the task or the language than others. (Read, 1986, p. 47)
This idea that spelling errors provide an index of children's metalinguistic understanding of language has allowed subsequent
researchers to categorize developmental strategies in spelling. Henderson and Beers (1980) analyzed samples of children's
creative writing and assigned each error to a category according to the completeness of phonetic information mapped by the
misspelling. On the basis of their work and that of Bissex (1980) and Gentry (1978, 1982), it is now generally agreed that
children move through five distinct stages of spelling, namely, precommunicative, semiphonetic, phonetic, transitional, and
correct spelling. Precommunicative spellings are characterized by the strategy of randomly selecting letter strings to represent
words (e.g., spelling monster as BTRSS, or chirp as 1MMPMPH). Although at this stage children can produce letters in writing,
their spellings reflect

Page 273
a complete lack of letter-sound or letter-name knowledge. Semiphonetic spellings contain a partial mapping of phonetic content.
At this stage (a) the speller begins to conceptualize that letters have sounds that are used to represent the sounds in words; (b)
the letters used to represent words provide a partial but not total mapping of the phonetic representation of the word being
spelled; and (c) a letter name strategy is very much in evidence where possible the speller represents words, sounds, or syllables
with letters that match their letter names (e.g., R for are; U for you; LEFT for elephant). At the phonetic stage, (a) phonological
segmentation of spoken words is usually evident and, as a result; (b) spellings contain a complete description of the sequence of
sounds in pronunciations; (c) all of the surface sound features of the words are represented in the spelling; (d) children
systematically show particular spellings for certain details of phonetic form such as tense vowels, lax vowels, preconsonantal
nasals, syllabic sonorants, -ed endings, etc.; but (e) letters are assigned strictly on the basis of sound without regard for
acceptable letter sequence or other conventions of English orthography (e.g., IFU LEV AT THRD STRET IWEL KOM TO
YOR HAWS THE ED for "If you live at Third Street I will come to your house. The End." [Bissex, 1980, p. 13]). In the
transitional stage, the child begins to adhere to more basic conventions of English orthography: Vowels appear in every syllable
(e.g., EGUL instead of the phonetic EGL for eagle); nasals are represented before consonants (e.g., BANGK instead of the
phonetic BAK for bank); both vowels and consonants are employed instead of a letter-name strategy (e.g., EL rather than L for
the first syllable of ELEFANT for elephant); common English letter sequences are used in spelling (e.g., YOUNITED for
united), especially liberal use of vowel digraphs such as ai, ea, ay, ee; silent e pattern becomes fixed as an alternative for
spelling long vowel sounds (e.g. TIPE in place of the phonetic TIP for type), etc. Transitional spellers present the first evidence
of a new orthographic strategy, moving from phonological to morphological and orthographic spelling (e.g., EIGHTEE instead
of the phonetic ATE for eighty), but they have not fully developed their knowledge of environmental factors such as the
graphemic environment of the unit, position in the word, stress, morpheme boundaries, and phonological influences. Acquisition
of this knowledge, along with an extended knowledge of word structure (e.g., prefixes, suffixes, contractions, and compound
words), an increased accuracy in using silent consonants and in doubling consonants, and simply in knowing when words "just
don't look right" (i.e., in having a complete visual orthographic description of them) are elements that allow the mastery of
correct spelling.
These descriptive categories for this developmental sequence are quite fine grained. They can be more coarsely summarized as a
shift from Stage I of early attempts (which may be precommunicative or based on visual copying of whole word or symbol
patterns [symbolic or logographic], or

Page 274
very rudimentary prephonetic attempts representing perhaps just the first sound of a word), through Stage II reflecting varying
degrees of mastery of the alphabetic principle, to Stage III of correct orthographic or morphemic spelling (Ehri, 1986; Frith,
1985; Gentry, 1982).
Stages of Reading Development

Analyses of children's early reading errors (Biemiller, 1970; Torrey, 1979; Weber, 1970) led Marsh, Friedman, Welch, and
Desberg (1981) to propose that the first stage of reading could be characterized as one of linguistic substitution when the child
uses a strategy of rote association between a simple, unanalyzed visual stimulus and an oral response. "The child typically
centres on one aspect of the visual stimulus such as the first letter and associates that with the oral response . . . Their natural
strategy is congruent with the `whole word' approach to teaching reading" (pp. 201-202). Frith (1985) called this the logographic
stage. If the child does not know the word, she may guess on the basis of contextual cues.
Marsh et al. (1980, 1981) suggested that the child next progresses to discrimination net substitution in which "the number of
graphemic features a young child can process is limited initially to the first letter, and it is only later that additional features such
as word length, final letter, etc. are added. The child at this stage appears to be operating according to a `discrimination net'
mechanism in which graphemic cues are processed only to the extent necessary to discriminate one printed word from another."
(p. 203).
Marsh et al. (1980, 1981) called the next stage that of sequential decoding, Frith (1985) termed it alphabetic, Gough and
Hillinger (1980) deciphering, Harris and Coltheart (1986) phonological recoding. In all of these models, this is characterized by
the use of individual graphemes and phonemes and their correspondences. "It is an analytic skill involving a systematic
approach, namely decoding grapheme by grapheme. Letter order and phonological factors play a crucial role. This strategy
enables the reader to pronounce novel and nonsense words" (Frith, p. 308).
Both models hold that the final stages of skilled reading take place by the use of orthographic strategies. Marsh et al. (1980,
1981) characterized this as being an extension to the simple decoding strategy (which was based on one-to-one correspondence)
where the child now learns more complex rules of orthographic structure the units are letter groups, and higher order conditional
rules (such as the magic e rule) are used. Frith (1985), however, suggested that skilled reading involves orthographic strategies
where the words are instantly analyzed into orthographic units without phonological conversion: "The orthographic units ideally
coincide with morphemes. They are internally represented as abstract letter-by-letter strings. These units

Page 275
make up a limited set that in loose analogy to a syllabary can be used to create by recombination an almost unlimited number of
words.'' (Frith, p. 308). Both the models of Frith and Marsh et al. emphasize analysis of multiple-letter orthographic units, but
Frith is implying that practice at the analysis of orthographic sequences will eventually allow nonphonological, whole-
morpheme, direct lexical access, with postlexical, phonological retrieval.
Interactions in the Development of Reading and Spelling

The importance of Frith's (1985) model is that it provides a theoretical framework within which spelling and reading interact to
advance the learner toward increased proficiency in each ability. It is a developmental model in which reading and spelling both
progress through stages of logographic, then alphabetic and, finally, orthographic strategies. However, her model does more
than this in that it suggests reasons for the move from one stage to the next. The crux of her argument is that normal reading and
spelling development proceed out of step and that the adoption and use of a strategy in one domain may serve as a pacemaker
for development of that strategy in the other. This is illustrated in Fig. 14.1, which shows the points at which the domains are
misaligned and the cross-domain influences that are suggested to occur.
Claim 1. Frith (1985) suggested that the beginnings of literacy lie in logographic reading when the child has a finite whole-word
reading vocabulary and that the development of this means of reading results in its adoption as a strategy for spelling. Thus,
Claim 1 is that logographic reading is the pacemaker for the use of a logographic strategy in spelling. Evidence for a
logographic stage of reading comes from Seymour and Elder (1986) who showed that 5-year-old beginning readers, who were
taught by a predominantly look-and-say method, were able to read aloud words that they had been taught but, nevertheless,
possessed no usable procedures for pronouncing new words. However, evidence for logographic reading driving the
development of a logographic strategy of spelling is harder to come by. There is little doubt that children are first brought to
realize the importance of the written word by seeing them, but several investigators, most notably Goswami and Bryant (1990)
seriously questioned the existence of a logographic stage of spelling.
Claim 2. At some points of development, reading and spelling are paradoxically out of synchronization. As Goswami and
Bryant (1990, p. 148) put it, ''It is still not clear why children are so willing to break up words into

Page 276
Fig. 14.1.
Firth's (1985) model of reading and spelling acquisition (strategies acting as
"pacemakers" at each step are italicized).
phonemes when they write, and yet are so reluctant to think in terms of phonemes when they read. But there can be little doubt
that at first children's reading and spelling are different and separate. The most dramatic demonstration of this separation is the
fact that young children often cannot read some words which they know how to spell and also fail to spell some words which
they can read." Perhaps the major tenet of Frith's (1985) model, as with that of Ehri (1986), is that children first gain explicit
insight into the alphabetic code through practice at spelling and that this causes a shift from a logographic reading strategy to an
alphabetic approach. Claim 2 is that alphabetic spelling is the pacemaker for the use of an alphabetic strategy in reading. This
can be broken down into two components: Claim 2a is that phonological awareness is much more related to early spelling than
it is to early reading; Claim 2b is that this mutually supportive growth of phonological awareness and spelling acts as a
pacemaker for the adoption of an alphabetic strategy of reading.

Page 277
The flesh of the argument of Claim 2 is as follows: It assumes that phonological awareness is important in the development of
alphabetic reading (Bradley & Bryant, 1983; Ellis & Large, 1987, 1988), but it is when first attempting to spell rather than to
read that the child realizes that a phonological strategy is useful in breaking up print-speech correspondences (Bryant & Bradley,
1980; Frith, 1981; Smith, 1978; Snowling & Perin, 1983). Early spelling practice typically involves dividing spoken words into
phonemes and representing these phonemes with letters. In this way, experience in spelling words affords the opportunity for
making comparisons between the phonetic information in individual letters and sounds as they are embedded in the spoken
word. Spelling practice helps to establish this abstract concept through two very concrete means articulatory and kinesthetic
rehearsal and by cross-modal correspondences so that orthographic units guide parsing breaks and provide labels for the sound
chunks produced by the novice segmenter. (Indeed, when considering phonology, we cannot ignore our knowledge of
orthography once it has been acquired, as is evidenced by fourth graders and above, who think that pitch has four phonemes and
rich only three whereas, in both of these words, a phonetic element corresponding to the extra letter t is present in articulation;
Ehri & Wilce, 1985). Through repeated practice in spelling, the child may come to appreciate the subtle relationship between a
symbol in the written word and its corresponding sound in the context of the spoken word. The discovery of this relationship is
the key to alphabetic insight. The crux of the problem is "knowing how to combine the letters into units appropriate for speech"
(Liberman & Shankweiler, 1979, p. 141). As children struggle to decompose words into individual phonemic units, they
commonly experiment with various articulatory rehearsals of word parts, and they search for distinguishable articulatory units
that correspond to letter-sound units. This process of their separating sounds in a word through consciously monitoring their
own articulations may serve a dual purpose: It may help the development of phonological awareness, and it may enhance
knowledge of the alphabetic principle. As children refine their ability to detect and isolate the sound content of spoken words
through repeated practice in spelling, they build a store of knowledge about the relationships among sounds, letters, and
pronunciations that can be applied to the task of reading (Chomsky, 1977; Ehri, 1986).
Claim 3. Finally, both Frith (1985) and Ehri (1986) supposed that considerable practice at reading by means of an alphabetic
strategy encourages sufficient analysis of letter sequences in words to allow the reader to develop internal representations that
are well specified in terms of letter-by-letter detail. These orthographic representations acquired through reading are then exact
enough to be transferred to spelling and to constitute the knowl-

Page 278
edge that allows the shift from phonetic through transitional to correct spelling described previously. Thus, Claim 3 is that
reading is the pacemaker for the development of orthographic spelling.
Frith's (1985) model is also important in that it explains different disorders of developmental literacy in terms of arrest at
different stages of development. Because the present purpose is to assess it as an explanation of normal literacy development,
these other claims are not addressed further here.
If we want to study development, then we must do so directly. Only when the same persons are tested repeatedly over time does
it become possible to identify developmental changes and processes of organization within the individual. Cross-sectional
studies that compare different groups of people at different points of acquisition must always come in a poor second when small
but reliable changes with age are to be detected, when teaching methods and teachers change with time, and when we do not
wish to make the artificial assumption that the abilities of a younger cross-section were present in the older cross-section at a
previous time (Ellis & Large, 1987). For these reasons, we only weigh the model against the two types of research paradigm
that allow some degree of causal interpretation: longitudinal studies and experimental investigations of the effects of training.
Tests from Longitudinal Studies of Developmental Phonological Awareness and Alphabetic Reading
There have now been a number of longitudinal demonstrations that early phonological awareness (PA) predicts later reading
achievement even when prior IQ is controlled (Bradley & Bryant, 1985; Ellis & Large, 1987; Lundberg, Olofsson, & Wall,
1980; Mann, 1984; Stanovich, Cunningham, & Cramer, 1984; see Goswami & Bryant, 1990, and Wagner & Torgesen, 1987, for
reviews). It is clear, therefore, that there is a causal developmental sequence whereby alphabetic reading capitalizes on prior
phonological abilities.
Reciprocal influences are also evident, as metaphonological skills are made relevant and are practiced in alphabetic reading so
they themselves are enriched (see Morais, Alegria, & Content, 1987, for review). For example: (a) Portuguese adult illiterates
who had never attended school for social reasons scored only 19% correct on tests of phoneme addition and deletion, whereas
matched participants who had been taught to read and write in special classes in adulthood scored at over 70% levels. Thus
Morais, Cary, Alegria, and Bertelson (1979) concluded that the ability to deal explicitly with the segmental units of speech is not
acquired spontaneously in the course of cognitive growth but demands some specific training, which, for most persons, is
provided by learning to read and write; (b) Alegria, Pignot,

Page 279
and Morais (1982) showed that 6-year-old children, trained to read using phonic methods, were much better (58%) at phonemic
segmentation than whole-word trained readers (15%); (c) Chinese literates who were taught to read pinyin, an alphabetic script,
scored 83% on a test of segmental analysis, whereas nonalphabetic literates who could read logographs only scored 21% (Read,
Zhang, Nie, & Ding, 1986)-phonological segmentation is, to some considerable extent, a consequence of alphabetic literacy; (d)
In a longitudinal investigation of 40 children learning to read English, Ellis and Large (1988) found that reading ability at 5- and
6-years-old better predicted phoneme segmentation skill one year later than phoneme segmentation predicts later reading; and
(e) Perfetti, Beck, Bell, and Hughes (1987) used time-lag, partial correlation analyses of first-grade children's development to
demonstrate that phoneme blending predicted later reading ability and this, in turn, predicted later proficiency at phoneme
segmentation. They suggested that young children come to possess a basic, primitive awareness of the sounds of language on
which alphabetic reading capitalizes. However, learning to read fosters attention to constituent principles of words, and this
results in the development of sophisticated segment analysis ability. As this is acquired, so there are further gains in reading
itself.
Ellis and Cataldo (1990) put a similar case concerning different aspects of phonological awareness and the differential
interactions of these with developing reading. Whereas early research into the relationship between reading and phonological
awareness did not discriminate between different types of phonemic awareness tasks (Bradley & Bryant, 1985; Lewkowicz,
1980), more recent work suggests that the level of phonemic awareness demanded by the phonological tasks influences the
strength of the relationship between reading and phoneme awareness (Backman, 1983). Stanovich, Cunningham, and Cramer
(1984) asked children to perform tasks involving the analysis of words for explicit sound content (nonrhyming tasks) and for the
perception of overall similarity of sound content (rhyming tasks). They found that the nonrhyming, or analytic, and productive
phonological tasks formed a cluster of related skills and that the rhyming tasks did not correlate strongly with the nonrhyming
tasks. Snowling and Perin (1983) found that children's ability to perform a segmentation task was not significantly different
from their ability in spelling, the close connection between these skills indicating the necessity of explicit PA in spelling. Thus,
there seem to be two developmentally different, measurable levels of PA.
Children's first awareness of the sound properties of speech is implicit and perceptual. Spontaneous play with rhyming and
nonsense words is thought to reflect an overall sensitivity to the sound content of words (Chukovsky, 1968; Clark, 1978; Slobin,
1978). At this point, they are not yet able to consciously reflect on language (Andresen cited in Valtin, 1984; Shankweiler,
Liberman, & Savin, 1972). Valtin (1984) described a 3-stage

Page 280
model for the development of phonological awareness. Initially the child is not aware of the sound value of speech; he or she
senses when, but not why, speech acts fail to be communicative. During the next stage, "children become increasingly able to
abstract the language from the action and the meaning context and to think about some of the properties of the form of
language. Their knowledge of language units is still implicit, however, and related to psycholinguistic units of speech" (Valtin,
p. 214). Once the child achieves conscious awareness, he or she demonstrates explicit phonological awareness and can reflect
on, produce, and manipulate phonemic units within spoken words. MacLean, Bryant, and Bradley (1987) demonstrated in a
longitudinal study that children's early experience of nursery rhymes results in their gaining initial implicit phonological
awareness. Bradley and Bryant (1983) and Ellis and Large (1987) reported that very early reading capitalizes on this implicit
phonological awareness. Finally, the aforementioned studies, particularly Perfetti et al. (1987), found that experience in
alphabetic reading promotes more analytic and explicit phonological awareness at subsyllabic levels.
Stanovich (Cunningham & Stanovich, 1990, 1993; Stanovich & West, 1989) presented a body of evidence that appears,
interestingly, to contradict the previous conclusions. They used tests of recognition of famous authors and titles to assess the
amount to which individuals have been exposed to print and, by inference, the amount of reading that they have done. Although,
as is described later, print exposure is highly predictive of orthographic knowledge and vocabulary even when intelligence is
controlled, it does not reliably predict phonological processing abilities (e.g., phoneme deletion or transposition) either in
children (Cunningham & Stanovich, 1990, 1993) or in adults (Stanovich & West, 1989). These disparate results do not seem
attributable to measurement instruments because they use similar sorts of phoneme segmentation tasks to those in the
aforementioned studies. An alternative explanation is that the amount of exposure to the written language per se is not the
relevant variable. Rather, in normal (as opposed to dyslexic) children, refined explicit phonological awareness at the level of the
phoneme is acquired as a result of alphabetic literacy skill that, in turn, is engendered by an introduction to grapheme-phoneme
correspondences in spelling.
This leads us to the particular question of Claim 2: What is the role of spelling in the coming together of phonological skills and
reading?
Alphabetic Spelling, Phonological Awareness, and Reading

There is now a useful collection of longitudinal studies that address the development of phonological awareness, reading, and
spelling. Unfortunately, although they are all informative, they also all have their problems. This

Page 281
section briefly describes some representative studies and weighs Claim 2 against their evidence.
Lundberg et al. (1980) presented the first longitudinal study in which phoneme segmentation and synthesis skills measured
during kindergarten in 143 Swedish children were used as predictors of reading and spelling development at the end of Grade 1
and the beginning of Grade 2. There were highly significant correlations between the PA measures and later reading and
spelling, demonstrating the extreme importance of PA in literacy acquisition. Unfortunately, the study was marred by too easy a
spelling test at Grade 1, resulting in ceiling effects on this variable and lack of discrimination. It is difficult, therefore, to assess
whether PA is more involved in early spelling than in reading (Claim 2a) and impossible to look at causal interactions between
early spelling and later reading (Claim 2b).
Tornus (1984) investigated the causal relationships between intelligence (IQ), general language development (L), phonological
awareness (PA: sound blending and segmentation), reading (R), and spelling (S) in 46 dyslexic and 44 control Swedish children
at the end of Grade 1 (IQ1, L1, PA1, R1, S1) and at the beginning (IQ2, R2, S2) and middle of Grade 2 (S3). A number of
causal models of interactions in their development were tested using LISREL. There was a very strong causal path from PA1 to
spelling (as a latent factor that represented ability over tests S1-S3), but only a slight influence of general linguistic and
cognitive development on spelling. The causal model for reading was similar, with the exception that it also included a direct
causal path from cognitive development. Moreover, it was clear from Tornus that the associations between various aspects of
PA and spelling at all grades measured (12 correlations ranging from .51 to .73) were considerably stronger than those between
PA and reading (8 correlations ranging from .33 to .53), a set of findings consistent with Frith's (1985) Claim 2a that
phonological awareness is initially much more involved in spelling than in reading. Unfortunately, there were no models that
looked for interactions in effect from reading to spelling, and vice versa, so it is difficult to assess Frith's (1985) Claim 2b about
pacemakers from this study (the failure to obtain an R3 measure also limits these possibilities here). Tornus did provide tests of
reciprocal models whereby PA could affect S (or R) and S (R) could, in turn, affect PA. These suggested much more of a causal
role in development from PA to reading and spelling than the reverse. However, there are grave problems with these analyses:
(a) Solutions to such reciprocal path models are notoriously unstable in causal path analysis; (b) PA was measured at Time 1
only, the composite S variable was measured at Times 1, 2, and 3; and the composite R variable at Times 1 and 2. Therefore,
the models that were fitted were looking for causal affects of, for example, spelling midgrade 2 on Grade 1 PA. Because neither
time nor development run backwards, it is not surprising that there were low beta-weights on these

Page 282
paths. Against Tornus' claims, these models are not fair tests of, for example, Ehri's (1979) claim that phonological awareness
arises from acquaintance with orthography gained from practice in spelling.
Juel, Griffith, and Gough (1986) assessed a well-motivated range of variables including IQ, listening comprehension, PA
(phoneme segmentation and phoneme substitution), exposure to print, nonword reading, spelling recognition and production,
reading comprehension, and writing in more than 100 children over the first two grades of their schooling in Texas.
Unfortunately, they analyzed the data with a series of cross-sectional rather than time-lag models, and reanalysis is impossible
because they do not provide time-lag correlations. Their cross-sectional data do replicate high correlations between PA and both
reading and spelling. However, these correlations were of similar magnitude in both grades (Juel et al., 1986), and there is no
evidence from their study that explicit PA is more related to Grade 1 spelling than to reading (Claim 2a). The lack of cross-
lagged correlations precludes a test of Claim 2b.
Mommers (1987; Mommers, van Leeuwe, Oud, & Janssens, 1986) described a longitudinal investigation of the first 3 years of
development of PA, reading, and spelling in approximately 500 Dutch children. The study is admirable in that (a) word
identification, spelling, and reading comprehension were measured in parallel at least at 5 separate points in development, and
(b) the investigators analyzed the results with proper longitudinal causal path models. Unfortunately, (a) PA was only measured
at the first point, and so reciprocal effects of reading and spelling on PA growth cannot be assessed; (b) as in many of the
previous studies, the reading and spelling tests included both regular and irregular words in terms of grapheme-phoneme
correspondence, and so we cannot disentangle alphabetic and orthographic strategies; (c) such complex data limit the authors to
reporting only the final models resulting from many iterative stages of model refinement. Even so, there is some tentative
support for aspects of Frith's (1985) model. In the first place, there were strong effects of initial PA on immediately subsequent
single-word reading and spelling abilities even though there was no evidence of a greater contribution on spelling (Claim 2a). In
the course of model refinement, the investigators found it necessary to fit a path at the second point of measurement (4 months
after the start of formal reading instruction) between spelling ability and word reading (SPo to DSo in their Fig. 5), "a much
stronger one than in the reverse direction" (Mommers, p. 136). As this influence of spelling on reading is unique to this early
stage of literacy development, this finding is highly supportive of Frith's (1985) suggestion of the influence of spelling on
alphabetic reading (Claim 2b). At all subsequent stages, the best fitting causal models had significant, although not large,
influences in the reverse direction, that is, from word reading to spelling: "There also exists an influence of decoding skill on
spelling. In both

Page 283
decoding skill and spelling, orthographic representations stored in the lexicon play a part . . . The repeated reading of words can
only to some extent improve the quality of the orthographic representations." (Mommers, p. 140). These paths, although small,
are supportive of Claim 3.
Cataldo and Ellis (1988; Ellis & Cataldo, 1990) charted the development of reading, spelling, and phonological awareness in a
group of 28 children during their first 3 years in school. During this time, the children were tested at four intervals in reading
and spelling real words and nonsense words, phoneme segmentation, and auditory categorization. The Wechsler Preschool
Primary Scale of Intelligence (WPPSI) was included in the set of initial assessments. A test of phoneme segmentation was given
as a measure of explicit phonemic awareness, and a test of auditory categorization was taken as a measure of implicit
phonological awareness. The majority of the sample had only begun to attend school when the initial assessments were taken at
the beginning of the school year when their mean age was 4 years 6 months. The children were retested at the end of their first
school year, at the beginning of the second year, and at the beginning of the third school year. Exploratory (LISREL) causal path
analyses were used to investigate the contribution of each ability to the subsequent growth of skill in word recognition, spelling,
and phonological awareness over three measured phrases of development. Phase One spanned the children's first year in school.
Phase Two charted the development from Spring of the first school year to Autumn of the second year. Phase Three looked at
development from the beginning of the second year in school to the beginning of the third year. The Phase 1 pathweights from
spelling to reading real words (0.31) and nonsense words (0.23) identified spelling as an important contributor to the early
formation of reading. This pattern of influence was repeated in the second phase (spelling to reading real words 0.64 and
nonsense words 0.60). The pronounced influence of spelling on reading contrasted with the meager contributions of reading to
spelling (Phase 1: 0.10 real words, 0. 06 nonsense; Phase 2: 0.14 real, 0.00 nonsense), thus confirming Claim 2b. Implicit PA
initially predicted early attempts to read (0.36, 0.41) as well as to spell (0.38, 0.31) but lost its influence on both reading and
spelling in the following two phases. In contrast to the diminishing predictive power of implicit PA, explicit PA consistently
predicted spelling in all three phases (confirming Claim 2a), this influence increasing with phase. Explicit PA only emerged as a
strong predictor of reading in Phase Three. To summarize the Ellis and Cataldo results, the early flow of information between
reading and spelling appeared to be asymmetrical: Knowledge gleaned from spelling contributed to reading. Similarly, both
implicit and explicit PA affected spelling development, with explicit PA increasing its influence as the contribution of implicit
PA diminished. Only later in the developmental sequence did explicit PA begin to contribute directly to reading.

Page 284
Berninger, Abbott, and Shurtleff (1990) tracked visual language processes (1-second delayed visual recognition memory for a
word or for a letter in a word), oral language processes (vocabulary and phoneme segmentation and deletion), reading (word
naming and lexical decision), and spelling (written reproduction subsequent to seeing the word) in 42 children during their first
grade of school in the United States. Written and oral language abilities at the end of kindergarten predicted reading and spelling
at the beginning of first grade. PA skills at this first measurement interval correlated .63 with whole word presentation (WW)
reading (lexical decision) but somewhat more so with WW spelling (.74) and WW reading for naming (.77). At the end of first
grade the concurrent correlations with PA followed the same patterns with lexical decision reading (.47) lower than spelling
(.57) and lower than reading for naming (.74). It seems, therefore, that PA is initially more involved in spelling than in lexical
access for reading (although it is involved in reading for naming). This accords with but qualifies Claim 2a. Visual word
recognition memory abilities at the end of kindergarten correlated .55 with WW lexical decision reading, .63 with WW spelling,
and .63 with WW naming reading, but these dropped dramatically in the concurrent correlations at the end of first grade to .26,
.04, and .28 respectively. The fact that there are these visual correlates of both reading and spelling at the start of Grade 1 but
not at the end may lend some support to Claim 1 that early reading and spelling are visual, or logographic, in nature and that
there is then a shift from this to alphabetic processing. However, this interpretation should be treated with some caution because
the visual memory task used by Berninger et al. was not necessarily a hygienic measure of logographic strategy use, and
performance on this task could reflect additional influences from alphabetic or orthographic knowledge.
Goulandris (1991) reported a small longitudinal study of 27 British children to assess Claim 2b. Verbal intelligence
(vocabulary), nonword spelling, reading, and spelling were used to predict reading and spelling 1 year later. Even when verbal
intelligence, reading age, and spelling age were partialled out, nonword spelling still predicted reading and spelling 1 year later,
clearly demonstrating that a child's ability to generate phonetic spellings is the precursor of the eventual acquisition of alphabetic
reading (Claim 2b).
Finally, and most recently, there are the studies of Wimmer, Landerl, Linortner, and Hummer (1991). The importance of these
studies is that the initial testing was done on children within 1 month of joining school in Salzburg before any reading
instruction had taken place. In the first study of 50 children, vowel substitution (their measure of PA) was a significant predictor
of end of first-grade spelling (r = .49) and reading (r = .45), and these effects were significant even after IQ, initial letter
knowledge, and initial, nonword reading abilities were partialled out. Study 2 investigated 42 children and the degree to which
initial PA and rudimentary logographic

Page 285
(identification of ubiquitous logos such as Coca Cola or Milky Way) and alphabetic (the reading of logos printed in upper case
to distort their original distinctive whole-word shape) reading skills predicted later reading of familiar words and nonwords and
alphabetic and orthographic spelling. PA later predicted alphabetic spelling (.31) and reading accuracy (.30) but not
orthographic spelling (.15). Logographic reading did not predict later reading (.08), but alphabetic reading did more so (.19),
suggesting a change of reading strategy over this first year from logographic to alphabetic strategies. Logographic reading did,
however, predict later orthographic spelling (.42). Most striking was the finding that PA skill at the end of the year was only
moderately predicted by prior PA skill (.31) but was highly correlated with reading accuracy at that time (.53) and even more so
with nonword spelling (.66). In other words, over the course of the children's first year of entry into literacy, PA and spelling
became strongly enmeshed abilities as a result of a symbiotic developmental relationship. Wimmer et al. summarized their
article in the title ''The relationship of phonemic awareness to reading acquisition: More consequence than cause but still
important'' and concluded that, in most children, quite limited exposure to reading and spelling instruction is sufficient to induce
explicit awareness of phonemic segments of words in a language, at least in languages such as German that have fairly regular
grapheme-phoneme correspondences.
These findings are compared and brought together with the findings of training studies in the conclusions of this chapter.
Orthographic Reading and Orthographic Spelling
We have already described Mommers' (1987) finding of causal paths from word-reading skill to spelling at later stages of
development. Stanovich (Cunningham & Stanovich, 1990, 1993; Stanovich & Cunningham, 1992) produced a number of group
studies that give further confirmation to Claim 3. Stanovich and Cunningham (1992) argued that, when reading, "whatever
cognitive processes are engaged over word or word-group units (phonological coding, semantic activation, parsing, induction of
new vocabulary items) are being exercised hundreds of times a day. It is surely to be expected that this amount of cognitive
muscle-flexing will have some specific effects" (p. 51). They demonstrate, in multiple regression analyses, that adults who read
a lot (who have high print exposure, as measured on Author and Magazine Recognition Tests) are better spellers even when
nonverbal intelligence is controlled. The same is true for third- and fourth-grade children. Cunningham and Stanovich (1990)
found that even after partialling out IQ, memory ability, and phonological processing abilities, print exposure (the amount of
reading the child did) accounted for significant variance in orthographic knowledge.

Page 286
This result is clear confirmation of Claim 3 that the move to an orthographic strategy of spelling was driven by reading at later
stages of development. However, Cunningham and Stanovich (1993) also showed that it is true in the case of first-grade
children. They demonstrated that 6-7-year-old children's ability to spell phonologically irregular words such as read, talk,
mouse, rough, which require consultation of an orthographic lexicon for conventional rather than phonetic spelling, was a
separable component of variance in word recognition from phonological awareness. Furthermore, even at these young ages,
significant variance in the ability to perform correct orthographic spelling was accounted for by print exposure after
phonological processing skill had been partialled out.
Thus, there does seem to be clear evidence for Claim 3 that exposure to the letter sequences of words in reading allows the child
to develop orthographic representations that can then be used in spelling. However, these results qualify the claim in one
important aspect: It seems that, in normal children, this is happening from quite early on in the development of literacy (Grade
1) rather than being solely an aspect of a later third stage of reading and spelling development.
Tests from Training Studies

Studies that involve training children in PA are fairly consistent in supporting Claim 2a of Frith's (1985) model in which
phonological awareness is more involved in early spelling than in early reading-training in PA has its first effect on children's
spelling.
Bradley and Bryant (1983) taught metaphonological skills to 4- and 5-year-old children who could not read and who were at
least two standard deviations below average in sound categorization ability. Some children were trained in categorizing common
beginning (hen, hat), middle (hen, pet), and end (hen, man) sounds; others received this training and, with the help of plastic
letters, how each common sound was represented by a letter of the alphabet. The training had a positive effect on both reading
and spelling measured over the next 4 years, and Bradley and Bryant (1985) concluded that "although others have suggested a
link between phonological awareness and reading, our study is the first adequate empirical evidence that the link is causal" (p.
41). However, there is more in their study than this. In the first place, there was much greater benefit from training sound
categorization in conjunction with plastic letters that labelled the sounds, confirming the emphases of Ehri (1984) and Frith
(1985) that phonological awareness is more readily acquired when it is related to orthography. Second, this training had more
effect on later spelling (17 months' gain) than it did on later reading (8.5 months' gain).
Tornus (1984) reported a training intervention whereby 38 first graders were assigned to either phonological awareness training
or to a general

Page 287
language activity control for 8 weeks. They were pre- and posttested on reading, spelling, sound blending, and segmentation.
Training of phonological skills was effective and, among the children with the lowest phonological awareness pretest
performance, specific phonological training improved spelling performance more than did general language activities.
Phonological training did not, however, directly affect reading.
Experimental evidence for Claim 2b is provided by Ehri and Wilce (1987a, 1987b) who taught kindergarten children to spell
words by attending to constituent letter-sound sequences and, when necessary, to phonetic, phonemic, and articulatory cues.
These children learned to read words better than children who were taught isolated letter-sound relationships. Thus, children
trained in spelling were superior to matched controls in their ability to use phonetic cues and letter-sound constituents when
learning to read.
Lundberg, Frost, and Petersen (1988) trained over 200 Danish preschool children who had no reading instruction in phonemic
awareness, and they assessed the later effects on reading and spelling in first and second grades. Training had a small effect on
rhyming and syllable manipulation abilities and a dramatic effect on phoneme segmentation abilities. These improvements, in
turn, had a large facilitative effect on first-grade spelling (p < .001) but only a marginal immediate effect on first-grade reading
(p < .10). By the end of Grade 2, however, the effect of training on spelling persisted (p < .001) and, by now, had an additional
effect on reading (p < .01).
Lie (1991) assessed the effects on later reading and spelling of the training of metaphonological skills in approximately 100
first-grade Norwegian children. Training both in phoneme identification and in phoneme segmentation and blending had a
facilitative effect on later Grade 1 and Grade 2 reading and spelling but, again, the initial effect on spelling was somewhat
greater than that on reading. At the end of Grade 1, students who had been trained in sequential phoneme segmentation scored
significantly higher in spelling than students who had received positional (phoneme isolation) training, a result that stresses that
it is important to phonologically analyze a word sequentially in order to spell it.
The results of Bradley and Bryant (1983), Tornus (1984), Lundberg et al. (1988), and Lie (1991) were all consistent with Claim
2a, that phonological awareness is much more involved in early spelling than in early reading. The findings of Ehri and Wilce
(1987a) and Lundberg et al. supported pacemaker Claim 2b that the acquisition of phonological awareness in spelling drives the
development of alphabetic reading.
Longitudinal Studies of Single Cases

No chapter on longitudinal tests of developmental models is complete without describing one other essential source of evidence
that of detailed longitudinal clinical investigations of single cases. Unfortunately, there is simply

Page 288
not the space to do justice to this body of research, and, although this small subsection may allow an assertion of formal
completeness, it is readily acknowledged that this chapter is far from satisfactory in this respect.
Just one study is used to illustrate the power of this approach, particularly with respect to Frith's (1985) claims about
developmental arrest. Seymour and Evans' (1988) study was relevant to Claim 1. They analyzed reading and spelling processes
in a case (AT) of developmental disability associated with Klinefelter XXY syndrome. Between the ages of 7 and 11, this boy's
reading was almost entirely logographic; he had a sight vocabulary of over 500 words that he had been taught, but he was
unable to do any sounding and blending, with error rates on nonwords between 95% and 99%. He could spell approximately 170
words, but he only managed 1 out of 200 nonwords correctly. Thus, AT showed an almost complete absence of alphabetic
functions, that is, his development was arrested at a logographic stage.
Conclusions
The longitudinal studies reviewed previously give considerable support for several of the claims of Frith's (1985) model. There
is some evidence of a logographic first stage of reading (Berninger et al., 1990; Seymour & Evans, 1988), although no study can
be found that provides corroboration of logographic reading acting as a pacemaker for a logographic stage of spelling (Claim 1).
Phonological awareness does seem more related, in early development, to spelling than to reading (Tornus, 1984; Wimmer et
al., 1991; cf. Juel et al., 1986; Mommers, 1987), and training in PA first affects the development of spelling rather than reading
(Bradley & Bryant, 1983; Lie, 1991; Lundberg et al., 1988; Tornus, 1984; Claim 2a). The acquisition of PA through spelling
engenders development of an alphabetic strategy of reading (Cataldo & Ellis, 1988; Ehri & Wilce, 1987; Ellis & Cataldo, 1990;
Goulandris, 1991; Lundberg et al., 1988; Mommers, 1987; Claim 2b; see Ellis & Cataldo, 1990, for the pedagogical implications
of this). The acquisition of orthographic knowledge through reading promotes orthographic spelling (Mommers, 1987; Stanovich
& Cunningham, 1992; Claim 3).
These data also suggest some additions and qualifications. First, they very clearly show that there are different facets of PA, an
early implicit awareness of syllables and rhyme, and a later sophisticated explicit ability at segmentation at the level of
phonemes. Even very early reading seems to capitalize on this implicit phonological awareness that plays a role in the
logographic reading, otherwise characterized as being primarily visual in nature. The aforementioned studies also demonstrate
that experience in alphabetic spelling and reading promotes the more analytic and explicit phonological awareness at subsyllabic
levels. In normal children, this is acquired very quickly and, although it remains to be determined by just how much, it seems
likely

Page 289
that this comes more from alphabetic spelling instruction than from practice in alphabetic reading. Finally, although experience
with reading allows the child to abstract knowledge of orthographic sequences that can then be applied in spelling, it seems that
this is happening not just with mature readers who are solidly at an orthographic stage of reading but also with first-grade
children at the beginnings of literacy. Frith's (1985) model, literally interpreted, is a strong stage model suggesting three very
different strategies of reading and spelling at three discrete stages of development. Although this is probably true as a broad
characterization, it seems that there are mutual influences between the alphabetic and orthographic aspects of reading and
spelling at all stages of development. These qualifications to a stage model of development are taken up by other chapters in this
volume. Treiman and Cassar (this volume) showed that there can be some rudimentary influences of orthographic knowledge on
spelling right from the beginning: Even early on, children may notice (a) that English words may start but not end with capital
letters; (b) that words may end but not start with double consonants; and (c) that letters such as e and s may double but that
letters such as a and v rarely do. Rieben and Saada-Robert (this volume) similarly concluded from their word-search and word-
copying data that children are more flexible than a fixed-stage model predicts and that their development is better described in
terms of phases of dominance of strategy.
A new skill initially builds on whatever relevant abilities are already present, then, as it is used, it may well legitimatize and
make more relevant (Istomina, 1975) those prior skills and cause their further development. Stanovich (1986) persuasively
argued the case for reciprocal relationships and bootstrapping effects, "In short, many things that facilitate further growth in
reading comprehension ability general knowledge, vocabulary, syntactic knowledge are developed by reading itself" (p. 364);
"interrelationships between the various subskills of reading and intelligence increase with age, probably due to mutual
facilitation" (Stanovich, Cunningham & Feeman, 1984, p. 278, my emphasis). What is true of reading and intelligence also
applies to the symbiotic development of spelling, reading, and PA: They interact reciprocally over time.
Finer-grained studies are now needed to further determine the contributions from one to the other, the representations of the
mutual exchanges (visual, phonological, orthographic), and their particular content at each point in the acquisition of literacy.
Ellis (1994) described the type of longitudinal study that remains to be done in order to properly chart the developmental
interactions between these representational systems.
There is also a role for detailed modelling work in order to understand the formal properties of these connections. How does the
nature of the phonological and orthographic representations available to the child determine the kinds of sound-spelling
correspondences that are learned and

Page 290
used at different developmental stages? More specifically, do children begin to spell using high-level (e.g., rime-based) sound-
spelling correspondences, or lower level (e.g., phoneme-grapheme) correspondences, or are both used as a result of interactions
between phonological and orthographic systems at various levels of representation? It is helpful to approach this issue by
considering the conflicting levels of representation in the different modalities. In terms of phonological representations, there is
considerable evidence that the first sublexical representations to develop are onsets and rimes. Thus, Goswami (1986, 1993), on
the basis of a study of analogical transfer by children, argued that rime-based, spelling-sound correspondences are among the
first to be used because these are the most relevant phonological units that children initially possess that can be used to link to
orthographic patterns. Yet, in terms of orthographic representations, the most salient representational unit in early spelling
instruction is the individual letter of the alphabet.
When children attempt to make connections between sound patterns and print, they are faced with the impossible task of
mapping between incompatible representations. Initially, they have no orthographic rime, letter-cluster units that can be made to
correspond with phonological rime units, and they have no phoneme representations that can be mapped on to letter units. To
learn to spell, then, the child must not only realize that there is some connection between orthographic and phonological forms
but must also develop representations that allow mapping between orthography and phonology at compatible levels. Brown and
Ellis (1994b) presented an analysis of the ways in which the different sound-spelling and spelling-sound correspondences could
be used, depending on the availability of appropriate representations, and some of the ways in which the availability of units of
representation in one domain might encourage the equivalent level of representation in the other. For example, if a child has a
phonological representation for the rime /Int/ and is faced with the task of finding some orthographic input pattern that reliably
predicts this rime, this may lead to development of a recognition unit for the orthographic pattern int (Goswami, 1993). More
detailed models of developmental changes in orthographic and phonological representations are needed to forward our
understanding of the developmental reciprocity between representational systems for spelling, reading, and phonology. The
reader is referred to Brown and Ellis (1994a) for a review of some of the current computational models of these interactions.
References
Alegria, J., Pignot, E., & Morais, J. (1982). Phonetic analysis of speech and memory codes in beginning readers. Memory and
Cognition, 10, 451-456.
Backman, J. (1983). The role of psycholinguistic skills in reading acquisition: A look at early readers. Reading Research
Quarterly, 18, 466-479.

Page 291
Berninger, V. W. (1986). Normal variation in reading acquisition. Perceptual and Motor Skills, 62, 691-716.
Berninger, V. W., Abbott, R. D., & Shurtleff, H. A. (1990). Developmental changes in interrelationships of visible language
codes, oral language codes, and reading or spelling. Learning and Individual Differences, 2, 45-66.
Biemiller, A. (1970). The development of the use of graphic and contextual information as children learn to read. Reading
Bissex, G. L. (1980). GNYS AT WRK: A child learns to read and write. Cambridge, MA: Harvard University Press.
Bradley, L., & Bryant, P. E. (1983). Categorising sounds and learning to read: A causal connection. Nature, 301, 419-421.
Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading and spelling. International Academy for Research in Learning
Disabilities, No. 1. Ann Arbor: University of Michigan Press.
Brown, G. D. A., & Ellis, N. C. (Eds.). (1994a). Handbook of spelling: Theory, process and intervention. Chichester, UK:
Wiley.
Brown, G. D. A., & Ellis, N. C. (1994b). Issues in spelling research: An overview. In G. D. A. Brown & N. C. Ellis (Eds.),.
Handbook of spelling: Theory, process and intervention (pp. 3-25). Chichester, UK: Wiley.
Bryant, P. E., & Bradley, L. (1980). Why children sometimes write words which they do not read. In U. Frith (Ed.), Cognitive
Cataldo, S., & Ellis, N. (1988). Interactions in the development of spelling, reading and phonological skills. Journal of Research
in Reading, 11, 86-109.
Chomsky, C. (1977). Approaching reading through invented spelling. In L. B. Resnick & P. A. Weaver (Eds.), Theory and
practice of early reading: Vol. 2 (pp. 43-59). Hillsdale, NJ: Lawrence Erlbaum Associates.
Chukovsky, K. (1968). From two to five. Berkeley: University of California Press.
Clark, E. V. (1978). Awareness of language: Some evidence from what children say and do. In A. Sinclair, R. J. Jarvella, & W.
J. M. Levelt (Eds.), The child's conception of language (pp. 17-38). New York: Springer-Verlag.
Cunningham, A. E., & Stanovich, K. E. (1990). Tracking the unique effects of print exposure in children: Associations with
vocabulary, general knowledge and spelling. Journal of Educational Psychology, 83, 264-274.
Cunningham, A. E., & Stanovich, K. E. (1993). Children's literacy environments and early word recognition subskills. Reading
and Writing, 5, 193-204.
Ehri, L. C. (1979). Linguistic insight: threshold of reading acquisition. In T. G. Waller & G. E. MacKinnon (Eds.), Reading
research: Advances in theory and practice: Vol. 1 (pp. 63-114). New York: Academic Press.
Ehri, L. C. (1984). How orthography alters spoken language competencies in children learning to read and spell. In J. Downing
& R. Valtin (Eds.), Language awareness and learning to read (pp. 119-147). New York: Springer-Verlag.
Ehri, L. C. (1986). Sources of difficulty in learning to spell and read. Advances in Developmental and Behavioural Paediatrics,
7, 121-195.
Ehri, L. C., & Wilce, L. S. (1985). Movement into reading: Is the first stage of printed word learning visual or phonetic?
Ehri, L. C., & Wilce, L. S. (1987a). Does learning to spell help beginners learn to read words? Reading Research Quarterly, 22,
47-65.
Ehri, L. C., & Wilce, L. S. (1987b). Cipher versus cue reading: An experiment in decoding acquisition. Journal of Educational
Ellis, N. C. (1994). Longitudinal studies of spelling development In G. D. A. Brown & N. C. Ellis (Eds.), Handbook of spelling:
Theory, process, and intervention (pp. 154-177). Chichester, UK: Wiley.

Page 292
Ellis, N. C., & Cataldo, S. (1990). The role of spelling in learning to read. Language and Education, 4, 1-28.
Ellis, N., & Large, B. (1987). The development of reading: As you seek so shall you find. British Journal of Psychology, 78, 1-
28.
Ellis, N., & Large, B. (1988). The early stages of reading: A longitudinal study. Applied Cognitive Psychology, 2, 47-76.
Frith, U. (1981). Experimental approaches to developmental dyslexia: An introduction. Psychological Research, 43, 97-110.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, M. Coltheart, & J. Marshall (Eds.), Surface
Gentry, J. R. (1978). Early spelling strategies. The Elementary School Journal, 79, 88-92.
Gentry, J. R. (1982). Analysis of developmental spelling in GNYS AT WORK. The Reading Teacher, 36, 192-200.
Goswami, U. (1986). Children's use of analogy in learning to read: A developmental study. Journal of Experimental Child
Goswami, U. (1993). Towards an interactive analogy model of reading development: Decoding vowel graphemes in beginning
reading. Journal of Experimental Child Psychology, 56, 443-475.
Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. Hove, UK: Lawrence Erlbaum Associates.
Gough, P. B., & Hillinger, M. L. (1980). Learning to read: An unnatural act. Bulletin of the Orton Society, 30, 171-176.
Goulandris, N. K. (1991). Alphabetic spelling: Predicting eventual literacy attainment. In C. M. Sterling & C. Robson (Eds.),
Psychology, spelling, and education (pp. 143-158). Clevedon, Avon, UK: Multilingual Matters.
Harris, M., & Coltheart, M. (1986). Language processing in children and adults. London: Routledge & Kegan Paul.
Henderson, E. H., & Beers J. W. (Eds.). (1980) Developmental and cognitive aspects of learning to spell: A reflection of word
knowledge. Newark, DE: International Reading Association.
Istomina, Z. M. (1975). The development of involuntary memory in preschool-age children. Soviet Psychology, 13, 5-64.
Juel, C., Griffith, P. L., & Gough, P. B. (1986). Acquisition of literacy: A longitudinal study of children in first and second
grade. Journal of Educational Psychology, 78, 243-255.
Lewkowicz, N. K. (1980). Phonemic awareness training: What it is and how to teach it. Journal of Educational Psychology, 72,
686-700.
Lundberg, I., Frost, J., & Petersen, O-P. (1988). Effects of an extensive program for stimulating phonological awareness in
Lundberg, I., Olofsson, A., & Wall, S. (1980). Reading and spelling skills in the first school years predicted from phonemic
awareness skills in kindergarten. Scandinavian Journal of Psychology, 21, 159-173.
Liberman, I. Y., & Shankweiler, D. (1979). Speech, the alphabet, and teaching to read. In L. B. Resnick & P. A. Weaver (Eds.),
Theory and practice of early reading (pp. 109-132). Hillsdale, NJ: Lawrence Erlbaum Associates.
Lie, A. (1991). Effects of a training program for stimulating skills in word analysis in first-grade children. Reading Research
Quarterly, 26, 234-250.
MacLean, M., Bryant, P., & Bradley, L. (1987). Rhymes, nursery rhymes, and reading in early childhood. Merrill-Palmer
Quarterly, 33, 255-281.
Mann, V. A. (1984). Longitudinal prediction and prevention of early reading difficulty. Annals of Dyslexia, 34, 117-136.
Marsh, G., Friedman, M. P., Welch, V., & Desberg, P. (1980). The development of strategies in spelling. In U. Frith (Ed.),
Cognitive processes in spelling. London: Academic Press.

Page 293
Marsh, G., Friedman, M. P., Welch, V., & Desberg, P. (1981). A cognitive-developmental theory of reading acquisition. In T.
G. Waller & G. E. MacKinnon (Eds.), Reading research: Advances in theory and practice: Vol. 3 (pp. 199-221). New York:
Academic Press.
Mommers, M. J. C. (1987). An investigation into the relationship between word recognition, reading comprehension and
spelling skills in the first two years of primary school. Journal of Reading Research, 10, 122-143.
Mommers, M. J., Van Leeuwe, J. F., Oud, J. H., & Janssens, J. M. (1986). Decoding skills, reading comprehension and spelling:
A longitudinal investigation. Tijdschrift voor Onderwijsresearch, 11, 97-113.
Morais, J., Alegria, J., & Content, A. (1987). The relationships between segmental analysis and alphabetic literacy: An
interactive view. Cahiers de Psychologie Cognitive, 7, 415-438.
Morais, J., Cary, I. L., Alegria, J., & Bertelson, P. (1979). Does awareness of speech as a sequence of phones arise
spontaneously? Cognition, 7, 323-331.
Morris, D. (1983). Concept of word and phoneme awareness in the beginning reader. Research in the Teaching of English, 17,
359-373.
Perfetti, C. A., Beck, I., Bell, L. C., & Hughes, C. (1987). Phonemic knowledge and learning to read are reciprocal: A
longitudinal study of first-grade children. Merrill-Palmer Quarterly, 33, 283-319.
Read, C. (1975). Children's categorizations of speech sounds in English. Urbana-Champaign, IL: National Council of Teachers
of English.
Read, C., Zhang, Y., Nie, H., & Ding, B. (1986). The ability to manipulate speech sounds depends on knowing alphabetic
reading. Cognition, 24, 31-44.
Seymour, P. H. K., & Elder, L. (1986). Beginning reading without phonology. Cognitive Neuropsychology, 3, 1-36.
Seymour, P. H. K., & Evans, H. M. (1988). Developmental arrest at the logographic stage: Impaired literacy functions in
Klinefelter's XXXY syndrome. Journal of Research in Reading, 11, 133-151.
Shankweiler, D., Liberman, I. Y., & Savin, H. B. (1972). General discussion of papers. In J. F. Kavanagh & I. Mattingly (Eds.),
Language by ear and by eye (pp. 327-330). Cambridge, MA: MIT Press.
Slobin, D. J. (1978). A case study of early language awareness. In A. Sinclair, J. Jarvella, & W. J. M. Levelt (Eds.), The child's
conception of language (pp. 46-61). New York: Springer-Verlag.
Smith, F. (1978). Understanding reading: A psycholinguistic analysis of reading and learning to read (2nd ed.). New York:
Holt, Rinehart & Winston.
Snowling, M., & Perin, D. (1983). The development of phoneme segmentation skills in young children. In J. Sloboda (Ed.), The
acquisition of symbolic skills (pp. 87-99). London: Plenum.
Stanovich, K. E. (1986). Matthew effects in reading: Some consequences of individual differences in the acquisition of literacy.
Stanovich, K. E., & Cunningham, A. E. (1992). Studying the consequences of literacy within a literate society: The cognitive
correlates of print exposure. Memory and Cognition, 20, 51-68.
Stanovich, K. E., Cunningham, A. E., & Cramer, B. B. (1984). Assessing phonological awareness in kindergarten children:
Issues of task comparability. Journal of Experimental Child Psychology, 38, 175-190.
Stanovich, K. E., & West, R. F. (1989). Exposure to print and orthographic processing. Reading Research Quarterly, 24, 402-
433.
Tornus, M. (1984). Phonological awareness and reading: A chicken and egg problem? Journal of Educational Psychology, 76,
1346-1358.
Torrey, J. W. (1979). Reading that comes naturally: The early reader. In T. G. Waller & G. E. MacKinnon (Eds.), Reading
research: Advances in theory and practice, Vol. 1 (pp. 117-128). New York: Academic Press.

Page 294
Valtin, R. (1984). The development of metalinguistic abilities in children learning to read and write. In J. Downing & R. Valtin
(Eds.), Language awareness and learning to read. New York: Springer-Verlag.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of reading
skills. Psychological Bulletin, 101, 192-212.
Weber, R. M. (1970). A linguistic analysis of first-grade reading errors. Reading Research Quarterly, 5, 427-451.
Wimmer, H., Landerl, K., Linortner, R., & Hummer, P. (1991). The relationship of phonemic awareness to reading acquisition:
More consequence than precondition, but still important. Cognition, 40, 219-249.

Page 295
Chapter 15
Relations Between Word-search Strategies and Word-copying Strategies in Children Aged 5 to 6 Years Old
Laurence Rieben
Madelon Saada-Robert
Reading and Writing Words: Autonomy or Interdependence?

It is possible to distinguish two trends under discussion concerning the relation between the reading and writing of words. A
number of researchers focused on the differences between the two areas of processing, sometimes even postulating the existence
of two separate lexical memories (Bryant & Badley, 1980; Campbell, 1987; Frith, 1980). Others considered that the reading and
writing of words are complementary activities based on the same orthographical and phonological knowledge that is stored in a
single system (Ehri, 1980, 1989, 1991a; Gough, Juel, & Griffith, 1992; Perfetti, 1992; Waters, Bruck, & Seidenberg, 1985). As
shown, the results of our longitudinal study comparing word-search strategies and word-copying strategies favor the second
position.
There are, in fact, significant correlations between the reading of words and the writing of words (Gough, Juel, & Griffith, 1992;
Zutell, 1992) despite the fact that it is possible to identify participants who exhibit a contrasting pattern, that is, those who
achieve high scores in reading tasks but low scores in spelling tasks1 (Bruck & Waters, 1990; Frith, 1980; Waters, Bruck,
1 In fact, the reverse pattern does not seem to exist. Bruck and Waters (1990) pointed out that of 343 third graders and
357 sixth graders, there were no examples of students being good spellers while being bad readers.

Page 296
& Seidenberg, 1985). Such results have generally been interpreted in terms of models that postulate the existence of two routes
for the reading and writing of words. One route is based on the rules of correspondence between sounds and letters
(phonological route), which allows participants to read and write pseudowords or uncommon words; the other route is based on
the idea of access to a mental lexicon that contains orthographic information about words (lexical route), allowing participants to
read or write irregular words (Barron, 1986; Peereman, 1991; Segui, 1991). Thus, Frith studied and compared three groups of
12-year-old children: good readers and writers (++ Group), poor readers and writers (-- Group), and a group consisting of good
readers but poor writers (+- Group). Her results lead her to the following two conclusions: (a) The phonological route is active
during spelling because a large proportion of the spelling errors encountered in the ++ and +- Groups contain phonological
errors; (b) the processes involved in reading and spelling are independent because she assumed that, although good 12-year-old
readers employ the phonological route for spelling, they use the direct lexical path for reading. Waters, Bruck, and Seidenberg
found these conclusions unconvincing, in particular because there was no experimental verification of the assumption that
participants use the direct path for reading. These authors studied three similar but younger (third graders) groups of children
and controlled the reading and spelling processes more systematically. They concluded that all the children tended to use the
phonological route both in reading and spelling.
It should be noted that the dual-route theoretical framework has stimulated much more work in reading than in spelling.
However, the collection of papers on spelling processes, edited by Frith (1980), influenced the use of the model for spelling
research, and stimulated research focusing on the relationship between reading and spelling. The standard two-process model
has been refined and extended since the early 1980s. One example of this development can be found in research demonstrating
the importance of analogy strategies for reading and spelling words in both adults (Glushko, 1981) and children (Goswami &
Bryant, 1992; Gombert, Bryant, & Warrick, chap. 12, this volume).
However, even if the separability of the two routes finds good support in neuropsychological reports (see Zesiger & de Partz,
chap. 3, this volume), there is more doubt about the independence of the two paths in normal readers and spellers (Barry, 1994).
The traditional interpretation holding that there are two encapsulated processing systems was contested by Seymour (1992). He
suggested that there may be a single system managing constraints that are generated by both morphemic (lexical, semantic) and
submorphemic (syllabic, phonemic) sources. In effect, postulating that these two sources function interdependently makes it
possible to account for the results of experiments that show how participants spell a particular dictated

Page 297
pseudoword is influenced by the spelling of an orally printed prime word (Campbell, 1983). For example, participants who hear
the word sweet before writing the pseudoword /pri:t/ tend to write PREET, whereas those who hear the word treat tend to write
PREAT. The orthodox dual-route interpretation may also be incompatible with Perfetti, Zhang, and Berent's (1992) findings that
show the pervasiveness of phonological processes in reading with respect to writing systems and to individuals.
One further type of result that is frequently cited as evidence of the relative autonomy of reading and spelling: Words that are
read correctly are not necessarily written correctly, and vice versa. Such results have been obtained by Bryant and Badley (1980)
who observed that when 6-year-old children were asked to read and write the same set of 30 words, an average of 6.3 words
could be read but not written, and an average 3.9 words could be written but not read. However, a dcalage in favor of spelling
over reading is very infrequent; furthermore, Bryant and Badley's results showed that it was no longer present in 7-year-old
children. In a recent study using the same experimental paradigm, but with a double pass of the reading and spelling tests,
Gough, Juel, and Griffith (1992) confirmed the rarity of this phenomenon and demonstrated that an equivalent percentage of
words is either not read or not written in the same way in the two passes. These authors concluded that such results may simply
be interpreted as indicative of the instability of performance that is frequently observed in children engaged in a learning
process. Although the question of the independence of the reading and writing of words is still largely undecided, there are two
reasons for doubting that these two activities function autonomously: First, it is difficult to demonstrate that children
systematically use different strategies for reading and spelling; second, even with such a demonstration, it is far from certain that
these different strategies operate independently.
Are Stage-based Models Good Models for Describing the Acquisition of Reading and Spelling?
The literature contains numerous independently proposed stage-based models for the identification (e.g., see Ehri & Wilce,
1985; Frith, 1985; Marsh, Friedman, Welch, & Desberg, 1981) and the writing of words (e.g., see Ehri, 1991b; Ferreiro, 1978;
Frith, 1985; Henderson, 1992). These models tend to agree on the distinction between three phases logographic, alphabetic, and
orthographic although differences remain in the number of stages identified, the importance attributed to each of them, and the
transitional mechanisms that allow participants to move between stages (Ehri, 1991c; Rieben, 1991; Rieben & Saada-Robert,
1991).

Page 298
However, the independence between the stages of reading and the stages of spelling acquisition has been questioned. Ehri
(1980, 1989, 1991a) proposed a process by which the orthographic aspects of words are amalgamated with other aspects already
present in lexical memory, in particular with phonological aspects but also with syntactic and semantic aspects, the spelling
being used as a visual symbol to aid in the retention of the sounds in memory.
Frith (1985) proposed a joint stage-based system for reading and spelling and suggests an out-of-phase model that allows for
dcalages favoring reading and spelling alternately: Reading reaches the logographic phase before spelling, and spelling
overtakes reading at the point when participants accede to the alphabetical phase before reading gains ground once more as they
enter the orthographic stage (see Ellis, chap. 14, this volume).
The necessity of passing through a logographic stage has also been called into question, in particular with reference to reading
(Ehri & Wilce, 1985; Stuart & Coltheart, 1988; Seymour, 1992; Sprenger-Charolles & Casalis, 1995; Wimmer & Hummer,
1990) because some children make use of phonological mediation from the very beginning of learning.
Furthermore, if stages provide adequate models for the description of changes in strategy use for the reading and writing of
words, the question of their overlap must be studied more systematically (Rieben, 1995a). Traditionally, the idea of stage is
interpreted as describing relatively abrupt and generalized qualitative changes. Although this type of qualitative leap has never
really been observed, the generally accepted interpretation is that one type of strategy is replaced in a fairly radical manner by
another, and we expect to be able to assign a child to a reasonably unambiguous reading/spelling stage (Frith, 1985). However,
results exist (Rieben, 1995a; Rieben & Saada-Robert, 1991; Saada-Robert & Rieben, 1993; Seymour & Elder, 1986; Seymour &
Evans, 1992) that point to the coexistence of the logographic and alphabetic stages. Such an observation concerning the
variability of strategies is not specific to the field of written language. It is also encountered in other fields of learning and, in
particular, in the arithmetical domain (Siegler, 1987).
From Reading and Writing Words to Searching and Copying Words

In most research, the experimental paradigm consists of reading lists of words and spelling from dictation. The participants'
strategies are inferred either on the basis of an analysis of the errors they make or on the basis of differences in the processing
time required for words or pseudowords, regular or irregular words, common or rare words. However, this type of

Page 299
paradigm is not suitable for children at the start of the learning process because their lexical knowledge is at once too limited
and too idiosyncratic. For this reason, we use a paradigm in which we observe how young children search for and then copy
words in a text production situation. In this way, the activity forms part of a wider context that is meaningful for them. Little
research has been devoted to the question of word-search studies. A similar paradigm consists of writing a short sentence and
then reading it to the child before asking the participant whether a particular word has been written and, if it has, where it
appears in the written sentence. This situation has been used by Ferreiro (1978) and Morris (1992) to study children's awareness
of the elements that have to be written an interesting result from this research is the finding that 4-year-old children generally
think that there is no need to write articles or prepositions and of the segmentation of utterances into lexical units.
Ehri and Sweet (1991) reported a study closer to our perspective. In their experiment, a short text was read to the participants
over a number of training sessions with the result that they knew its contents practically by heart. The children were then
confronted with a variety of postexperimental tests, one of which, a word-recognition task, required them to identify isolated
words within the text. Ehri and Sweet showed that the results, like those obtained in more traditional reading tasks, are closely
linked to phonological awareness and letter knowledge. A major difference between our situation and the one studied by Ehri
and Sweet however, is the role played by memorization of the text. Our experiment made use of long texts (approximately 200
words), collective productions that were read to the students without being memorized. The word-search task required the
mobilization of active problem-solving strategies based on cues relating to the properties of the words and texts.
Word copying is another area that has not been the object of extensive research. Originally, word copying was viewed as a tool
for the study of the graphomotor aspects of writing. More recently, researchers have come to believe that text copying may
provide an indication of children's level of acquisition and representation of written language (Fijalkow & Liva, 1988). For
example, the longitudinal study conducted by Humblot, Fayol, and Lonchamp (1994) used word copying to study the
characteristics of the units processed in the written production of 6- and 7-year-old children. This research shows that copying is
far from being a simple, mechanical process, drawing instead on lexical knowledge (the best known words tend to be copied
after only a single reference to the written model) and sublexical knowledge (multiple references to the written model implying
segmental analysis in the case of unfamiliar words). In children aged 6 and 7, the syllable seems to play an important role in the
process of word decomposition.

Page 300
Our Study
Our study of word-search strategies and word-copying strategies approached the theoretical questions discussed previously
within a specific perspective that can be summarized in three points. First, because research within a developmental framework
that define stages involved in the acquisition of reading/spelling does not pay sufficient attention to individual differences, our
perspective is simultaneously developmental and differential. Second, we observe children in functional reading/writing
situations (i.e., in communication) rather than in test situations. Third, studies in school allow observations in the students'
primary learning environment (Rieben, Meyer, & Perregaux, 1991; Rieben, 1989).
Method
Our research involved the longitudinal classroom observation of 21 students at four points during the school year in a situation
of written utterance, a didactic approach based on the work of Clesse (1977) and Hbrard (1977). In this approach, young
prereaders are encouraged to reconstruct and dictate to their teacher a story taken from a children's book previously presented in
picture form and then read. This results in the production of a text displayed on three large, wall-mounted pages, the Reference
Text (RT). The children are then asked to draw a picture representing one episode of the story and then to write an individual
commentary of their drawing, using the RT as a ''dictionary'' in which they can be fairly sure of finding the words they need.
During this activity, the word-search strategies and word-copying strategies used by the participants were observed at four
points during the school year, approximately every 6 weeks between November and May (T1, T2, T3, T4, amounting to a total
of approximately 540 minutes of observation for each student).
Participants
At the start of the school year (T1), the participants were nonreaders (with the exception of 4 children who could read a few
words but could not read and understand a written instruction indicating how to complete a drawing). The group consisted of 11
students in the second year of kindergarten (K2) between 5.0 and 5.11 at the start of the experiment (mean age 5.5) and 10 first
graders (G1) between 6.0 and 6.9 (mean age 6.4).2 The children were
2 It should be noted that in French-speaking Switzerland, reading is mostly taught in G1. Our observations also included
K1 children, not because we wanted to conduct a systematic comparison between these two school grades but because we
wanted to identify the strategies used by younger children and, most importantly, children who were less directly
concerned with the acquisition of reading in the classroom situation.

Page 301
evenly distributed between the different socioeconomic categories (33% high, 33% average, 33% low socioeconomic
background).
Results
Observation of the children's actions and verbal productions allowed us to infer 7 types of word-search strategy and 7 types of
word-copying strategy. Separate analyses were conducted and published for the word-search strategies (Rieben & Saada-Robert,
1991) and word-copying strategies (Rieben & Saada-Robert, 1993). Here, the most important results revealed by these separate
analyses are summarized before we move on to a study of the relations between the two types of strategy.
Word-Search Strategies
Seven types of word-search strategy were distinguished, based on: (a) the localization of words in the title that the child has
generally memorized (TITLE); (b) the localization of words in the text with the child being aware of the text organization and
knowing that the required word is to be found at the beginning, middle, or end of the story or on a particular page (TEXT); (c)
visual cues relating to the length or other salient features of the word (VISUAL); (d) grapheme phoneme or syllable
correspondence (CORRESPONDENCE); (e) a blind search for and copying of strings of words from the text rereading shows
that the children "read" what they wanted to write and not what they have actually written (BLIND); (f) the deliberate search for
and copying of a string of words from the text with the child being able to reread what he or she has written (COPY); (g) the
direct, correct identification of the required words in the text without sounding (ORTHO). For the 21 subjects and the four
periods of observation, 1,044 strategies were counted. These strategies were categorized by two independent judges who agreed
on 79% of cases.
Table 15.1 presents the percentages for the different types of strategy observed from point T1 though to T4. As expected,
ORTHO and COPY strategies increased as the school year developed, whereas use of the BLIND strategy tended to wane.
CORRESPONDENCE strategies increased until T3 and then decreased. At no point were the other three strategies (TEXT,
TITLE, and VISUAL) particularly frequent, and they did not undergo marked development during the course of the school year.
Multidimensional Scaling Analysis (MSA) was performed in order to go beyond the isolated analysis of each strategy
individually. The method used was Correspondence Factorial Analysis (CFA) (Benzecri, 1980; for a discussion of the advantage
of the method in developmental studies, see de Ribaupierre, & Rieben, 1985; Lautrey, de Ribaupierre, & Rieben, 1986), which
permits the joint study of the

Page 302
TABLE 15.1
Percentages of Word-Search Strategies From T1 to T4
BLI TIT TEX VIS COR COP ORT
T1 24.0 7.0 6.0 5.0 16.0 23.0 19.0
T2 15.0 3.0 12.0 8.0 23.0 22.0 17.0
T3 8.0 1.0 8.0 4.0 24.0 23.0 32.0
T4 11.0 4.0 6.0 5.0 14.0 28.0 32.0
Note. BLI = Blind; TIT = Title; TEX = Text; VIS = Vosia; COR =
Correspondence; COP = Copy;
ORT - Ortho.
strategy and participant spaces. The first extracted factor (which explains 33.5% of the variance) contrasted the very earliest
strategies (BLIND) with the most highly developed (ORTHO and COPY) while passing via TITLE, VISUAL, TEXT, and
CORRESPONDENCE strategies; the second factor (which explains 23.9% of the variance) distinguished between the strategies
that result from a process of automatization or memorization (COPY, ORTHO, TITLE) and those that demand active problem-
solving behavior (CORRESPONDENCE, TEXT, VISUAL).
Turning to individual participants, we were able to compare their trajectories as defined by their position at T1, T2, T3, and T4
in the four quadrants formed by the space of the first two factors. In general, these trajectories corresponded to the steps
described in the stage-based models: transition from the strategies of prereaders (BLIND, TITLE), through the nonautomatized
strategies of beginning readers (VISUAL and TEXT logographic phase and reference to context) followed by nonautomatized
strategies based on phoneme grapheme correspondence (CORRESPONDENCE alphabetic phase) to, finally, the automatized
strategies of quasi-readers (ORTHO, COPY alphabetic/orthographic phase). It is important to note that the phases described
earlier cover the period of learning to read in the narrow sense and exclude a period during which this new knowledge is
consolidated. This is because, on the one hand, the observed strategies apply to a still limited lexicon bearing on a familiar text
and, on the other, because these strategies still result in frequent word-identification errors.
Moreover, the children observed in this study differed in the speed with which they made the transition between the various
phases and did not make equal use of all the strategies characterizing the individual phases. We hypothesized that the
participants may also differ in the shape of their trajectory, with some children omitting the phase in which strategies linked to
the alphabetical code are dominant. This hypothesis finds limited support in the results, although it is not possible to assert that
these children totally "skipped" the alphabetical phase. When searching for words, most children

Page 303
made frequent use of the alphabetical code even if in class they received no systematic instruction in the rules of phoneme
grapheme correspondence.
However, the presence of acquisitional trends should not mask the existence of within-subject variability. Although the
strategies correspond to different levels of learning, the children did not make exclusive use of a single type of strategy during
any given period of observation. Instead, we noted the presence of considerable within-subject variability in strategy use, which
is reflected in the high median values for the number of different strategies used during any given period. For the 7 types of
strategy, the median number of different strategies used was 4 at T1 and 5 at T2, falling to 4 again at T3 and T4. Thus, children
exhibited considerable flexibility during the early part of the school year when the most elementary strategies persisted, despite
the fact that more advanced strategies were beginning to appear. This flexibility may result from differences between the words
that the children processed. However, it may also be considered as a characteristic of learners during the early phases of
learning.
Word-Copying Strategies
Turning now to the question of word-copying, strategy categorization was based both on qualitative (nature of the sequence of
letters transported from the RT to the children's individual sheet of paper3) and quantitative (number of letters in the sequence)
criteria. In most cases, we observed children's physical movement from their table to the RT, and it was easy to note how many,
and which letters, they collected in one move. However, when a child was sitting nearer to the RT, the reference to the text was
established on the basis of head movements. We were able to distinguish between 7 types of strategy embracing the following
14 individual strategies: (a) letter-by-letter transportation (either LETTER-U, letters that were unknown by the child, or
LETTER-K, known letters). The distinction was established in view of the results of letter-knowledge tasks used in parallel at
T1, T2, T3, and T4; (b) transportation of double letters (DOUBLE, two identical letters e.g., the nn of bonnet; DB+, two
identical letters plus another letter e.g., nne of bonnet); (c) transportation of digrams (DIGRAM e.g., an of guirlande); (d)
transportation of a syllable (SYLLABLE2, two-letter syllable; SYLLABLE3, syllable of three letters or more); (e) transportation
of morphemes (MORPHEME1, single letter e.g., plural s of enfants4; MORPHEME2, two letters e.g., es of petites;
MORPHEME3, three letters or more e.g., ait of
3 This is a slightly delayed copy because the children transpose the letters or clusters of letters from the text displayed on
the classroom wall to their own sheet of paper.
4 MORPHEME1 strategy was not counted if the remaining part of the word was transported on a letter-by-letter base.

Page 304
finissait); (f) transportation of a word (WORD2, two-letter word; WORD3 word of three letters or more); (g) transportation of
other groups (OTHER2, two letters; OTHER3, three letters or more).
For the 21 subjects and four periods of observation, we counted 1,989 instances of strategies (the judges achieved 92%
agreement in their assessment of similar data drawn from another sample). Table 15.2 presents the percentages for the 14
categories of strategy at points T1 to T4.5 Two of these categories occurred with a relatively high frequency throughout the
year: LETTER-K (which reached its maximum at T2 and then decreased regularly as more complex strategies emerged) and
WORD2. LETTER-U strategy was observed at T1 only. All the other strategies occurred less frequently and were not
characterized by any marked development during the course of the school year.
In order to study the relations between the strategies and the participants' progression, Factorial Correspondence Analysis (FCA)
was performed using the same procedure that was applied for the search strategies. The first two factors explained 43.5% and
14.5% of the variance respectively. The first revealed a very clear contrast between transportation of an unknown letter by
unknown letter (LETTER-U) and all the other strategies, and this factor might be interpreted as representing the transition from
a phase of logographic spelling (absence of code knowledge) to the alphabetical spelling phase. The second distinguished
between the strategies as a function of the number of letters transported together in one move, and it may be supposed that this
factor accounts for a gradual progression from partial alphabetic cues to more complete orthographic cues. We also expected a
more qualitative differentiation that would distinguish between strategies based on submorphemic constraints (LETTER-K and
SYLLABLE) and those dependent on morphemic constraints (MORPHEMES and WORDS). Such an interpretation could be
derived from the third factor, which explained only 8.2% of the variance.6 Given the age and school grade of the children, it is
not very surprising that the marks of the transition from the alphabetic to the orthographic phase are not yet very substantial.
Within the two-dimensional space
5 There were two ways of analyzing these strategies, one based on word-specific level and the other based on the
language system level. An example taken from our observations helps clarify this distinction. During observation period
T3, child 50 transposed the French word garon in three blocks: gar - - on. If analysis was conducted with reference to
the word, this copy will be analyzed as SYLLABLE3 LETTER-K DIGRAM. However, if the whole language system
provided the reference, then the final block may also be interpreted as WORD2 because, in French, "on" is a simple,
common word that may well be known to the child. We consequently performed two sets of analyses and discovered that
the results were not fundamentally different. In this chapter, we refer to results obtained using language level as the
reference.
6 Another Correspondence Factorial Analysis, grouping together all the strategies without taking account of the number of
letters transposed, promotes this type of factor to second place. However, it still accounts for only 18.8% of the variance.

Page 305
TABLE 15.2
Percentages of Copying Strategies From T1 to T4
LTU LTK DBL DB+ DIG OT2 OT3 SY2 SY3 MR1 MR2 MR3 WO2 WO3
T1 28.0 33.0 3.0 1.0 0.5 4.0 1.0 7.5 2.5 1.0 1.0 0.5 14.0 3.0
T2 9.0 38.0 1.5 2.5 3.0 4.0 3.0 6.0 4.0 3.0 5.0 0.5 16.0 5.0
T3 6.0 31.0 6.0 4.0 2.5 4.0 4.0 6.0 4.5 4.0 3.0 3.0 15.0 6.5
T4 7.0 25.0 1.0 0.0 4.0 6.0 5.0 8.0 7.0 4.0 6.0 2.0 16.0 9.0
Note. LTU = Letter-U; LTK = Letter-K; DBL = Double; DB+ = Double+; DIG = Digram; OT2 = Other2; OT3 = Other3; SY2 =
Syllable2; SY3 = Syllable3; MRI =
Morphemel; MR2 = Morpheme2; MR3 = Morpheme3; WO2 = Word2; WO3 = Word3.

Page 306
formed by the first two factors, we observed the progression of the children through the quadrants that defined the following
phases: (a) transportation of unknown letter by unknown letter; (b) transportation of double letters; (c) transportation of known
letter by known letter; (d) transportation of clusters consisting of two, three, or more letters. An unexpected finding was both the
relatively high frequency and the precocious emergence of double letter transportation. It was a salient feature of some words
that children took into consideration even when they ignored the name or the sound of the double letters. Could this very early
sensibility to double letters be related to the observation made by Caramazza and Miceli (1990) showing that geminate letters
were extremely resistant to errors in a patient presenting an orthographic buffer deficit? (See Zesiger & de Partz, chap. 3, this
volume).
As was the case in our analysis of word-search strategies, we noted that there were differences between the participants at T1
and that they progressed through the various phases at different speeds. However, certain qualitative differences were also
observed, with phonographic strategies dominant in some children and orthographic strategies dominant in others.
As was the case for word-search strategies, the children made use of a wide range of different strategies during each of the
periods of observation. With a maximum number of 14 types of strategies, the median number of different strategies used was 5
at T1, 7 at T2, 11 at T3 and 9 at T4. The greatest variability was observed in the middle of the school year (at T3 and, therefore,
slightly shifted with reference to the search strategies for which this phenomenon was observed at T2). These results were
perfectly compatible with those obtained in connection with word search and again called into question any model that holds
that the stages are defined by the exclusive use of a particular type of strategy.
The Relations Between Word-Search Strategies and Word-Copying Strategies
Two complementary methods were used to study the relations between word search and word copying. First, we compared the
results of the separate analyses of the two strategy categories. Second, we presented the results of an FCA consisting of a joint
analysis of the two categories of strategy (7 in the case of word search, 14 in the case of word copying). The two types of
analysis are not redundant. The first relies on a priori correspondences between two systems of classification established
separately; the second tends to provide a posteriori empirical validation of the relations between the whole set of strategies.
As far as the separate analyses are concerned, the development of both word-search strategies and word-copying strategies can
be described in terms of four operationalized phases represented by the quadrants defined by the first two factors of each FCA.
These phases correspond quite closely

Page 307
to the models described in the literature and may be defined as follows: prereading/spelling (BLIND for word search; LETTER-
U for word copying), logographic (TITLE-TEXT-VISUAL for word search; DOUBLE for word copying), nonautomatized
alphabetical (CORRESPONDENCE/for word search; LETTER-K for word copying), and automatized alphabetical/orthographic
(ORTHO-COPY for word search; any clusters of two or more letters for word copying). This progression relates only to an
initial phase in the learning of reading and spelling at the end of which the number of words stored in lexical memory is still
limited. In other words, identifying certain orthographic strategies with reference to certain words in no way signifies that we are
dealing with expert readers/writers.
Using the participants' positions in the quadrants defined by the two Factorial Correspondence Analyses mentioned previously,
we can construct double entry tables in which the horizontal entries contain the results in connection with word search (1, 2, 3,
4) and the vertical entries present the word copying results (A, B, C, D). Table 15.3 indicates participant frequencies7 in the
various cells for each of the periods of observation (T1, T2, T3, T4).
A strictly stage-based model predicts that only the diagonally aligned cells will contain observations. The cells located below the
diagonal correspond to a situation in which word copying was more advanced than word search, and the cells above the
diagonal correspond to a situation in which word search was further advanced. According to Frith's (1985) model, the presence
of observations in cells departing from the diagonal in one direction or the other depending on the stage of acquisition is
acceptable. Figure 15.1 illustrates the difference between these two models. Whereas Frith's model attempted to describe literacy
acquisition occurring over a few years, we are using it here to make predictions on changes appearing over a shorter period of
time. For our purposes, the best opportunity to observe the entire range of strategies is covered by a time interval slightly longer
than one year and situated, for the average children, between 4-6.5 and 5-6.6 years. The very first use of most sophisticated
strategy orthographic in Frith's stage theory (1985) and full alphabetic in Ehri's (1995) more recent phase model appears near
the end of this period and allows children to read a few words by sight, even irregular words, if they are familiar and were
encountered frequently. In other words, we assume that an "accelerated" progression over the word-reading phases can be
observed when limited to words encountered early and recurrently. A strong argument for this view comes from the fact that
during the same amount of time, most children progress dramatically in letter knowledge, in phonological awareness, and in
their understanding of the alphabetic system.
7 The number of children entered in these tables varies because some children were absent at different times during the
school year.

Page 308
Fig. 15.1.
Theoretical models of the relationship between reading and spelling of words.
Note: PRE = Preliterate; LOG = Logographic; ALP = Alphabetic; ORT = Orthographic.

Page 309
As Table 15.3 shows, only 65% of the participants occupy the diagonal at T1, 50% at T2, 61% at T3, and 63% at T4. What is
more, if we follow the development of these same participants through the four tables, we find that of the 19 participants, 21%
are always positioned on the diagonal, 37% display more highly developed searching strategies than copying strategies, and 21%
display more highly developed copying patterns than searching patterns during at least one period of observation. Finally, in the
remaining 21%, copying is sometimes ahead of searching, whereas at other times, searching overtakes copying. The dcalages
are not systematic in the direction predicted by Frith's (1985) model. Instead, only slightly more than half of the dcalages
observed (16 out of 29) correspond to this model. These results provide no support for either the strictly stage-based models or
for Frith's model. Instead, they conform more closely to a model that postulates an interactive construction and presupposes that,
at the margins of the acquisitional trends, there is considerable flexibility in the strategies that are undergoing construction.
An alternative method of analyzing these links between searching and copying considers each participant's pattern of search and
copying strategies within a single Factorial Correspondence Analysis. In terms of the strategies, this analysis enables us to check
whether the correspondences established between behaviors in the contingency tables possess any empirical confirmation. When
we turn our attention to the participants, this analysis enables us to develop a description of the global progression of
reading/spelling in each child. Figure 15.2 consists of a projection of each strategy (indicated by its code) and each participant at
T1, T2, T3, and T4 into the space defined by the first two factors. Each child appears four times in Fig. 15.2 (the G1 participants
are indicated by a black square, the K2 participants by a cross), with the exception of two children who were present at only
three of the four observation periods. The first (horizontal) factor, which explains 33.5% of the variance, orders the strategies in
the sequence in which they are acquired: on the left, where the most elementary strategies are located, we find LETTER-U (for
copying) and BLIND (for search), whereas at the opposite end, corresponding to the most developed strategies, we find the
transportation of clusters of three or more letters, MORPHEME3, WORD3, SYLLABLE3, and OTHER3, for word copying, and
ORTHO and COPY for word search. Between these two poles, we find all the other strategies in the order in which they
become dominant during the school year. The second (vertical) factor explains 12% of the variance and can be interpreted in a
way similar to that proposed for word search. In effect, it contrasts those strategies that essentially require the mobilization of
memorization processes involving both long-term and working memory (WORD3, SYLLABLE3, OTHER3 for copying and
ORTHO in the case of word search) with strategies requiring active problem-solving behavior that was

Page 310
Fig. 15.2.
Factorial Correspondence Analysis: location of strategies and subjects in the
space defined by Factor 1 (horizontal) and 2 (vertical).
more specific to searching than copying (CORRESPONDENCE, TEXT, VISUAL for word search and LETTER-K for word
copying). If we now examine the situation of each strategy within the two-dimensional space, we find that the grouping revealed
by this analysis corresponds to that established earlier on the basis of the two separate analyses.
Again, in comparison with the separate analyses, the order of the search strategies is retained for both factors, whereas the order
of a number of very close strategies, that is, strategies with extremely similar coordinates, is inverted. Even though there is no
important modification of the order, it is interesting to note that the TITLE strategy, which appears to be one of the strategies
involving memorization when viewed in terms of the subset of search strategies, ranks alongside the problem-solving strategies
when viewed in terms of the overall set of strategies (closer to VISUAL and TEXT than to BLIND on the second factor). On the
basis of the interpretation of the first two factors, the quadrants can be characterized as follows: (a) automatic
prereader/prewriter strategies; (b) nonautomatized logographic strategies; (c) nonautomatized alphabetical strategies; (d)
partially automatized alphabetic/orthographic strategies of quasi-readers/quasi-writers. When we examine the position of the
participants, Factorial Correspondence

Page 311
TABLE 15.4
Subjects' Trajectory From T1 to T4
Subjects T1 T2 T3 T4
58, 53, 57 IV IV IV IV
55 I IV IV IV
08, 54 III III IV IV
52, 10 II II IV IV
56, 59 III III III IV
04 II II IV III
01, 03 II III III III
09 I III III III
05 II II - III
02 I II IV II
50 II I II II
11 II II III I
07 II - II I
06 I II I I
51 I I II I
Analysis clearly reveals the differences between the K2 and G1 children, most of the latter being found in quadrant IV. This
reflects their schooling differences: Children at level G1 have already experienced many activities related to written language,
whereas K2 children have been primarily engaged in activities involving spoken language.
It is possible to identify each child's progress throughout the year by studying the trajectory defined by their positions within the
two-factor space at T1, T2, T3, and T4. By way of example, Fig. 15.2 presents the trajectories of three children. Child 55 (G1)
exhibited rapid progress. In this subject, prereader/prewriter strategies (LETTER-U and BLIND) were dominant at T1, with
quasi-reader/quasi-writer strategies approaching dominance at T2 before becoming clearly established at T3 and T4. Also
interesting is that this child did not pass though a phase in which active problem-solving strategies of the logographic or
alphabetical type were dominant. Participant 03 (K2) exhibited a slower acquisition pattern and, more particularly, passed
through a phase at which logographic strategies were dominant (T2) and a long period during which alphabetic strategies
dominated (T2 and T3) before approaching quadrant IV at T4. Finally, participant 06 (K2) provided an example of a child who
did not move from quadrant I throughout the entire school year.
Table 15.4 presents the trajectories of the 21 participants.8 Their starting points at T1, their trajectories, and their points of
arrival at T4 vary considerably. The participants were classified into three groups depending on their level at T4 the advanced
group reaching level IV, the average group reaching
8 The K2 children are coded from 01 to 11, the G1 children from 50 to 59.

Page 312
level III, and the slow group reaching level I or II. A study of the progression patterns allows us to check, at the level of
individual participants, our model postulating a movement away from quadrant I toward quadrant IV. Of the 21 participants, 14
(67%) exhibit ''error-free'' trajectories, without regression. It is significant that of the 7 subjects who exhibit regressions, 6 are
found in the slow group. Our hypothesis is that these unstable patterns can be explained equally in terms of the individual
characteristics of school children experiencing a greater-or-lesser degree of learning difficulty and in terms of the less reliable
observations that result from the fact that these participants have searched for and written fewer words than the others.
The Relations Between Strategies and the Reading and Writing of Words
In order to confirm the relevance of word-search strategies and word-copying strategies for the study of beginning
reading/spelling, Table 15.5 presents the results obtained at T4 by the three groups previously defined in a word reading (12
items) and pseudoword reading task (12 items) taken from Alegria, Content, and Leybaert (1989). It can be seen that the three
groups differ in the results achieved in these two tasks, which are the classic tools for the study of reading. In the absence of any
similar tests for the evaluation of spelling, an indication of progress was provided by the number of words that the children
wrote correctly without having searched for them in the reference text. Table 15.5 shows that the advanced group is clearly
distinguished from the other two groups.
TABLE 15.5
Results in the Reading Tests for the Fast, Medium and Slow Groups
Fast Group Medium Group Slow Group
Q IV at T4 Q III at T4 Q I or II at T4
n = 10 n=5 n=6
Pseudoword Reading Median 8 2 0
Min-Max 5-12 0-5 0-0
Word Reading Median 10 2 0
Min-Max 4-12 0-3 0-0
Word Spelling Median 7 1 2
Min-Max 3-14 0-4 0-5
Conclusions
Let us return to the question of the tools that enable us to describe the development of reading/spelling in learners at the very
beginning of acquisition. Because the entries in their lexicon are both limited and idiosyncratic,

Page 313
we were dissuaded from studying young children simply on the basis of word-reading and spelling tests. The analysis of the
relationships between the strategies and the reading and writing of words and pseudowords has revealed a high degree of
concordance, thus validating our method.
Regarding the validity of stage-based models, our longitudinal study of word-search strategies and word-copying strategies has
revealed a relatively clear acquisitional framework. However, there was considerable variability within each participant at each
observation point. These two observations might be seen as contradictory because the greater the variability, the more unlikely it
is that the participant is at any particular stage. However, other empirical results have required complexity of numerous sources
of variation for description of developmental facts (Lautrey, 1990; Reuchlin, 1978; Rieben, 1995b; Rieben, de Ribaupierre, &
Lautrey, 1990; Siegler, 1989, 1995). As far as the learning of the reading and writing of words is concerned, we conclude that
the notion of stage cannot accommodate the data. Instead, the developmental trends show what we call phases of dominance.
Because our results suggest considerable flexibility of strategies, phases of dominance must allow the simultaneous use of
elementary and more advanced strategies. In particular, the coexistence of alphabetical and orthographic strategies seems to lend
more support to a theory of amalgamation such as that proposed by Ehri (1980, 1989, 1991a) than to a model that sees a clear
distinction between these two stages (Frith, 1985). Furthermore, our data give support to the idea of qualitative changes that are
simultaneously discontinuous (to explain the emergence of new strategies) and continuous (to explain the co-occurrence of
strategies from different levels). Viewed from this perspective, our data are in agreement with Perfetti's (1991) conceptualization
of continuous changes in the quality of the mental lexicon through the mechanism of an increase in the level of precision and
redundancy.
If we consider the flexibility in strategy use as a psychological phenomenon rather than just background noise, then we must
also find an explanation for it. Explanations are possible at a variety of levels. First, at the level of the participant, variability
may be a general functional characteristic of all learning. Such variability has an adaptive value and is an indispensable element
in any explanation of the emergence of novelty (Piaget, 1975) or an explanation of the way general knowledge interacts with
specific knowledge (Siegler, 1989).
A second source of variability can be identified in teaching methods whose influence remains to be adequately defined.
However, the work of Seymour and Elder (1986) showed that when the teaching method focuses on the visual learning of a
word list, logographic strategies are clearly dominant, whereas in a class in which the alphabetical code is systematically taught
during the course of the school year, these logographic strategies coexist with alphabetical strategies. The school, and in
particular the

Page 314
class in which our observations were gathered, was characterized by its openness to different pedagogical trends. Despite the
fact that the teacher proposed an approach to written language based on oral expression and favored meaning-based language
processing, code knowledge, which was not systematically taught, was most clearly present in interactions between teachers and
students. Such an environment may contribute significantly to the high level of intrasubject variability that we have identified.
A third source of variability resides in the linguistic material, that is, the words. Although effects due to frequency and regularity
are well documented, we do not consider that they alone are able to account fully for the flexibility observed at the very
beginning of the acquisition of reading and spelling. The reason for this is that the first written words memorized by children are
probably not representative of the set of words belonging to a given language in terms of their frequency or regularity. With the
exception of words such as daddy, mummy, utility words (the, a, some, etc.) or certain environmental words (Coca Cola, Lego,
etc.), the mental lexicons of children aged 4-to-5 consist of subject-specific items, for example, their first names, or surnames,
or words encountered in connection with certain books. In fact, what is needed in the field of written language is research
equivalent to that which we possess in connection with spoken language (Nelson, 1973), studies that describe, within a
longitudinal perspective and for a set of individual cases, what words children attempt to read/write and how they set about
doing so. Such research makes it possible to date words in the order in which they enter lexical memory and observe their
specific strategic history. Furthermore, such studies help explain within-subject and between-word variability at the beginning of
learning.
We return finally to the central question of the relations between the reading and writing of words. The concordance obtained
between the study of strategies and the reading of words and pseudowords justifies our belief that comparisons of word-search
strategies and word-copying strategies constitute a proxy for comparisons of reading and spelling and contribute to the debate
on the dependence or independence of the two activities. The groupings of the two sets of strategies resulting from the Factorial
Correspondence Analysis exhibited sufficient coherence at each phase to support the idea of a common acquisitional path.
However, as for each of the families of strategies analyzed separately, the results show a high degree of variability that
sometimes sees reading move ahead of spelling, and vice versa. In this regard, our results do not support the hypothesis that
beginning reading and spelling develop in perfect synchrony or the hypothesis of alternating dcalages as proposed in Frith's
(1985) model.
To conclude, we want to emphasize the rich potential of the observations that can be made in classroom situations and the
importance of interpreting data relating to the learning of reading or spelling in full knowledge of the

Page 315
context within which learning takes place. We believe that these advantages far outweigh the drawbacks that result from the fact
that the control of variables in such a situation is a somewhat more delicate task. It is clear that this is not a question of
opposing experimental studies to field studies but far more a question of viewing the two as complementary parts of the same
whole.
Acknowledgments
The research that is presented here would not have been possible without the active cooperation of the students of the Maison
des Petits and their teacher, Nicole Elliott, to whom we owe our sincere thanks. We should also like to thank Nilima
Changkakoti, Colette Grobty, Martine Mornacchi, Ladislas Ntamakiliro, Christiane Perregaux, and Gianreto Pini for their
assistance in the gathering and analysis of this data, and Mara Georgi for editing assistance in English. We are also grateful to
Linnea Ehri for her insightful comments.
References
Alegria, J., Content, A., & Leybaert, J. (1989). The development of reading mechanisms during the first school years: Effects of
the teaching method. Unpublished document. Universit Libre de Bruxelles.
Barron, R. W. (1986). Word recognition in early reading: A review of the direct and indirect access hypotheses. Cognition, 24,
93-119.
Barry, C. (1994). Spelling routes (or roots or rutes). In G. D. A. Gordon & N. C. Ellis (Eds.), Handbook of spelling: Theory,
process, and intervention. Chichester: Wiley.
Benzecri, J. P. (1980). L'analyse des donnes [The practice of data analysis]. Paris: Dunod.
Bruck, M., & Waters, G. S. (1990). An analysis of the component spelling and reading skills of good readers good spellers,
goods readers poor spellers, and poor readers poor spellers. In T. H. Carr & B. A. Levy (Eds.), Reading and its development (pp.
161-206). San Diego: Academic Press.
Bryant, P., & Bradley, L. (1980). Why children sometimes write words which they cannot read. In U. Frith (Ed.), Cognitive
Campbell, R. (1987). One or two lexicons for reading and writing words: Can misspellings shed any light? Cognitive
Clesse, C. (1977). Apprendre lire en parlant. Exprimentation dans un cours prparatoire [Learning to read by speaking.
Experimentation in first grade]. In L. Lentin (Ed.), Du parler au lire (pp. 91-152). Paris: ESF.
Ehri, L. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive processes in spelling (pp. 311-338).

Page 316
Ehri, L. (1989). Movement into word reading and spelling. How spelling contributes to reading. In Mason (Ed.), Reading and
writing connections (pp. 65-81). Boston: Allyn & Bacon.
Ehri, L. (1991a). Learning to read and spell words. In L. Rieben & C. Perfetti (Eds.), Learning to read: Basic research and its
implications (pp. 57-73). Hillsdale, NJ: Lawrence Erlbaum Associates.
Ehri, L. C. (1991b). The development of reading and spelling in children: An overview. In M. Snowling & M. Thomson (Eds.),
Dyslexia: Integrating theory and practice (pp. 63-79). London: Whurr.
Ehri, L. C. (1991c). Development of the ability to read words. In R. Barr, M. Kamil, P. Mosenthal, & P.D. Pearson (Eds.),
Handbook of reading research, Vol. II (pp. 383-417). New York: Longman.
Ehri, L. C. (1995). Phases of development in learning to read words by sight. Journal of Research in Reading, 18, 116-125.
Ehri, L., & Sweet J. (1991). Fingerpoint-reading of memorized text: What enables beginners to process the print. Reading
Ehri, L., & Wilce, L. S. (1985). Movement into reading: Is the first stage of printed word learning visual or phonetic? Reading
Ellis, A. W. (1984). Reading, writing and dyslexia. London: Erlbaum. (trad. Delachaux & Niestl, 1989).
Ferreiro, E. (1978). What is written in a written sentence: A developmental answer. Journal of Education Boston University,
160, 4, 25-39.
Fijalkow, J., & Liva, A. (1988). La copie de texte comme indicateur de l'apprentissage de la langue crite par l'enfant [Text
copying as indicator of literacy acquisition in children]. European Journal of Psychology of Education, 3, 431-449.
Academic Press.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, J. Marshall, & M. Coltheart (Eds.), Surface
Glushko, R. J. (1981). Principles for pronouncing print: The psychology of phonography. In A.M. Lesgold & C.A. Perfetti
(Eds.), The Acquisition of reading (pp. 61-84). Hillsdale, NJ: Lawrence Erlbaum Associates.
Goswami, U., & Bryant, P. (1992). Rhyme, analogy, and children's reading. In P. Gough, L. Ehri, & R. Treiman (Eds.), Reading
acquisition (pp. 49-63). Hillsdale, NJ: Lawrence Erlbaum Associates.
Gough, P.B., Juel, C., & Griffith, P. L. (1992). Reading, spelling, and the orthographic cipher. In P.B. Gough, L. Ehri, & R.
Hbrard, J. (1977). Rle du parler dans l'apprentissage de l'crit [Role of speaking in learning to read]. In L. Lentin (Ed.), Du
parler au lire (pp. 57-90). Paris: ESF.
Henderson, E.H. (1992). The interface of lexical competence and knowledge of written words. In S. Templeton & D.R. Bear
(Eds.), Development of orthographic knowledge and the foundations of literacy (pp. 1-30). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Humblot, L., Fayol, M., & Lonchamp, A. (1994). La copie de mots en CP et CE1 [Word copying in first and second grade].
Reperes, 9, 47-57.
Lautrey, J. (1990). Esquisse d'un modle pluraliste de dveloppement cognitif [Sketch of a pluralistic model of cognitive
development]. In M. Reuchlin, J. Lautrey, C. Marendaz, & T. Ohlmann (Eds.), Cognition: L'individuel et l'universel (pp. 185-
216). Paris: PUF.
Lautrey, J., de Ribaupierre, A., & Rieben, L. (1986). Les diffrences dans la forme du dveloppement cognitif valu avec des
preuves piagtiennes: Une application l'analyse des correspondences [Differences in the form of cognitive development
assessed with Piagetian tasks: An application of correspondence analysis]. Cahiers de Psychologie Cognitive, 6, 575-613.
Marsh, G., Friedman, M., Welch, U., & Desberg, P. (1981). A cognitive-developmental theory of reading acquisition. In G. E.
Mackinnon & T. G. Waller (Eds.), Reading research: Advances in theory and practice, Vol. 3 (pp. 355-370). New York:
Academic Press.

Page 317
Morris, D. (1992). Concept of word: A pivotal understanding in the learning-to-read process. In S. Templeton & D. R. Bear
(Eds.), Development of orthographic knowledge and the foundations of lateracy (pp. 53-77). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society for Research in Child Development,
38(1/2, Serial No. 149), 1-35.
Peereman, R. (1991). La mdiation phonologique dans la reconnaissance des mots [Phonological mediation in word
recognition]. In R. Kolinsky, J. Morais, & J. Segui (Eds.), La reconnaissance des mots dans les diffrentes modalits
sensorielles: tudes de psycholinguistique cognitive (pp. 119-163). Paris: PUF.
Perfetti, C. A. (1991). Representations and awareness in the acquisition of reading competence. In L. Rieben & C. Perfetti
(Eds.), Learning to read: Basic research and its implications (pp. 33-44). Hillsdale, NJ: Lawrence Erlbaum Associates.
Perfetti, C. A. (1992). The representation problem in reading acquisition. In P.B. Gouch, L. C. Ehri, & R. Treiman (Eds.),
Perfetti, C. A., Zhang, S., & Berent, I. (1992). Reading in English and Chinese: Evidence for a "universal" phonological
principle. In R. Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 227-248): Amsterdam: North-
Holland.
Piaget, J. (1975). L'quilibration des structures cognitives: Problme central du dveloppement [The development of thought:
Equilibration of cognitive structures]. Paris: PUF.
Reuchlin, M. (1978). Processus vicariants et diffrences individuelles [Vicarious processes and individual differences]. Journal
de Psychologie Normale et Pathologique, 2, 133-145.
Ribaupierre, A. de, & Rieben, L. (1995). Individual and situational variability in cognitive development. Educational
Psychologist, 30, 5-14.
Rieben, L. (1989). Individual differences in word recognition acquisition: A path from an interactive model of reading to an
interactive instructional setting. European Journal of Psychology of Education, 4, 329-347.
Rieben, L. (1991). Les modles en "stades" de l'acquisition de la lecture [Stage-based models in learning to read]. Les Cahiers
de Beaumont, 52/53, 33-37.
Rieben, L. (1995a). Word-search strategies in young children inside and outside the classroom. In L. Verhoeven & A.
Teberosky (Eds.), Proceedings of the Workshop on Understanding Early Literacy in Developmental and Cross-Linguistic
Approach (pp. 137-165). Strasbourg: European Science Foundation Scientific Networks.
Rieben, L. (1995b). Diffrences individuelles dans la reconnaissance des mots crits chez l'enfant [Individual differences in
word recognition by children]. In J. Lautrey (Ed.), Universel et diffrentiel en psychologie (pp. 193-221). Paris: PUF.
Rieben, L., de Ribaupierre A., & Lautrey, J. (1990). Structural invariants and individual modes of processing: On the necessity
of a minimally structuralist approach of development for education. Archives de Psychologie, 58, 29-53.
Rieben, L., Meyer, A., & Perregaux, C. (1991). Individual differences and lexical representations: How five 6-year-old children
search for and copy words. In L. Rieben & C. Perfetti (Eds.), Learning to read: Basic research and its implications (pp. 85-
101). Hillsdale, NJ: Lawrence Erlbaum Associates.
Rieben, L., & Saada-Robert M. (1991). Developmental patterns and individual differences in the word-search strategies of
beginning readers. Learning and Instruction, 1, 67-87.
Saada-Robert, M., & Rieben, L. (1993). Evolutions des stratgies d'criture-copie et units graphiques du franais [Development
of word copying strategies and French graphic units]. Etudes de Linguistique Applique, 90, 84-96.
Segui, J. (1991). La reconnaissance visuelle des mots [Visual recognition of words]. In R. Kolinsky, J. Morais, & J. Segui
(Eds.), La reconnaissance des mots dans les diffrentes modalits sensorielles: tudes de psycholinguistique cognitive (pp. 99-
117). Paris: PUF.

Page 318
Seymour, P. H. K. (1992). Cognitive theories of spelling and implications for education. In C. M. Sterling & C. Robson (Eds.),
Psychology, spelling and education (pp. 51-70). Clevedon: Multilingual Matters.
Seymours, P. H. K., & Elder, L. (1986). Beginning reading without phonology. Cognitive Neuropsychology, 3, 1-36.
Seymour, P. H. K., & Evans, H. M. (1992). Beginning reading without semantics: A cognitive study of hyperlexia. Cognitive
Siegler, R. S. (1987). The perils of averaging data over strategies: An example from children's addition. Journal of Experimental
Psychology: General, 116, 250-264.
Siegler, R. S. (1989). How domain-general and domain-specific knowledge interact to produce strategy choices. Merrill-Palmer
Quarterly, 35, 1-26.
Siegler, R. S. (1995). How does change occur: A microgenetic study of number conservation. Cognitive Psychology, 28, 225-
273.
Sprenger-Charolles, L., & Casalis, S. (1995). Reading and spelling acquisition in French first graders: Longitudinal evidence.
Reading and Writing, 7, 1-25.
Stuart, M., & Coltheart, M. (1988). Does reading develop in a sequence of stages? Cognition, 30, 139-181.
Treiman, R. (1984). Individual differences among children in spelling and writing styles. Journal of Experimental Child
Zutell, J. (1992). An integrated view of word knowledge: Correlational studies of the relationships among spelling, reading, and
conceptual development. In S. Templeton & D. R. Bear (Eds.), Development of orthographic knowledge and the foundations of
literacy (pp. 213-230). Hillsdale, NJ: Lawrence Erlbaum Associates.

Page 319
Chapter 16
Foundations of Orthographic Development
Philip H. K. Seymour
University of Dundee, Scotland
In this chapter I aim to present the outline of a theory of orthographic development, that is, to provide an account of the
progression from preliteracy to skilled reading and spelling (Perfetti, 1992). I make the simplifying assumption that these two
processes reading and spelling are both dependent on the development of a single central orthographic resource.
Written language systems have been the object of large amounts of linguistic analysis and discussion (Henderson, 1982;
Kavanagh & Mattingly, 1972). A distinction has been made between logographic systems, in which symbols represent whole
words or concepts, and phonographic systems, in which component sounds are represented. Phonographic writing may take the
form of a syllabary or of an alphabet. English is generally held to be an alphabet, although the correspondence between
phonemes and graphemes is far from straightforward. Some people take the view that the script is morphophonemic, implying
that distinctions of spelling are used to indicate lexical identities, word derivations, and morphological structure. One
consequence has been the conclusion that the treatment of English orthography as a simple alphabet is, in principle, inadequate
as an approach to the achievement of competence. Almost any word might, legitimately, be written in a number of ways.
Because each word is assigned a unique and conventionally agreed spelling, it becomes essential to know the precise
arrangement of letters that is appropriate in each case.

Page 320
Individual Variation
The nature of English spelling imposes constraints on the format of theories of the process. For many people, the implication
has been that a theory might need to account for two distinct aspects of competence: (a) the establishment of a knowledge of
sound letter correspondences (representing the alphabetic basis of the language), and (b) the storage of a large amount of word-
specific and morphological information regarding the actual spellings of simple and complex words (the lexical and morphemic
basis). One possibility is a dual-route system in which abstract phoneme grapheme correspondences are separately represented
from lexical and morphological aspects (Ellis & Young, 1988; Morton, 1980, 1989; Seymour & Porpodas, 1980). Others have
argued that the two aspects of orthographic knowledge might be represented within a single network (Brown & Loosemore,
1994; Seidenberg & McClelland, 1989), or as a letter-sequence generator that operates under various types of (phonological,
lexical semantic, and morphological) constraint (Seymour & Evans, 1994a).
A key question in the evaluation of the theoretical accounts is whether significant individual variations in spelling difficulties
are identifiable. If different types of disability occur, affecting mainly the alphabetic basis in some individuals and the lexical or
morphological basis in others, then this will offer strong circumstantial support for models that contain distinct systems or
sources of constraint that might be independently vulnerable. This argument was initially set out by reference to cases of
neurological damage in adulthood (Ellis, 1993; Shallice, 1988; Zesiger & de Partz, chap. 3, this volume). It was found that some
cases presented a phonological pattern in which the ability to write dictated nonwords was lost, although a large vocabulary of
real words was preserved (Shallice, 1981), whereas others presented with a pattern of lexical (or surface) dysgraphia in which
knowledge of specific word spelling had been lost, although the alphabetic basis of spelling was retained (Hatfield & Patterson,
1983).
The position has been less clear-cut with respect to developmental forms of spelling disability. Boder (1973) suggested the
existence of dyseidetic and dysphonetic reading-spelling patterns. However, the dominant proposal has been that literacy
depends on the formation of an orthographic cipher (Gough & Hillinger, 1980) that encodes the grapheme phoneme
correspondences underlying written English. This process is considered to be impaired in cases of dyslexia (Pennington, 1991;
Stanovich, 1988). Gough argued that the primary indicator of the possession of the cipher is the ability to read unfamiliar
nonwords. This has led to the expectation that dyslexic individuals should appear uniquely deficient in their ability to read or
write nonwords (Rack, Snowling, & Olson, 1992).
The argument against this single process theory (Mitterer, 1982) rests mainly on studies of variations in the relative degrees of
difficulty encoun-

Page 321
tered in reading or spelling irregular words (an index of lexical competence) and nonwords (an index of alphabetic competence).
Ellis (1985) suggested that the distinction might be captured by plotting the locations of cases with respect to two theoretical
dimensions: a whole word (lexical) dimension and a phonic (nonlexical) dimension. A few empirical investigations of the actual
scatter of cases on these dimensions have now been reported.
Manis, Szeszulski, Holt, and Graves (1990) obtained scores for phonological processing (e.g., nonword reading) and
orthographic processing (e.g. homophone discrimination) from reading disabled samples and reading-level-matched controls.
Performance of the cases of disability was expressed as a deviation from the control mean on these two dimensions. The results
show evidence of a scatter, with cases differing in the relative degree of impairment on the two dimensions.
Castles and Coltheart (1993) contrasted reading performance on lists of irregular words and nonwords in samples of reading
disabled children and age-matched controls. A regression analysis indicated that some individuals performed at a normal level
with nonwords but were impaired in irregular word reading, whereas others showed the opposite pattern. This was taken as
evidence for the existence of developmental forms of surface dyslexia (impaired acquisition of lexical reading) and of
phonological dyslexia (impaired acquisition of alphabetic reading).
Seymour and Evans (1994a) plotted the scatter of cases of disability on dimensions defined by error scores for nonword spelling
and lower frequency word spelling. Some individuals showed a positive lexicality effect (nonword spelling worse than word
spelling), whereas others showed a negative lexicality effect (low frequency word spelling worse than nonword spelling). This
contrast was reflected in reading, partly in the magnitude of the discrepancy between error rates for words and nonwords, but
also in more detailed features of processing, such as reaction time profiles and word length effects (Seymour & Evans, 1993).
Two contrasting cases, DK and RC, who exemplified phonological and lexical dysgraphia, have now been followed over a
period of years (Seymour & Bunce, 1994) during which time the underlying patterns have persisted.
These studies strongly favor the view that developmental orthographic impairments may selectively affect either the lexical or
the alphabetic aspects of written language. It follows that theoretical accounts should, ideally, include mechanisms that allow for
differential impairments and contrasting patterns of disability.
The Role of Phonological Awareness

A common theme in theories of literacy acquisition has been an emphasis on the contribution of phonological abilities. Most
commentators agree that acquisition of an alphabetic writing system is contingent on the capacity to

Page 322
develop an awareness of speech segments, especially at the phonemic level (Bertelson, 1987; Liberman, Shankweiler, Liberman,
Fowler, & Fischer, 1977; Morais, 1991). It is also broadly agreed that phonemic awareness does not normally arise as a
spontaneous product of maturation (Morais, Cary, Alegria, & Bertelson, 1979) but as a response to demands imposed by
instruction in the alphabetic basis of literacy (Morais, Alegria, & Content, 1987; Perfetti, Beck, Bell, & Hughes, 1987). This
interactive link between phonemic awareness and literacy is supported by correlational studies and some training studies
(Lundberg, Frost, & Petersen, 1988).
One important area of dispute relates to the distinction between phonological skills that are available prior to literacy and those
that only emerge during acquisition. The contribution of the work of Bryant and his colleagues has been to demonstrate that
some phonological skills, especially in the area of rhyming and alliteration, may be available to preliterate children and may also
predict later reading achievement (Bradley & Bryant, 1983; Bryant, MacLean, Bradley, & Crossland, 1990). These early skills
have typically been measured using tasks such as rhyme generation or selection of the odd-man-out in a list of spoken words.
Bryant has raised the question of the relationship between early (preliterate) phonological skills and later development of
alphabetic literacy and phonemic awareness. One possibility is that a pathway can be traced that relates early speech games such
as chanting nursery rhymes to rhyming ability and, thence, to a capacity to develop phonemic awareness during the acquisition
of literacy. Using stepwise, multiple regression techniques, Bryant has presented evidence to support this hypothesis. However,
in other analyses, he has secured evidence that indicates that there may be a contribution from rhyming ability to literacy over
and above the connection between rhyming and phonemic awareness.
This conclusion led Goswami and Bryant (1990) to propose an alternative theory of literacy acquisition in which the role of
rhyme structures is emphasized. The main assumptions of this theory are: (a) Children arrive at the task of acquiring literacy
with rhyming skills in place, (b) rhyme-based phonology provides a basis for categorizing words in terms of similarity of sound
and orthography, and (c) this basis is used to build a lexicon of words by a process of detection of analogies between known
words and unfamiliar words.
It appears, therefore, that there are two accounts of the way in which phonological processes have an impact on literacy. These
can be clarified by referring to the hierarchical model of the internal structure of the syllable (see Fig. 16.1) that has been
developed by Treiman (1992) and others. It is supposed that the syllable subdivides naturally into an onset (the initial consonant
group) and a rime (the vowel and any following consonants). This can be referred to as a two-dimensional (2D) level because it
is based on the supposition that English syllables can be constructed by combining ele-

Page 323
Fig. 16.1.
Hierarchical model of the structure of the syllable.
ments from a list of onsets with elements from a list of rimes. Subdivision of the rime into a vowel and terminal consonant
group defines a further level that can be thought of as three-dimensional (3D) because it is implied that syllables are constructed
by combining elements from lists of initial consonant groups (ICs), vowels (Vs), and terminal consonants (TCs). A fourth level
can be identified at which the IC and TC structures are broken into sequences of phonemes.
The debate regarding the manner in which phonological awareness influences literacy concerns the level within this hierarchy at
which the critical units are represented. On the one hand, there is the view that phonological development naturally progresses
down the hierarchy, from syllables, through onset and rime structures, and eventually to phonemes (Treiman, 1987). It is this
view that motivates Goswami and Bryant's (1990) contention that it is the high-level structures that are important. Against this,
there is the view that it is the lowest phonemic level that is essential because, in an alphabetic language, it is the phonemes that
stand in correspondence with the letters of the alphabet.
Dual-foundation Model of Orthographic Development

In the remainder of the chapter, I discuss a theory of orthographic development that I have previously referred to as a dual-
foundation model (Seymour, 1990, 1993). The theory is intended to provide an account of the development of an orthographic
system (or cipher, in Gough's terms) that takes note of the points that have already been raised, notably: (a) that English
orthography encodes information at lexical and morphemic levels as well as at an alphabetic level, (b) that spelling disabilities
may take different forms that relate to the lexical versus nonlexical contrast, and (c) that linguistic awareness may affect literacy
at an onset rime level as well as at a phonemic level.

Page 324
A diagrammatic representation of the model is given in Fig. 16.2. The model contains five distinct components, each of which
may be thought of as referring to a processor or module. Two of these, the Logographic Process and the Alphabetic Process, are
held to perform a foundation role (i.e., to provide a basis for more advanced developments). This does not mean that their
existence is restricted to the early stages. In fact, they persist over time and make continuing contributions to development. The
third component, Linguistic Awareness, performs an enabling function because interactions with this component are necessary
for development in the other systems. Finally, there are two central structures the Orthographic Framework and the
Morphographic Framework that perform a representational function because it is in these systems that an individual's abstract
knowledge of the orthography is stored. The formation of the central systems depends on the availability of the foundations and
on the possibility of developing appropriate forms or levels of linguistic awareness.
Logographic Process
The term logographic is used here as a reference to a process that is concerned with the direct recognition and storage of words.
The Shorter Oxford Dictionary defines a logography as ''a method of printing with entire words'' and logographic as "consisting
of characters or signs, each of which represents an entire word." The assumption is that English words, although composed of
letters, may be treated as units in this sense. This does not
Fig. 16.2.
Diagrammatic representation of the dual-foundation model of
orthographic and morphographic development.

Page 325
imply adherence to the connotations that have (erroneously in my view) been attached to the word logographic in recent
debates. Hence, it is not equated with recognition of logos or reliance on visual cues even though these primitive processes may
represent the early stages in the development of the process (Seymour & Elder, 1986).
This account of logographic literacy includes both the visual cue and the phonetic cue stages in Ehri's (1992) theory (i.e.
encompasses what Ehri called the alphabetic stage of sight word reading). The process is a complex one that advances in
precision and completeness as development proceeds (Seymour & Elder, 1986; Stuart & Coltheart, 1988). The main point is that
word representations are encoded, normally consisting of partial or complete letter identity information. This process is
considered vital for orthographic development because it is the way in which the exemplars, from which the orthographic
information is to be extracted, are internalized. As new words are encountered, they enter the logographic process in a more-or-
less complete form and provide the data for an ongoing process of orthographic (and morphographic) construction (indicated by
the arrows in Fig. 16.2).
It is hypothesized that, in some individuals, the logographic process may be inherently underresourced. This will limit the scope
for processing words and letter groups holistically and for storing the outcomes of such processes. A logographic impairment of
this kind will be evident as a difficulty in acquiring a sight vocabulary. The long-term consequence will be a distortion in the
development of the orthographic system that is manifest as a surface dyslexic pattern of disabled reading and spelling.
The logographic process is not thought to require phonological organization below the level of segmentation of spoken language
into lexical units. This is suggested by observations that an early focus on whole word methods does not support the
development of speech segmentation. For this reason, no interactive relationships with Linguistic Awareness are indicated in
Fig. 16.2. However, the process might well develop an internal sensitivity to graphotactic constraints. If this occurred, we could
expect to find that logographic readers were able to discriminate between orthographically legal and illegal letter arrays and to
produce orthographically well-formed (though phonologically inappropriate) letter sequences when attempting to write
unfamiliar items.
Alphabetic Process
The alphabetic process is founded on a knowledge of the letters and their equivalent sounds. As such, the term alphabetic
corresponds to the dictionary definition of "any set of characters representing the simple sounds in a language." Acquiring the
letters can be viewed as a second form of logographic learning. The letters are arbitrary signs that have to be linked to their
names. There is some disagreement as to whether these names should

Page 326
be the alphabet names "ay-bee-see-dee," and so on, or whether it is preferable to teach children a set of sound names "ah, buh,
kuh, duh." The point at issue here is which system may be more helpful in allowing children to infer the phonemic equivalent of
each letter. The target phonemes in both systems are often embedded in the names, although in a more systematic and complete
way in the case of the sound names.
The acquisition of the letter identities presents a problem in only a minority of cases of reading disability. The difficulty seems
to arise in applying this knowledge in a systematic way by reading sequentially from left to right, deriving a sequence of sounds,
stripping away the schwa vowels, and organizing the whole into a spoken response. It is this operation of grapheme phoneme
translation that is the essence of the alphabetic process.
In the model, the alphabetic process is represented as standing in an interactive relationship with the phonological component of
linguistic awareness. The essential proposal is that acquisition of the alphabetic process entails the isolation of phonemic
structures in speech. This is consistent with the numerous demonstrations that the emergence of an ability to manipulate
phonemes coincides with the introduction of alphabetic teaching (Alegria, Pignot, & Morais, 1982).
In the early stages, alphabetic processing is likely to be accompanied by overt sounding and appears to be functionally distinct
from the logographic process (Seymour & Evans, 1992). Thus, a familiar word can be recognized directly by the logographic
process, whereas an unfamiliar word is submitted to a sequential process of letter sounding and assembly of a response. The
model allows that the alphabetic process may survive into later development as a procedure that can be used to translate
unfamiliar forms on a letter-by-letter basis.
Many people view a difficulty in establishing the alphabetic process as the key characteristic of reading or spelling difficulty.
This could be due to limitations on the decoding of unfamiliar words or because alphabetic processing is needed in order to
establish attention to the identities and positions of all of the letters contained in a word (Frith, 1985). Others argue that the
grapheme phoneme links, once established, become part of the word recognition process, resulting in lexical access pathways
that are precise, phonemically based, systematic, and overdetermined (Ehri, 1992; Perfetti, 1992; Stuart & Coltheart, 1988).
The mechanism proposed in the present account is slightly different. It is agreed that letter identities may be incorporated into
the logographic process and that this leads to the internalization of increasingly detailed and specific representations. However,
the major contribution of the alphabetic foundation is held to be in the formation of the orthographic framework, especially the
implementation of a core structure that is the basis for subsequent developments. The proposed mechanism is described more

Page 327
fully later in this chapter when the orthographic and morphographic systems are discussed.
Linguistic Awareness
Literacy development is thought to require the establishment of linguistic segments via an interactive process of matching
orthographic segments against speech segments. The nature of the demand to develop new forms of segmentation varies as a
function of developmental level. One possibility is that a distinction is needed between a natural course of (implicit)
phonological development, which proceeds from larger units (words and syllables) through intermediate units (onsets and rimes)
to small units (phonemes), and a demanded sequence of literacy-related explicit awareness that develops in the opposite
direction, from small units toward larger ones (Seymour & Evans, 1994b).
The reason for this conflict is to be found in the priority that is accorded to the alphabetic foundation. As has been discussed,
the formation of the alphabetic process involves an interaction with Linguistic Awareness that has as its primary outcome the
emergence of an awareness of phonemes. Thus, although there might be a natural proclivity for a progressive development from
syllables to onsets and rimes, this tendency is overridden by the demands of acquisition of the alphabetic principle. This does
not mean that higher order structures are not important in literacy development, only that their importance arises later rather than
earlier. The reason for this is that the higher order structures do not become relevant until the stage when the orthographic
framework is being formed.
These points can be clarified by referring back to the hierarchical representation of the syllable shown in Fig. 16.1. What is
suggested is that the natural development of implicit phonological awareness normally proceeds down the hierarchy from whole
syllables through onsets and rimes toward the phonemic level. The acquisition of alphabetic literacy, by contrast, imposes a
demand for the development of explicit phonological awareness of speech segments that starts at the phonemic level and
proceeds upwards through the hierarchy to the three-dimensional and, eventually, to the two-dimensional levels.
Evidence bearing on these arguments has recently been presented by Seymour and Evans (1994b) and by Duncan, Seymour, and
Hill (in press). In the first study, Scottish primary school children were asked to perform a segmentation task under three
different instructions, involving division of a spoken word into two parts, three parts, or as many parts as possible. These
instructions can be equated with the 2D, 3D, and phonemic levels of the syllabic hierarchy (Fig. 16.1). The items were
monosyllables (words or nonwords), some of which contained consonant clusters. The results suggest

Page 328
that preliterate children were unable to segment at any of the three levels. As literacy was acquired during the first 2 primary
school years, there was a gain in segmentation ability that favored small unit (phonemic) division over large unit (onset-rime)
division.
In the second study, children in the first primary year performed a common unit task. A pair of spoken words was presented and
the child was asked to report back the shared segment. This could be a large segment such as the rime, or a smaller segment
including a vowel, consonant, or consonant cluster. The children were more adept at reporting back small shared segments than
large ones. A further experiment tested their ability to pick out the letters in a word that corresponded to a spoken segment.
Children were better able to identify small orthographic units than large ones. These experiments indicate that initial teaching
that includes an alphabetic component tends to direct developing phonological awareness toward the phonemic (small unit)
level.
In the model, a second form of linguistic awareness is hypothesized, referring to the morphological structure of words. This is
needed for the most advanced level of development at which a capacity to spell complex words composed of combinations of
syllables, prefixes and suffixes, is required (Seymour & Evans, 1994a).
Orthographic Framework
The heart of the theory of literacy acquisition is the formation of the orthographic framework. This is viewed as a structure that
encodes a generalized knowledge of the correspondence system together with word-specific features, as has been suggested by
Seidenberg and McClelland (1989). For the purposes of the present discussion, I make the simplifying assumption that this
information refers to English monosyllables and disyllables.
It is assumed that the Orthographic Framework develops progressively, dealing at first with quite simple structures but
undergoing a gradual elaboration that incorporates multiletter structures, including consonant groups and vowel spellings.
During the time in which the framework is being formed, the logographic system contains representations of a wide range of
words, only some of which offer appropriate examples for the orthographic representation that is currently being built up. Thus,
the vocabulary available within the logographic system is typically in advance of the region of the orthography represented in
the orthographic system, at least in the early stages of development. This relationship between the logographic store and the
orthographic framework is somewhat reminiscent of Perfetti's (1992) distinction between a functional and an autonomous
lexicon. However, in the present account, the logographic entries may be relatively complete and well specified, and the
distinction between the systems is that one holds word-specific representations, whereas the other is engaged in the construc-

Page 329
tion of an abstract description of the orthography. Hence, a process akin to Karmiloff-Smith's (1986) notion of representational
redescription may be a closer parallel (Seymour, 1993).
In the model, it is suggested that the development of the Orthographic Framework proceeds through a number of steps or stages.
These are referred to as the core, the intermediate, and the advanced stages. The actual progression is likely to be more complex
than this, involving a larger number of levels and some intervening or transition stages. The origin of the Orthographic
Framework is a core structure. This is assumed to emerge out of the basic grapheme-phoneme knowledge that results from the
formation of the alphabetic process. The supposed mechanism is one in which the simple set of correspondences (i.e., the list of
the letters and their predominant sounds) is reorganized in the light of the hierarchical structure of the syllable (Seymour, 1993).
This entails the segregation of a subgroup of letters, which function as vowels, and the derivation of two sets of consonants,
those which can occur in the initial position and those which can occur in the terminal position. This organization is founded on
the three-dimensional structure of the syllable and might initially refer only to simple vowels and consonants.
The suggestion is that the formation of a core orthographic framework involves a 3D phonological model of the syllable as an
initial consonant (IC), vowel (V), terminal consonant (TC) structure, and the establishment of a parallel orthographic
representation (i.e., the letters or letter groups that predominantly correspond to the elements of the phonological structure). This
structure can be set up by carrying out an internal search for reading words that have the appropriate phonological structure and
that are also represented in the logographic system. As these words are located, their positions in the emerging orthographic
framework are marked. A search might also be made for words in the logographic store that are composed from the appropriate
orthographic elements. Some of these words may be found to be irregular because their pronunciation is in conflict with the
dominant correspondences represented in the framework. It is assumed that exceptional items of this kind are located within the
structure together with an index that signals the divergence from expected pronunciation. Hence, it is assumed that the
framework codes both the general relationship between orthography and phonology and the identities of individual lexical items.
The core is regarded as the essential initial requirement for formation of an orthographic system. Once this has been achieved,
the direct reading of both words and nonwords that satisfy the core definition can be achieved. It should be possible to
demonstrate that all relevant words and nonwords can be read or written without the need for support by alphabetic processes
(sequential letter sounding) and that full generalization from taught to untaught items has occurred. This generalization is one
property that distin-

Page 330
guishes orthographic acquisition from logographic acquisition. The capacity to read items directly and without the need for
sequential sounding of letters is the feature that distinguishes the orthographic process from the alphabetic process.
Beyond the core, orthographic development is held to involve the elaboration of the framework by a process of introduction of
increasingly variable and complex structures. This may be in part a matter of the teaching that is provided. If additional vowel
groups such as oo and ee are explicitly identified, these will be added into the alphabetic process and eventually into the core,
leading to the assimilation of a further set of words that are held in the logographic vocabulary. Other structures may be
acquired more or less naturally. These could include some initial consonant groups and final consonant groups. What is required
in these cases is an elaboration of the three-part phonological representation of the syllable to represent complex consonantal
structures coupled with selection of words from logographic storage and entry of the letter groups into the framework. Other
developments are required to accommodate the final -e vowel spellings and further features of vowel lengthening, including
consonant doubling and the numerous complex vowel spellings that form the main area of inconsistency in English spelling
(Barry & Seymour, 1988).
The precise hierarchy of acquisition of the orthographic structure is not known and may be a matter of individual variation.
However, it seems likely that the process is extended in time and may not be complete until toward the end of primary
schooling. Seymour (1993) presented a function relating reading age to error scores on initial consonant groups, vowels, and
terminal consonant groups that had been classified as belonging to the orthographic core or to the more advanced levels. A
process of progressive internalization extending over the range 6-to-10 years is suggested for the advanced structures in spelling
as well as reading.
The criteria for identification of intermediate or advanced levels of orthographic development are similar to those suggested for
the core. It should be possible to define the orthographic elements that are added to the framework at each level and then to
show that real words or nonwords that contain those elements can be read or spelled by a direct form of processing. In addition,
instruction directed toward the elaboration of the framework at a given level should generalize beyond the particular exemplars
used in order to bring about improvements in the whole structure.
Morphographic Framework
As so far described, the orthographic system is mainly to be viewed as a structure that encodes English mono- (and di-)
syllables. This system is not capable of dealing with words that are composed of two or more syllables,

Page 331
including words that have a complex morphemic structure (a stem combined with prefixes and suffixes). Frith (1985) and
Morton (1989) and others suggested that mastery of these morphemic aspects of written language is the key feature of the
advanced levels of development (misleadingly referred to as orthographic in their accounts). Seymour and Evans (1994a)
provided examples of spelling errors produced when premorphographic children (age about 7 years) attempted to write
morphemically complex words.
In order to accommodate this requirement, the model contains an additional component referred to as the Morphographic
Framework. This might be thought of as being established on top of the Orthographic Framework, that is, as a structure that
references syllabic entities (represented at the lower level) and their combinations. The development of the morphographic level
is, therefore, dependent on the achievement of an adequate structure at the orthographic level. It is also dependent on an
interaction with Linguistic Awareness. This could include a phonological influence at the level of syllabic structure and a
morphemic influence in establishing the status of segments as word stems or affixes. The framework also requires input from
the logographic system where stored word forms provide the exemplars needed to establish the orthographic features of
morphemic and polysyllabic spelling.
Conclusions
In the chapter, I have presented the outline of a theory of orthographic development. The central ideas have been that literacy
acquisition involves the formation of a framework that encodes the abstract properties of written language and that this
development is dependent on contributions from logographic and alphabetic foundations as well as on an ongoing reciprocal
interaction with linguistic awareness.
Stages of Acquisition
A common feature of theories of literacy acquisition has been a belief that the process may involve a progression through a
succession of stages (Ehri, 1992; Frith, 1985; Gough & Hillinger, 1980). This is also true of the dual-foundation model because
it is assumed that: (a) the foundations must have reached a certain level before orthographic development can be initiated, and
(b) orthographic development must have proceeded some distance before the construction of the morphographic framework
becomes possible. However, no necessary sequence in the ordering of the foundation developments is proposed.

Page 332
It is also hypothesized that the construction of the Orthographic Framework is a stepwise process. Under normal circumstances,
the internalization of orthographic information seems to occur automatically and without a requirement for element by element
teaching. However, the theory makes the assumption that some structures, notably the core, are developmentally prior to others,
that morphographic development is contingent on prior orthographic development, and that there is a hierarchy of acquisition,
possibly related to the complexity or consistency or frequency of the orthographic elements. The main implication of the model
for the teaching of reading and the remediation of difficulties is that the materials used should be adapted to support the
hierarchical structure of development.
Reading and Spelling Difficulties
A spelling disability (or dyslexia) is defined as a failure to develop a normal orthographic (or morphographic) framework. Such
failures could arise for one or other of two reasons. One is that the neural resources underlying the framework are impoverished
and that the scope for internalization of information is accordingly restricted. The other is that the development of the
framework is distorted or held back as a consequence of a failure at the foundation level. Both of these situations could produce
differing dyslexic patterns. So far as the model is concerned, the favored proposal is that there may be individual variations
underlying the adequacy of either the logographic process or the alphabetic process.
A logographic impairment could reflect an inability to process and store multiletter segments that compose whole words or
morphemic or orthographic segments. The effect of this impairment is to curtail the range and quality of the exemplars available
to the developing orthographic system. A consequence is that the framework does not internalize word-specific data or
information about spelling structures and tends to retain a restriction to an alphabetic letter-sound based level of representation.
Thus, a logographic impairment is expected to emerge early in development as a difficulty in acquiring a sight vocabulary and,
later on, as a surface dyslexic reading and spelling pattern.
An alphabetic impairment might result from a difficulty in acquiring the letters or from problems affecting the systematic
application of letter sound knowledge. A difficulty of this kind might be expected to affect orthographic development by
inhibiting the establishment of the core structure, especially if there was an accompanying phonological difficulty that affected
the capacity to derive a phonemic and a three-dimensional structure for the syllable. If the core of the orthographic lexicon did
not develop, the whole of the subsequent development of an organized, phonologically based framework would be
compromised. The outcome might be a pattern of

Page 333
phonological dyslexia in which a word vocabulary is developed (based on the logographic system), although nonwords cannot
be read easily (on account of the lack of an orthographic framework) or only by a laborious process (equated with a
malfunctioning alphabetic process) (Seymour, 1990; Seymour & Bunce, 1994).
According to this account, the contrasting patterns of phonological and surface dyslexia reflect the effect of a selective
impairment of one or other of the two foundation processes on orthographic development. The other possibility is that the basis
of the orthographic framework is itself impoverished. If this view is taken, the explanation of the contrasting patterns of dyslexia
requires some additional assumptions, notably that different kinds of impoverished frameworks are possible. In models such as
the one described by Seidenberg and McClelland (1989), the network linking orthographic elements to phonemic elements
performs a dual function of deriving generalizations about the correspondence system while storing data about individual words.
This dual function depends rather critically on the way the network is constructed, especially the number of hidden units that are
interposed between the two levels. It seems possible that one could envisage a range of networks, some of which are specialized
in favor of the generalization function and against word-specific storage, and others are specialized in favor of item storage and
against generalization. If so, there might be a variation across individuals regarding the direction and degree of specialization
that was favored by the available network.
The model accordingly favors the view that differing patterns of dyslexia are observable. These could be classified with respect
to level, such that one might identify a child as being blocked at the foundation or at an orthographic or morphographic level.
They can also be classified in terms of bias, with a distinction between logographic and alphabetic impairments at the
foundation level, and a surface versus phonological contrast at the orthographic level. Morphographic impairments might also be
found to take differing forms, for example, a distinction between affixes and stems as the primary source of difficulty.
Linguistic Awareness
Linguistic awareness is held to be interactively involved in the development of the orthographic and morphographic systems at
all levels. As orthographic development progresses, there is always a corresponding advance in the relevant area of linguistic
awareness. At any stage, a failure of reciprocity, due to an inability to develop the relevant linguistic structures, can pose a
barrier against further development. The effects will vary, depending on the level that is affected. A difficulty affecting
morphemic awareness affects the establishment of the morphographic framework (thus damaging the

Page 334
ability to learn to spell complex words), but need not affect the processes occurring at the foundation and orthographic levels.
Similarly, a lack of awareness of 2D phonological structure (the onset-rime distinction) will affect the final stages of
orthographic development, including classification of consistent and inconsistent word families. The most serious difficulties are
those that affect 3D syllabic structure and phonemic segmentation because these have the potential to disrupt the alphabetic
foundation, the formation of the orthographic core, and the whole of the subsequent program of orthographic and
morphographic development.
Implications for Teaching
The model has some implications for the teaching of beginning reading or spelling and the remediation of difficulties. A first
point is that the componential character of the theory of acquisition helps in the identification of the foci of contrasting
instructional methods. Emphasis on whole words (in or out of context) favors the logographic process. Letter sounding develops
the alphabetic process and the emergence of phonemic awareness. A stress on word families promotes 2D organization of the
orthographic framework and onset rime awareness. Discussion of the derivational structure of complex words is directed toward
the morphographic framework and the development of morphemic awareness. The model, therefore, provides a basis for
discussion of contrasting instructionally driven pathways in literacy acquisition.
In relation to reading difficulties, the important distinctions relate to the foundation and the orthographic and morphographic
levels. If development is blocked at the foundation level (corresponding to a reading age of less than 7 years in a UK context),
then the appropriate intervention is an attempt to establish the elements of the foundation, with emphasis being directed toward
whichever subprocess (logographic or alphabetic) seems more seriously impaired. At more advanced levels, the aim must be the
step-by-step construction of an orthographic or a morphographic framework. Hence, a remedial teacher who wishes to work
under the guidance of a theoretical model of literacy acquisition requires diagnostic instruments and teaching materials that can
be mapped directly onto the components and the levels that have been identified.
References
Alegria, J., Pignot, E., & Morais, J. (1982). Phonetic analysis of speech and memory codes in beginning readers. Memory and
Cognition, 10, 451-556.
Barry, C., & Seymour, P. H. K. (1988). Lexical priming and sound-to-spelling contigency effects in nonword spelling.
Quarterly Journal of Experimental Psychology, 40A, 5-40.

Page 335
Bertelson, P. (1987). The onset of literacy: Liminal remarks. In P. Bertelson (Ed.), The onset of literacy: Cognitive processes in
reading acquisition (pp. 1-30). Cambridge, MA: MIT Press.
Boder, E. (1973). Developmental dyslexia: A diagnostic approach based on three atypical reading-spelling patterns.
Developmental Medicine and Child Neurology, 15, 663-687.
Bradley, L., & Bryant, P. E. (1983). Categorising sounds and learning to read: A causal connection. Nature, 301, 419-421.
Brown, G. D. A., & Loosemore, R. P. W. (1994). Computational approaches to normal and impaired spelling. In G. D. A.
Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process, and intervention (pp. 319-335). Chichester: Wiley.
Bryant, P. E., MacLean, M., Bradley, L.L., & Crossland, J. (1990). Rhyme and alliteration, phoneme detection, and learning to
read. Developmental Psychology, 26, 429-438.
Castles, A., & Coltheart, M. (1993). Varieties of developmental dyslexia. Cognition, 47, 149-180.
Duncan, L. G., Seymour, P. H. K., & Hill, S. (in press). How important are rhyme and analogy in beginning reading? Cognition.
Ehri, L. C. (1992). Reconceptualising the development of sight word reading and its relationship to recoding. In P. B. Gough, L.
C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 107-143). Hillsdale, NJ: Lawrence Erlbaum Associates.
Ellis, A. W. (1985). The cognitive neuropsychology of developmental (and acquired) dyslexia. Cognitive Neuropsychology, 2,
169-205.
Ellis, A. W. (1993). Reading, writing, and dyslexia (2nd ed.). Hove, UK: Lawrence Erlbaum Associates.
Ellis, A. W., & Young, A.W. (1988). Human cognitive neuropsychology. Hove, UK: Lawrence Erlbaum Associates.
Surface dyslexia: Neuropsychological and cognitive analyses of phonological reading (pp. 301-330). London: Lawrence
Erlbaum Associates.
Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. Hove, UK: Lawrence Erlbaum Associates.
Gough, P. B., & Hillinger, M. L. (1980). Learning to read: An unnatural act. Bulletin of the Orton Society, 30, 179-196.
Hatfield, F. M., & Patterson, K. E. (1983). Phonological spelling. Quarterly Journal of Experimental Psychology, 35, 451-468.
Henderson, L. (1982). Orthography and word recognition in reading. London: Academic Press.
Karmiloff-Smith, A. (1986). From metaprocesses to conscious access: Evidence from children's metalinguistic repair data.
Kavanagh, J. F., & Mattingly, I. G. (1972). Language by ear and by eye. Cambridge, MA: MIT Press.
Liberman, I. Y., Shankweiler, D., Liberman, A. M., Fowler, C., & Fischer, F. W. (1977). Phonetic segmentation and recoding in
the beginning reader. In A. S. Reber & D. L. Scarborough (Eds.), Toward a psychology of reading: The proceedings of the
CUNY conferences (pp. 207-225). Hillsdale, NJ: Lawrence Erlbaum Associates.
Lundberg, I., Frost, J., & Petersen, O.-P. (1988). Effects of an extensive programme for stimulating phonological awareness in
Manis, F. R., Szeszulski, P. A., Holt, L. K., & Graves, K. (1990). Variation in component word recognition and spelling skills
among dyslexic children and normal readers. In T. H. Carr & B. A. Levy (Eds.), Reading and its development: Component skills
approaches (pp. 207-259). San Diego, CA: Academic Press.
Mitterer, J. O. (1982). There are at least two kinds of poor readers: Whole word poor readers and recoding poor readers.
Canadian Journal of Psychology, 36, 445-461.
Morais, J. (1991). Constraints on the development of phonemic awareness. In S. A. Brady & D. P. Shankweiler (Eds.),
Phonological processes in literacy: A tribute to Isabelle Y. Liberman (pp. 5-27). Hillsdale, NJ: Lawrence Erlbaum Associates.

Page 336
Morais, J., Alegria, J., & Content, A. (1987). The relationships between segmental analysis and alphabetic literacy: An
interactive view. Cahiers de Psychologie Cognitive, 7, 415-438.
Morais, J., Cary, L., Alegria, J., & Bertelson, P. (1979). Does awareness of speech as a sequence of phones arise spontaneously?
Morton, J. (1980). The logogen model and orthographic structure. In U. Frith (Ed.), Cognitive processes in spelling. London:
Academic Press.
Morton, J. (1989). An information-processing account of reading acquisition. In A. Galaburda (Ed.), From reading to neurons.
Cambridge, MA: MIT Press.
Pennington, B. F. (1991). Diagnosing learning disorders: A neuropsychological framework. New York: Guilford.
Perfetti, C. A., Beck, I. L., Bell, L., & Hughes, C. (1987). Phonemic knowledge and learning to read are reciprocal: A
longitudinal study of first-grade children. Merrill-Palmer Quarterly, 33, 283-319.
Perfetti, C. A. (1992). The representation problem in reading acquisition. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.),
Rack, J., Snowling, M., & Olson, R. (1992). The nonword reading deficit in dyslexia: A review. Reading Research Quarterly,
27, 28-53.
Seymour, P. H. K. (1990). Developmental dyslexia. In M. W. Eysenck (Ed.), Cognitive psychology: An international review
(pp. 135-196). Chichester: Wiley.
Seymour, P. H. K. (1993). Un modle de dveloppement orthographique double fondation. In J.-P. Jaffr, L. Sprenger-
Charolles, & M. Fayol (Eds.), Lecture-criture: Acquisition. Les Actes de la Villette. Paris: Nathan Pedagogie.
Seymour, P. H. K., & Bunce, F. (1994). Application of cognitive models to remediation in cases of developmental dyslexia. In
M. J. Riddoch & G. W. Humphreys (Eds.), Cognitive neuropsychology and cognitive rehabilitation (pp. 349-377). Hove, UK:
Seymour, P. H. K., & Elder, L. (1986). Beginning reading without phonology. Cognitive Neuropsychology, 3, 1-36.
Seymour, P. H. K., & Evans, H. M. (1992). Beginning reading without semantics: A cognitive study of hyperlexia. Cognitive
Seymour, P. H. K., & Evans, H. M. (1993). The visual (orthographic) processor and developmental dyslexia. In D. Willows, R.
Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 347-376). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Seymour, P. H. K., & Evans, H. M. (1994a). Sources of constraint and individual variations in normal and impaired spelling. In
G. D. A. Brown & N. C. Ellis (Eds.), Handbook of spelling: Theory, process, and intervention (pp. 129-153). Chichester: Wiley.
Seymour, P. H. K., & Evans, H. M. (1994b). Levels of phonological awareness and learning to read. Reading and Writing: An
Interdisciplinary Journal, 6, 221-250.
Seymour, P. H. K., & Porpodas, C. D. (1980). Lexical and nonlexical processing of spelling in dyslexia. In U. Frith (Ed.),
Shallice, T. (1988). From neuropsychology to mental structure. Cambridge, UK: Cambridge University Press.
Stanovich, K. E. (1988). Explaining the difference between dyslexic and garden variety poor readers: The phonological-core
variable-difference model. Journal of Learning Disabilities, 21, 590-604.

Page 337
Treiman, R. (1987). On the relationship between phonological awareness and literacy. Cahiers de Psychologie Cognitive, 7,
524-529.
Treiman, R. (1992). The role of intrasyllabic units in learning to read and spell. In P. B. Gough, L. C. Ehri, & R. Treiman
(Eds.), Reading acquisition (pp. 65-106). Hillsdale, NJ: Lawrence Erlbaum Associates.


Page 339
Chapter 17
Beginning Reading and Spelling Acquisition in French: A Longitudinal Study
Liliane Sprenger-Charolles
CNRS & Universit Ren Descartes, Paris
Linda S. Siegel
University of British Columbia
Danielle Bchennec
National Institute for Pedagogical Research, Paris
Most research on reading and spelling acquisition has been conducted in the English language (see Perfetti, 1994). This state of
affairs poses a problem because developmental sequences seem to depend partially on the specific characteristics of languages
and orthographies rather than exclusively on the general principles common to all languages (see Frost, 1994). For example, the
comparisons between the acquisition of reading in English and in German, a language more transparent than English at the level
of grapheme phoneme correspondences, show that German-speaking children use phonological mediation even at the very
beginning of reading acquisition, without passing through a logographic stage of nonanalytic, direct visual processing (Wimmer
& Goswami, 1994); it is not necessarily the case for English-speaking children (Wimmer & Goswami). Therefore, it is
important to examine reading and spelling acquisition in French, a language that, similar to German, has a written system less
opaque than English.
Some Characteristics of the French Language

French is different from English at several levels. Specifically, the patterns of stress for words and phrases and the
distinctiveness of words in the

Page 340
speech stream vary between these languages. An English word has a certain phonetic independence in a sentence because every
word has its own stress. In French, stress occurs within a group of words instead of within a word; therefore, word boundaries
are not as easy to distinguish in French (see Encrev, 1988). Another difference between the two languages is in their syllabic
structure (Delattre, 1965; Kaye & Lowenstamm, 1984). English typically has closed syllables, and French primarily uses open
syllables (see Delattre). These two features of French make the outlines of a word sensitive to contextual influences. The final
consonant of a word is generally pronounced with the initial vowel of the following word, for example, pour une autre amie is
said /pu/ry/no/tra/mi/ and not /pur/yn/otr/ami/. Therefore, the graphic space used as a boundary for written words may not be
congruent with an audible reality.
For the spelling system, the basic principles in French are the same as in English. The useful units in writing are graphemes,
which can have both phonemic (phonograms) and morphemic references (morphograms; e.g., -s to mark plural; see Catach,
1980; Gak, 1976). In French, phonograms represent 83% of graphemes, according to Catach, who defined four levels of
phonograms. The basic units (level zero) are theoretical units representing a class of phonograms. For example, the
archiphonogram O covers all realizations of /o/. One of the main features of French is that, at this basic level, there is a large
proportion of complex units. Specifically, one third of the 33 archiphonograms are complex units that have no simple
equivalents (e.g., the ou /u/, on // vowels and the ch / / consonant). The three other levels correspond to graphemic
realizations characterized according to their frequency.
If one assumes that the basic spelling unit is the grapheme and takes into account the contextual rules that govern the
pronunciation of some of these units (c + i is read /si/ and c + a is read /ka/), then the grapheme phoneme correspondences are
highly regular and predictable in French; for example, o, au, and eau are always pronounced in the same way (/o/). On the other
hand, the phoneme grapheme correspondences are less predictable. In fact, in order to spell regular words such as domino
(/domino/), landau (/ldo/), or bateau (/bato/) without error, it is necessary to choose the correct grapheme for the phoneme /o/
among three usual allographes (o, au, and eau).
The main objective of the present study is to evaluate the effects of those features of oral and written French on the procedures
used by French-speaking beginner readers and writers.
General Framework of the Study
According to dual-route models (see Coltheart, Curtis, Atkins, & Haller, 1993; Ellis, 1984; Humphreys & Evett, 1985; Paap &
Noel, 1991) and their developmental adaptation by Frith (1985, 1986), Morton (1989), Harris and Coltheart

Page 341
(1986), and Seymour (1986), two procedures govern access to written words. The first involves making a direct match between
the written word and the orthographic lexicon, specified as whole words and lexical morphemes. This direct lexical process is
called the orthographic procedure in both Frith's (1985, 1986) and Morton's developmental models and does not correspond, in
theory at least, to prereading and prespelling logographic strategies, although these two types of processing may be influenced
by common factors. Specifically, in the logographic stage, children can directly ''read'' very frequent words that they know by
"sight." Unlike the orthographic procedure, the second process is an indirect and nonlexical procedure. In reading, the graphic
units (letters, graphemes, or syllabic units) are associated with phonic units (sounds, phonemes, or onset-rimes) that are then
assembled. In spelling, phonic units are connected to graphic units. This phonological mediation procedure, called the alphabetic
procedure in Frith's and Morton's developmental models, is slow and laborious at the beginning of reading and spelling
acquisition.
The study we summarize here was designed to examine the development of these procedures in reading and spelling in French.
It is possible that the regularity of the grapheme phoneme correspondences in French will lead to a greater reliance on
phonological mediation from the first stage of reading and spelling acquisition. In addition, the fact that the word unit is not
very accessible in the speech stream may reduce the use of the direct lexical route. Therefore, our main hypothesis is that the
first stage of reading and spelling acquisition in French is characterized by the exclusive use of phonological mediation and an
absence of influence from such lexical factors as frequency, lexicality, or analogy. This expectation for French is similar to the
predictions of Frith's (1985, 1986), Morton's (1989), and Harris and Coltheart's (1986) models, which proposed a stage in which
the children use only phonological mediation. On the other hand, this prediction is different from those predicted by the double
foundation model of Seymour (1990, 1994) that postulated a co-occurrence of the logographic and alphabetic procedures in the
first stage of acquisition. It differs also from the prediction of Goswami and Bryant's analogy-based model (1990) because the
lexical effect of analogy is not expected at the beginning. However, in later stages of acquisition, the effects of frequency,
lexicality, and analogy should appear, without the disappearance of the regularity effect.
Moreover, if processing units used by children in early reading and spelling are letters corresponding to sounds and not
graphemes connected with phonemes we should observe an effect of graphemic complexity for regular items. For example,
items such as table /tabl/ or lople /lcpl */ (called simple items) should be more accurately read and spelled than items such as
route /rut/ or moube /mub/ (called complex items) in which the ou is a two-letter grapheme corresponding to one phoneme (/u/).
Although these

Page 342
trends should be seen in both reading and spelling, the use of phonological mediation should be earlier and more marked in
spelling than in reading, in agreement with Frith's (1985, 1986) and Morton's (1989) models. Thus, we predict higher
performance in spelling than in reading for all items that can be processed by phonological mediation. Because grapheme
phoneme correspondences are more predictable in French than phoneme grapheme correspondences, a fact favoring reading, the
advantage of spelling over reading is predicted only for pseudowords. Regular words have several possible spellings but only
one possible pronunciation (bateau is always said as /bato/, even though it is possible to spell this word bato, batto, batau,
battau, batteau) and may be read better than spelled. Correspondingly, the presumably later developing orthographic direct route
should appear earlier and should be more marked in reading; therefore, on variables related to orthographic procedures
(frequency, lexicality, analogy), we should see a superiority of reading over spelling (see Frith; Morton).
Finally, we can hypothesize that phonological mediation is a mechanism that allows the construction of the orthographic lexicon
(Gough & Walsh, 1991; Share, 1995). In such a case, high correlations should be observed between early phonological skills
and later orthographic skills. To test this hypothesis, pseudoword and regular word performance were used as measures of
phonological skills. As a measure of orthographic skills, irregular word performance was used because reading this type of item
cannot, by definition, be entirely dependent on decoding skills.
These questions were examined in a study in which we evaluated how reading and spelling procedures develop at the beginning
of their acquisition. The analyses compared effects due to phonological mediation, such as regularity and graphemic complexity,
with effects due to orthographic procedures, such as frequency, analogy, and lexicality. Evaluating reading and spelling together
is optimal for two reasons. The first one is theoretical, as the different developmental models, particularly Frith's (1985, 1986),
Morton's (1989) and Goswami and Bryant's (1990), stressed the reciprocal relations between these two skills. The second reason
is methodological. Interpreting children's responses in tests that involve reading aloud may be biased by difficulties in listening
and transcribing. Recording the data may lessen these risks because it is possible to verify their answers. However, even careful
listening cannot completely avoid certain biases resulting, for example, from lexical reconstruction because we are inclined to
hear the l of table /table/ even if the child does not pronounce it (see Mowrey & MacKay, 1990; Treiman, Berch, Tincoff, &
Weatherston, 1993). This type of risk may be partially avoided when production and reception (i.e., reading and spelling) of the
same corpus are studied at the same time. If identical patterns for correct responses and errors are noted in reading and in
spelling, the validity of the observations is strengthened.

Page 343
This study is longitudinal because, in our opinion, longitudinal studies more accurately test developmental hypotheses. If the
same participants are examined at different times, the differences observed in performance can be ascribed to developmental
changes rather than to sample differences. However, the time period is necessarily compressed because of the rapid development
of children's language skills. In tasks that require spelling and reading words and pseudowords, children's performance,
especially in spelling, is very poor when they are just acquiring these skills, but, very quickly, performance in reading and, to a
lesser extent, in spelling reaches ceiling levels. Therefore, this study observed the development of reading and spelling skills in
first-grade children at the middle and the end of the year when their scores were not biased by either floor or ceiling effects.
Participants
Children had to meet certain criteria to be participants in the study. First, they had to be native French speakers. Second, children
with language disabilities, sensorimotor, or neural deficits detected by teachers, psychologists, or medical doctors, were
excluded. Finally, children with behavioral problems or those who lacked an adequate opportunity for schooling were not
considered.
Using these criteria, we examined a group of 100 children at the end of kindergarten. From this group, we chose 60 nonreaders
of middle-to-high cognitive level (50th percentile or higher on the Raven Progressive Matrices [Raven, 1947]) from different
socioeconomic levels representative of the French population. Their nonreading level was evaluated by the BAT-ELEM. This
test was used to confirm that they were nonreaders (Savigny, 1974).
It was possible to follow 57 of these children until the end of the first grade. For this first series of observations in primary
school (in January), the average age of these children was 6 1/2 years old (SD, 3 mos.). They were enrolled in 20 different
classes in nine different schools located in Parisian suburbs and received neither systematic teaching of reading nor teaching of
letter-to-sound correspondences in kindergarten. In first grade, the 20 different teachers used mainly mixed methods for reading
instruction, with variations in when grapheme-phoneme correspondence rules were explicitly taught.
The Tasks
The children first read and then spelled the same set of words and pseudowords, with the spelling task always second. Real-
word reading and spelling were separated by 1 week, with the pseudoword reading and spelling tasks given after an additional
week.

Page 344
For the reading tasks, the procedures used for words and pseudowords were the same. (For pseudowords, children were told
they would read [or spell] words that did not exist.) The children took the test individually on a computer that recorded their
responses. For the spelling tasks, the words to be spelled were pronounced in context to avoid confusion between homophones
(e.g., compte and conte /kt/). For reading and spelling of both words and pseudowords, no feedback was given to the children.
The tasks were given in January and then in June of first grade.
The Stimuli
Thirty-six words and 16 pseudowords were used (see the Appendix). The selection of words was based on their regularity,
graphemic complexity, and frequency. They fell into three categories: regular simple, regular complex, and irregular. The
criteria for regularity were based on the French language analyses of Catach (1980) and Gak (1976). The regular words, both
simple and complex, contained only frequent phonograms. For the 12 regular simple words, each letter matched a phoneme
except for a silent e at the end of a word. The regular complex words contained one of two complex phonograms. These two
phonograms, ou /u/ and ch / /, were chosen because they have no simpler equivalents. Other complex phonograms, such as au,
which has a simple and more frequent allograph (o), were not used because children could meet difficulties with those
phonograms in the spelling task.
The irregular words had either a low frequency phonogram with a specific pronunciation (e.g., the first e of femme pronounced
/a/) or a silent letter in a nonterminal position (e.g., the p in sept). Irregularities that were limited to the final consonant of a
word were avoided because one can correctly read a word such as porc (/pcr */) with a strictly sequential phonological decoding
that stops before the last letter.
Word frequency was established on the basis of Listes Orthographiques de Base (Basic Orthographic Lists, LOB, Catach, 1984)
that draws on several frequency tables (Gougenheim, Micha, Rivenc, & Sauvageot, 1964; Juilland, Brodin, & Davidovitch,
1970; Trsor de la langue franaise [French Language Treasure], 1971). Each of the three word categories contained half high-
frequency words and half low-frequency words. High-frequency words are in the first 2,000 items of the LOB. The low-
frequency words are not in the LOB but are typically in the spoken vocabulary of first graders, as verified by an oral
examination of children of the same age and school level as the ones observed in this study. Finally, the items were matched
according to their initial letter and total number of letters.
Sixteen pseudowords were constructed and matched in orthographic difficulty with regular words. Half these pseudowords were
analogs to the high-frequency realwords, four regular simple, and four regular complex

Page 345
pseudowords. The initial letter of the high-frequency words was altered to create these analog pseudowords (e.g., mable for
table or soche for poche). For nonanalog pseudowords (four regular simple items and four regular complex items), other letters
were changed so that there were no analog high-frequency words beginning or ending with the same letters.
Results and Discussion
Phonological Mediation and Orthographic Procedures in Reading and Spelling

Based on certain characteristics of French, we predicted that phonological mediation influences the beginning of spelling and
reading without any influence of such lexical factors as frequency, lexicality, or analogy. In the following stages of acquisition,
the effects of frequency, lexicality, and analogy are expected to appear and the effects of regularity are expected to remain. The
main predictions are summarized in Table 17.1.
In accordance with our predictions, the results of the first session (shown in Table 17.2) show that irregular words were not as
well read and spelled as regular simple or complex words. Moreover, frequency had no significant effect on performance, and
regular words were neither read nor spelled more accurately than their matched pseudowords. This pattern of results suggests
that the children primarily used phonological mediation at the beginning of reading and spelling. Lexicality did not matter for
reading and actually had a detrimental effect on spelling.
Contrary to one of our hypotheses, there was an effect of analogy in reading. Because of the results of regularity and frequency
reported previously, which argue against a lexical process, the analogy effect may not
TABLE 17.1
Summary of the Predictions for Regularity,
Frequency, Lexicality, and Analogy Effects
Middle of First End of First Grade
Grade
Reading Spelling Reading Spelling
RW > RW > IW RW > RW >
Regularity: Regular words (RW) IW IW IW
versus irregular words (IW)
HF = LF HF = LF HF > LF HF > LF
Frequency: High frequency words
(HF)
versus low frequency (LF)
WD = WD = WD > WD >
Lexicality: Words (WD) PW PW PW PW
versus pseudowords (PW)
AP = NP AP = NP AP > NP AP > NP
Analogy: Analog (AP)
versus non-analog (NP)
pseudowords

Page 346
TABLE 17.2
Regularity, Frequency, Lexicality, and
Analogy Effects in Reading and in Spelling
Middle of First Grade End of First Grade
Regularity: regular words (RW) RW > IW RW > IW RW > IW RW > IW
versus irregular words (IW) ** ** ** **
Frequency: high frequency words HF = LF HF = LF HF > LF HF > LF
(HF) versus low frequency (LF) ** **
Lexicality: words (WD) versus WD = PW WD < PW WD > PW WD = PW
pseudowords (PW) ** **
Analogy: analog (AP) versus AP > NP AP = NP AP > NP AP > NP
non-analog (NP) pseudowords * (p < .06) **
Note. Significant differences are indicated by ** for p < .01 or * for p < .05.
reflect facilitation via the orthographic lexicon. Instead, analogy in reading may have worked through the oral response, with
analog pseudowords being constructed on frequent words that have easily recoverable articulatory codes. If so, we should not
have observed any effect of analogy in spelling because the programming of the articulatory codes should not affect the written
response. In fact, this is what we observed in the first session in spelling. Moreover, the correlations between analog and
nonanalog pseudowords were very high (.90 in reading and .86 in spelling) and the effect of graphemic complexity a nonlexical
factor affected the analog and nonanalog pseudowords in the same way (the interaction between graphemic complexity and
analogy was not significant). These observations tend to support the hypothesis that pseudowords, whether they were analog or
nonanalog, were processed by an identical nonlexical procedure (i.e., by phonological mediation).
Table 17.2, which summarizes the effects observed in reading and in spelling, indicates that, except for the analogy effect in
reading, all our predictions for the first session are confirmed. French-speaking children primarily showed the use of
phonological mediation in the beginning of reading and spelling acquisition.
Nevertheless, we observed changes in performance from the middle of the year to the end of the year that indicate the
progressive establishment of the orthographic lexicon (see Table 17.2). In the second session, the high-frequency words were
better read and spelled than the low-frequency ones. Also, words were better read than pseudowords. In spelling, the reverse
lexicality effect favoring the spelling of pseudowords over words disappeared in the second session.
In Frith's (1985, 1986), Morton's (1989) and Seymour's (1986) models, frequency and lexicality effects were interpreted as
indicators of logo-

Page 347
graphic or orthographic procedures. Because we found no effect of frequency and no advantage for words in the first session,
the assumption of a logographic lexicon composed entirely of sight words can be rejected. On the other hand, the second session
results indicate a progressive development of the orthographic lexicon. The fact that, in this session, words were not more
accurately spelled than pseudowords, although they were better read, indicated the negative impact of orthography on spelling.
Finally, in accord with our hypothesis, more accurate performance (both in reading and in spelling) was observed for the analog
pseudowords in the second session, although this effect did not reach conventional levels of statistical significance in reading (p
< .06).
These results showed that the children were able to use an orthographic procedure at the end of first grade. However, the
orthographic procedure did not entirely replace phonological mediation. We still observed a regularity effect in the second
session. This effect was even stronger in session 2 than in session 1 because performance for regular words improved more than
for irregular words.
The increased importance of phonological mediation between sessions was also evident in the error analysis that found the mean
percentage of regularization errors (e.g., album /albcm */, read /albym/ or spelled albome) to increase between sessions from
4.9% to 35.2% in reading and from 27.4% to 64.4% in spelling. These results indicate that, in the second session, at the end of
first grade, reading and spelling are controlled by both orthographic and phonological procedures. They do not support the idea
of the existence of an orthographic phase in which phonological mediation exerts no influence (see Frith, 1985, 1986; Morton,
1989).
Relationships Between Reading and Spelling

In the developmental models of Frith (1985, 1986) and Morton (1989), phonological mediation was assumed to be used earlier
and more strongly in spelling than in reading, whereas an orthographic procedure should occur earlier in reading than in
spelling. Therefore, we might suppose that all the indicators of phonological mediation (i.e., correct responses for regular words
and for pseudowords and regularization errors) should be higher in spelling than in reading at the beginning. However, to
correctly spell a word, even a regular one, it is necessary to know its orthography; therefore, the superiority of spelling as
compared to reading can be expected only for pseudowords, which have no conventional spelling. On the other hand, indicators
of the orthographic procedure, such as frequency, lexicality, or analogy, should appear more strongly in reading than in spelling.
Consistent with our hypothesis, the frequency effect was more significant in reading than in spelling and the lexicality effect
was more pronounced in

Page 348
spelling, where the pseudowords were spelled more accurately than the words. Two other results support the hypothesis of a
greater recourse to orthographic procedures in reading. First, for all words (regular and irregular), the reading spelling difference
increased between sessions. In the first session, 29.4% of the reading responses and 26.1% of the spelling responses were
correct. In the second session, there were 70.8% correct responses in reading and only 53.2% in spelling. For pseudowords, there
was no corresponding reading advantage: 48.2% and 48.4% correct responses for reading and spelling, respectively, in the first
session and 80.0% and 76.3% in the second. The second supportive result was the increasing impact of orthography on spelling
compared with reading. The global deficit of spelling relative to reading was 5.6% for simple regular words, 12.3% for complex
regular words and 13.4% for irregular words. The difference for irregular words between reading and spelling increased from
2.2% in the first session to 24.7% in the second. These results suggest that the orthographic procedure was more marked in
reading than in spelling and that its significance to the benefit of reading increased with time.
Contrary to our hypothesis of an earlier and stronger recourse to phonological mediation in spelling as compared to reading in
the beginning of acquisition, the results of the correct-response analysis did not show better spelling of pseudowords. This result
might be due to the fact that graphemic complexity had a positive impact on pseudoword reading but not on pseudoword
spelling. This interpretation was reinforced by the results obtained in another study in which we observed better results in the
first session in the spelling of pseudowords that contained only simple graphemes (Sprenger-Charolles & Siegel, in press). It is
possible that the results of other studies finding a superiority of reading over spelling were influenced by graphemic complexity.
(For English-speaking beginners, see Seymour, 1990; Seymour & Evans, 1991; for German-speaking beginners, see Wimmer &
Hummer, 1990.) In the present study, an earlier and more marked use of phonological mediation in spelling might be inferred
from the regularization error analysis that showed more regularization errors in spelling than in reading in both sessions (35.2%
in spelling versus 4.9% in reading in the first session and 64.4% and 27.4% for the second session).
Table 17.3, which summarizes the entire results, shows that phonological mediation is used more strongly and earlier in spelling
than in reading, whereas the reverse is observed for the orthographic procedure. However, beyond these differences, the
correlations between reading and spelling were all positive and significant: first session, r = .70 for pseudowords, .82 for regular
words and .65 for irregular words; second session, r = .85, .80, and .72. These correlations show that, beyond their differences,
there are also strong links between reading and spelling, at least at the beginning of their acquisition in French.

Page 349
TABLE 17.3
Comparisons Between Reading and Spelling
Words
Inferiority in word reading as compared to word spelling
(**),a which
increased between sessions (**)
Pseudowords
No difference between pseudoword reading and
pseudoword
spelling; no change between sessions
Orthography Detrimental effect more marked in spelling than in
effect: reading (**), which
SW/CW/IWb increased between sessions (**)
Regularization Reading < Spelling **

errors: Reading < Spelling **
Middle of first
grade
End of first grade
Frequency effect
Facilitatory effect more marked in reading than in
spelling (p < .06);
no change between sessions
Lexicality effect
Detrimental effect more marked in spelling than in
reading (p < .06);
no change between sessions
Note. aSignificant differences are indicated by ** for p < .01 or * for p <
.05
bSW = regular word with ssimple grapheme; CW = regular word with
complex grapheme; IW
= irregular word.
The Processing of Complex Graphemes

If processing units used by children at the beginning of reading and spelling acquisition are letters in relation to sounds, and not
graphemes in relation to phonemes, we should observe an effect of graphemic complexity for regular items, which then
progressively disappears in both reading and spelling. In fact, we observed a difference between reading and spelling with regard
to the effect of graphemic complexity. Graphemic complexity had a detrimental effect only on performance for word spelling in
the first session. In June (the second session), in the pseudoword reading task, we even observed a superiority of the complex
items as compared to the simple ones. The results are summarized in Table 17.4. They show that, except for the word spelling
task in the first session, the children seemed to process graphemes rather than letters.
These results might have been due to the processing of complex graphemes per se or the result of compound factors. To check
this, we compared the processing of simple and complex graphemes in complex items. In these comparisons, we took into
account the mean percentage of orthographically correct responses for complex graphemes and for simple graphemes with the
exception of the final silent e in word reading and spelling. The same comparison was made for pseudowords. However,
because pseudowords have no canonical spelling, we took into account the mean percentage of phonologically correct responses.

Page 350
TABLE 17.4
Graphemic Complexity Effect in Reading and in Spelling
Middle of First Grade End of First Grade
SW = CW **SW > CW SW = CW SW = CW
Words with simple (SW)
versus complex (CW) grapheme
SP = CP SP = CP *SP < CP SP = P
Pseudoword with simple (SP)
versus complex (CP) grapheme
Note. Significant differences are indicated by ** for p < .01 or * for p < .05.
This direct comparison between simple and complex graphemes showed a very clear developmental trend for spelling. Complex
graphemes were less well spelled than simple graphemes in the January session both for words and pseudowords; in the June
session, there was no difference between simple and complex graphemes. These results show that children progressively set up
complex graphemes and that they were rapidly using graphemes, not letters, in spelling.
In reading, simple graphemes were better processed than complex graphemes only in the pseudoword task during the January
session, suggesting that children applied mainly graphemes in reading. Moreover, complex graphemes were more accurately
processed than simple ones in one case (word reading in June). This last result was not observed in the spelling tasks.
These results differ slightly for overall accuracy from those discussed previously. Such differences might result from the fact
that, in this complex grapheme analysis, only correct graphemes were assessed. For example, in table, read or spelled tabal(e),
all the expected graphemes were correct even though the response for the complete word was incorrect.
The Role of Phonological Mediation in the Construction of the Orthographic Lexicon

If phonological mediation is a mechanism allowing the setting up of the orthographic lexicon, we should observe a connection
between items that respond to the early phonological procedures and those that respond to later lexical-orthographic procedures.
For example, we should observe a connection between early success on regular words and pseudowords and later success in
reading and spelling of irregular words.
In fact, we observed significant positive correlations between performance on regular words and pseudowords in January and
performance in June for all types of items, including irregular words. However, scores for

Page 351
TABLE 17.5
Correlations Between Middle of First Grade and End of First Grade Scores
Reading
Pseudowords Regular Words Irregular Words
Pseudowords 0.61** 0.54** 0.27
Regular Words 0.55** 0.49** 0.25
Irregular Words 0.63** 0.69** 0.58**
Spelling
Pseudowords Regular Words Irregular Words
Pseudowords 0.61** 0.48** 0.27
Regular Words 0.48** 0.51** 0.15
Irregular Words 0.55** 0.48** 0.42**
Note. Significant differences are indicated by ** form p < .01.
irregular words in January were not correlated with those for regular words and pseudowords in June (see Table 17.5). We also
observed positive correlations between the regularization errors, another indicator of phonological procedures, and correct
responses in January for both reading (r = +.54) and spelling (r = +.42).
These results were consistent with those of some other studies showing that reading and/or spelling performance on
pseudowords or regular words, including regularization errors, predict all further results in reading and in spelling, even for
irregular words (see Byrne, Freebody, & Gates, 1992; Foorman, Francis, Novy, & Liberman, 1991; Foorman, Jenkins, &
Francis, 1993; Gough & Walsh, 1991; for a review, see Share, 1995). The reverse has not been established. The capacity to read
or spell irregular words does not seem to be a predictor of future skills in reading and/or spelling.
General Discussion
To evaluate whether French first-grade children use the indirect phonological route or the direct orthographic route, we asked
them to read and spell items we had selected or constructed so as to show the procedure required to read and spell them. For
example, sensitivity to frequency or lexicality is generally considered an indicator of direct matching between a word and the
orthographic lexicon because a high-frequency word, unlike a pseudoword or a low-frequency word, might have an entry in the
orthographic lexicon of the beginners. An analogy effect for pseudowords also suggests this lexical procedure because analog,
not nonanalog, pseudowords share their orthographic body with the high-frequency words from which

Page 352
they were constructed (e.g., mable from table). On the other hand, differences between regular and irregular words or
differences between regular words, according to whether they contain simple or complex graphemes, are indicators of
phonological mediation.
The principal hypothesis of this study on reading and spelling acquisition was that we should find signs of virtually exclusive
use of the nonlexical, indirect phonological route in the first stage of acquisition. This is because, in French, the
correspondences between grapheme and phoneme are regular and because the word unit is not very easy to determine in the
speech stream.
The data support our prediction. In the first session, performance was affected by regularity but not by frequency or lexicality.
Thus, regular words were better read and better spelled than irregular ones, and regularization errors (e.g., album /albom/ read
/albym/ or spelled albome) were made with the irregular words. Furthermore, high-frequency words were not more accurately
read or spelled than low-frequency ones, and comparisons between words and pseudowords showed no word superiority effect.
The pattern of results was completely consistent with the hypothesis of an exclusive phonological stage, with the single
exception that there was an effect of analogy in early reading; analog pseudowords were better read, but not better spelled, than
nonanalog pseudowords. This effect, observed in reading at a time when neither frequency nor lexicality had any impact on
performance, may be the result, as we have suggested, of a facilitation of the oral response rather than a true reading effect. The
analog items were derived from high-frequency words that may have preprogrammed articulatory codes. This interpretation was
confirmed by three aspects of the data. First, there was no effect of analogy in spelling, in which the programming of the
articulatory codes cannot have a direct effect. Second, there were very high correlations between analog and nonanalog
pseudowords in reading and in spelling. Third, graphemic complexity, which is a nonlexical factor, affected analog and
nonanalog pseudowords in the same manner. These results indicate that analog and nonanalog pseudowords were processed in
the same manner, by phonological mediation.
Taken together, the results show that French-speaking children mainly use phonological mediation at the beginning of reading
and spelling acquisition. These results then support the hypothesis of a phonological process as the sole operative process in this
early stage of reading and spelling acquisition as predicted in Frith's (1985, 1986), Morton's (1989), and Harris and Coltheart's
(1986) models. Alternatively, they are inconsistent with Seymour's (1990) hypothesis concerning the coexistence of logographic
and phonological procedures at the beginning of acquisition.
Although we found an early stage in reading and spelling acquisition during which the children mainly use phonological
mediation in French, we also concluded that an orthographic lexicon is quickly constructed. At the end of

Page 353
the first grade, words were better read than pseudowords, and frequency had an impact on the reading and spelling. Because
these effects were not seen earlier, they cannot be interpreted as the result of a logographic strategy. Rather, they indicate that an
orthographic lexicon gets progressively constructed, especially in reading. The fact that we did not observe any lexicality effect
in the second session in spelling is evidence of the negative impact of orthography on the spelling of even regular words.
Even if children in the later stage of reading and spelling acquisition are able to use a procedure of direct matching between a
word and its lexical entry, it does not entirely replace phonological mediation. This can be seen by the regularity effect observed
at the end of the first grade, which was even stronger that the one we had observed in January. The difference between regular
and irregular words increased, as did the mean number of regularization errors.
These data do not corroborate the existence of an orthographic stage in which phonological mediation plays no role (see Frith,
1985, 1986; Morton, 1989). Also contradicting the Frith and Morton models is the fact that there was no indication of a
logographic stage, one in which words known ''by sight'' would be read "holistically," whether regular or not. One may argue
that our first session was already too late in the development of the reading process to show logographic strategies. However, 6
months earlier, the children chosen for this study were all nonreaders, according to their results on a standard reading test. The
lack of a logographic stage in French (see Rieben, 1993; Sprenger-Charolles & Bonnet, 1996) may be explained by some of the
features of the orthography (the regularities of the grapheme-phoneme correspondences) and the language (the difficulty of
recognizing the word units in the speech stream). However, logographic reading might be rare in all languages (for German, see
Wimmer & Goswami, 1994; Wimmer & Hummer, 1990; for English, see Siegel, 1985).
Another hypothesis of the present longitudinal study is that phonological mediation allows the construction of the orthographic
lexicon (see Share, 1995). This hypothesis is corroborated by the discovery that early phonological decoding skills were
predictive of later reading and spelling performance, including performance with irregular words, but not the reverse. Moreover,
correlations between correct responses and regularization errors were, in the beginning of reading and spelling acquisition,
positive and significant. A similar result has already been noted in an earlier longitudinal study (Sprenger-Charolles & Casalis,
1995).
Similar tendencies were observed in another longitudinal study. In that study, we defined two groups of participants: a group of
"expert" spellers and a group of "poor" spellers (see Sprenger-Charolles, Siegel, & Bchennec, in press), established according
to their scores in an orthographic verification task. The experts in spelling are the participants who, at the end of

Page 354
second grade, made no errors in this task. Supposedly, they have a better orthographic lexicon than the poor spelling group. The
developmental trajectory of these two groups provides much support for the contribution of the phonological mediation
procedure in the establishment of the orthographic lexicon. First, the future experts in spelling were children who obtained
higher scores than the other participants in two metaphonological tasks at the beginning of their last year of kindergarten; at the
same time, those participants differed neither by their nonverbal cognitive level nor by their vocabulary level. Second, more than
a year later, at the end of first grade, the future experts in spelling were characterized by a strong reliance on phonological
mediation in silent reading as well as in oral reading and in the spelling of dictated words. Third, at the time when the two
groups were formed, at the end of second grade, the experts in spelling scored higher than the other participants in a silent and
in an oral reading comprehension task, but not in a spoken comprehension task.1 These results clearly showed that (a) only
reading specific components at the level of word recognition differentiated those two groups of children, (b) the early use of
phonological mediation contributed to the establishment of the orthographic lexicon, and (c) early metaphonological skills made
it easier to set up the phonological procedure.
These results might be explained by the fact that phonological mediation permits processing of all regular words, known or
unknown, as well as words that do not exist. Using phonological mediation and confronting the responses obtained by this
procedure with orally known words may permit the children to infer correspondences between graphemes and phonemes (GPC).
Noncontextual correspondences are probably learned before contextual ones (for example, c + i read /si/, whereas c + o is read
/ko/). However, as indicated by our data, some high-frequency contextual correspondences may be learned at the same time as
simpler ones (for example, o + u, read /u/). As French is a very regular language at the level of GPC correspondences, the
majority of words might be learned that way. As for irregular words, it is possible to consider that, on the one hand, these items
have only one irregular aspect (for example, in femme /fam/, only the first e has an irregular pronunciation). On the other hand,
the frequency of the irregular graphemes per se and the frequency of the words in which they are included, might influence
performance. Thus, children can learn progressively some rare correspondences (for example, e + m of femme is read /a/ as in
the -emment adverbs such as frquemment/frekam/). In this case, the inference of such a correspondence can be reasoned as
follows: The word /fem/ does not exist, but a very near word exists (/fam/), so e must be read a in this context. In this manner,
children
1 These tasks were elaborated by Lecocq (1996). See also Lecocq, Casalis, Watteau, and Leuvers (1996).

Page 355
probably learn the many relations between orthography and phonology, including the more opaque relations. Based on the
frequency of the graphemes and also on the word frequency, strong associations between phonological units and orthographic
units are thus created, permitting the progressive elaboration of an orthographic lexicon. However, even when the direct-
matching route between a word and the orthographic lexicon is functional, children may still use phonological mediation, which
becomes more and more automatic as the associations permitting the establishment of the grapheme phoneme correspondences
are reinforced (see Perfetti & Zhang, 1995). This learning scheme has the advantage of allowing the integration in the first stage
of the phonological mediation procedure of strategies using local alphabetical cues (see Ehri, 1992; Harris & Coltheart, 1986;
Stuart & Coltheart, 1988). However, it has two limitations. First, it does not allow an explanation of how, through associations,
through the reinforcement of the strength of some links, a qualitative change appears.
Second, this explanation cannot be used for spelling, in which, unlike reading, the oral lexicon is not of great help.2 For
example, knowing /bato/, which is the phonological form of the word bateau, can facilitate the reading, but not the spelling, of
this word. This is due to the asymmetry of the grapheme phoneme and phoneme grapheme correspondences. Whereas only one
reading is possible for this word, /bato/ could be spelled correctly, according to usual phonological rules, ba(t)to or ba(t)tau as
well as ba(t)teau (with one or two ts). Thus, the orthographic procedure develops later in spelling than in reading.
Our results have corroborated that hypothesis. On the one hand, the frequency effect was greater in reading than in spelling, and
if the lexicality effect was more marked in spelling, it was to the detriment of words. On the other hand, the words, but not the
pseudowords, were better read than spelled; the difference between word reading and word spelling increased between sessions.
Finally, the negative impact of orthography in spelling compared to reading was more noticeable in the spelling of complex
regular words and even more marked in the spelling of irregular words, with this last trend increasing over time. These data are
indicators that the orthographic procedure is more strongly used in reading. The reverse trend was noted for phonological
mediation in the fact that more regularization errors occurred in spelling as early as the first session. These results are in
accordance with those predicted by Frith's (1985, 1986) and Morton's (1989) models.
We observed another difference between reading and spelling in the processing of complex graphemes. At this level, our
hypothesis is that, if the processing units used by children at the beginning of acquisition are letters in relation to sounds, and
not graphemes in relation to phonemes,
2 Except in the case of morphological marks. For example, if you know the word chaton (kitten), you can correctly spell
chat with a t (see Leybaert and Alegria, 1995, on this topic).

Page 356
the graphemic complexity effect should disappear over time. In fact, the negative impact of graphemic complexity was observed
only in the first session, particularly in spelling. In the second session, complex graphemes never hurt the children's performance
and complex graphemes were processed even better than simple graphemes in word reading. These results might be explained
by the fact that, if the basic unit of GPC processing is the grapheme and the total number of letters is the same, when the item
contains one or more digraphs, the children have fewer units to assemble than when the items are only composed of simple
graphemes, for example, four units in route compared to five in table. In this case, they also have fewer phonemic units to
process to give an oral response (three in the first case and four in the second). Alternatively, in spelling, the use of phoneme
grapheme correspondences when items have complex graphemes necessitates the transformation of a simple unit (a phoneme) to
a complex one (a digraph). This latter operation might have a higher cognitive cost.
Beyond the differences noted between reading and spelling, the correlation analysis indicates a joint development of these two
skills. More generally, our data show that phonological mediation is dominant in the first stages of reading and spelling
acquisition in French and that the use of this procedure contributes to the establishment of the orthographic lexicon.
Conclusion
In sum, our data show that the primary process in the early stage of reading and spelling acquisition in French is phonological
mediation and that this mediation contributes to the development of the orthographic lexicon. Our results do not correspond
completely with any of those predicted in current developmental models. They are partly consistent with those stage models that
predicted the existence of a phonological mediation stage (see Frith, 1985, 1986; Harris & Coltheart, 1986; Morton, 1989;
Seymour, 1986). However they differ from these models in (a) finding no clear evidence of a logographic stage (see also
Sprenger-Charolles & Bonnet, 1996), and (b) observing, after the phonological mediation stage, a stage in which phonological
and orthographic procedures coexisted, with the weight of the orthographic procedure being less marked in spelling than in
reading.
Acknowledgments
This research was partially supported by a grant from the National Institute for Studies in Education, Paris, France to L.
Sprenger-Charolles and by a grant from the Natural Sciences and Engineering Research Council of Canada to L. S. Siegel. We
thank Eric Beltrando for designing the computer system for the reading tests.

Page 357
References
Byrne, B., Freebody, P., & Gates, A. (1992). Longitudinal data on the relations of word-reading strategies to comprehension,
reading time and phonemic awareness. Reading Research Quarterly, 27, 141-151.
Catach, N. (1980). L'orthographe franaise: Trait thorique et pratique [French orthography: Theory and practice]. Paris:
Nathan.
Catach, N. (1984). Les listes orthographiques de base du franais (LOB): Les mots les plus frquents et leurs formes flchies les
plus frquentes [Basic orthographic lists: High frequency words and their most frequent flexional forms]. Paris: Nathan.
Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993). Models of reading aloud: Dual route and parallel processing
approaches. Psychological Review, 100, 589-608.
Delattre, P. (1965). Comparing the phonetic features of English, French, German and Spanish. Heidelberg: Julius Gross Verlag.
Ehri, L. C. (1992). Reconceptualizing the development of sight word reading and its relationships to recoding. In P. Gough, L.
Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 107-143). Hillsdale, NJ: Lawrence Erlbaum Associates.
Ellis, W. E. (1984). Reading, writing and dyslexia: A cognitive analysis. London: Lawrence Erlbaum Associates.
Encrev, P. (1988). La liaison avec et sans enchanement: Phonologie tridimensionnelle et usages du franais [Liaison with and
without enchanement: Tridimensional phonology and French uses]. Paris: Seuil.
Foorman, B. R., Francis, D. J., Novy, D. M., & Liberman, D. (1991). How letter-sound instruction mediates progress in first-
grade reading and spelling. Journal of Educational Psychology, 83, 4, 456-469.
Foorman, B. R., Jenkins, L., & Francis, D. J. (1993). Links among segmenting, spelling and reading words in first and second
grades. Reading and Writing: An Interdisciplinary Journal, 5, 1-15.
Surface dyslexia: Cognitive and neuropsychological studies of phonological reading (pp. 301-330). London: Lawrence Erlbaum
Associates.
Frith, U. (1986). A developmental framework for developmental dyslexia. Annals of Dyslexia, 36, 69-81.
Frost, R. (1994). Prelexical and postlexical strategies in reading: Evidence from a deep and shallow orthography. Journal of
Experimental Psychology: Learning, Memory and Cognition, 20, 116-129.
Gak, V. G. (1976). L'Orthographe du franais, essai de description thorique et pratique [French orthography: A theoretical and
practical description]. Paris: S.E.L.A.F.
Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. Hillsdale, NJ: Lawrence Erlbaum Associates.
Gougenheim, G., Micha, R., Rivenc, P., & Sauvageot, A. (1964). L'laboration du franais fondamental (1er degr): Etude sur
l'tablissement d'un vocabulaire et d'une grammaire de base [Learning basic French (1st level): A study on setting up a basic
vocabulary and grammar]. Paris: Didier.
Gough, P. B., & Walsh, M. A. (1991). Chinese, Phoenicians, and the orthographic cipher of English. In S. A. Brady & D. P.
Shankweiler (Eds.), Phonological processes in literacy. A tribute to Isabelle Y. Liberman (pp. 199-209). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Harris, M., & Coltheart, M. (1986). Language processing in children and adults: An introduction. London: Routledge & Kegan
Paul.
Humphreys, G. W., & Evett, L. J. (1985). Are there independent lexical and nonlexical routes in word processing? An
evaluation of the dual-route theory of reading. Behavioral and Brain Sciences, 8, 689-740.

Page 358
Juilland, A., Brodin, D., & Davidovitch, C. (1970). Frequency dictionary of French words. The Hague: Mouton.
Kaye, J. D., & Lowenstamm, J. (1984). De la syllabicit [Sound pattern of language: Structures of phonological
representations]. In F. Dell, D. Hirst, & J.-R. Vergnaud (Eds.), Formes sonores du langage: Structures des reprsentations en
phonologie (pp. 123-159). Paris: Hermann.
Lecocq, P. (1996). L'E.C.O.S.S.E: Une preuve de comprhension syntaxico-smantique [S.S.C.T.: A syntactic-semantic
comprehension test]. Villeneuve d'Asq, Presses Universitaires du Septentrion.
Lecocq, P., Casalis, S., Watteau, N., & Leuvers, C. (1996). Apprentissage de la lecture et comprhension d'noncs [Learning to
read and understanding]. Villeneuve d'Asq, Presses Universitaires du Septentrion.
Leybaert, J., & Alegria, J. (1995). Spelling development in deaf and hearing children: Evidence for use of morpho-phonological
regularities in French. Reading and Writing, 7, 1-21.
neurons (pp. 43-66). Cambridge, MA: MIT Press.
Mowrey, R. A., & MacKay, I. R. A. (1990). Phonological primitives: Electromyographic speech error evidence. Journal of
Acoustic Society of America, 88, 1299-1312.
Paap, K. R., & Noel, R. W. (1991). Dual-route models of print to sound: Still a good horse race. Psychological Research, 53,
13-24.
Perfetti, C. A. (1994). Psycholinguistics and reading ability. In M. A. Gernsbacher (Ed.), Handbook of psycholinguistics (pp.
849-916). San Diego: Academic Press.
Perfetti, C. A., & Zhang, S. (1995). The universal word identification reflex. In D. L. Medin (Ed.), The psychology of learning
and motivation (Vol. 33, pp. 159-189). San Diego: Academic Press.
Raven (1947). Progressives matrices standard. E.A.P: Issy-les-Moulineaux (rdition, 1981).
Rieben, L. (October, 1993). Word-searching strategies in young children inside and outside the classroom. Second workshop of
the network on written language and literacy, European Science Foundation: Understanding early literacy in a developmental
and cross-linguistic approach, Wassenaar.
Savigny, M. (1974). Manuel (forme B) pour l'utilisation des tests BAT-ELEM [Manual (B form) for the use of the BAT-ELEM
test]. EAP: Issy-les-Moulineaux.
Seymour, P. H. K. (1986). Cognitive analysis of dyslexia. London: Routledge & Kegan Paul.
Seymour, P. H. K. (1990). Developmental dyslexia. In M. W. Eysenck (Ed.), Cognitive psychology: An international review
(pp. 135-196). Chichester: Wiley.
Seymour, P. H. K. (1994). Un modle du dveloppement orthographique double fondation [A double foundation model for
literacy acquisition]. In J. P. Jaffr, L. Sprenger-Charolles, & M. Fayol (Eds.), Lecture/criture: Acquisition; Les actes de La
Villette (pp. 57-79). Paris: Nathan.
Seymour, P. H. K., & Evans, S. H. M. (1991). Learning to read and write the names of classmates. Paper presented at the
meeting of the Cognitive Section of the British Psychological Society. Oxford, England, September 1991.
Share, D. L. (1995). Phonological recoding and self-teaching: Sine qua non of reading acquisition. Cognition, 55, 151-218.
Siegel, L. S. (1985). Deep dyslexia in childhood. Brain and Language, 26, 16-27.
Sprenger-Charolles, L., & Bonnet, P. (1996). New doubts on the importance of the logographic stage: A longitudinal study of
French children. CPC (Cahiers de Psychology Cognitive/Current Psychology of Cognition), 15, 173-208.
Sprenger-Charolles, L., & Casalis, S. (1995). Reading and spelling acquisition in French first graders: Longitudinal evidence.
Reading and Writing: An Interdisciplinary Journal, 7, 39-63.
Sprenger-Charolles, L., & Siegel, L. S. (in press). A longitudinal study of the effects of syllabic structure on the development of
reading and spelling in French. Applied Psycholinguistics.
Sprenger-Charolles, L., Siegel, L. S., & Bchennec, D. (in press). Phonological mediation, semantic and orthographic factors in
silent reading. SSR (Scientific Study of Reading).

Page 359
Treiman, R., Berch, D., Tincoff, R., & Weatherston, S. (1993). Phonology and spelling: The case of syllabic consonants.
Journal of Experimental Child Psychology, 56, 269-290.
Trsor de la langue franaise [Treasure of the French language]. (1971). Nancy: Centre National de la Recherche Scientifique.


Page 361
AUTHOR INDEX
A
Aaron, P. G., 126
Abbot, R. D., 284, 291
Abdi, H., 106, 113
Ager, J., 134, 148
Alegria, J., 117, 118, 119, 122, 126, 127, 225, 233, 278, 290, 293, 312, 315, 322, 326, 334, 336
Aliminosa, D., 47, 56
Allal, L., 133, 145, 148
Allen, D., 134, 148
Allport, D. A., 50, 55
Anderson, J. R., 99, 112, 113
Anderson, L. M., 129, 149
Anderson-Inman, L., 244, 265
Andr-Leickman, B., 12, 13, 18
Angelergues, R., 39, 56
Anis, J., 4, 18
Assal, G., 40, 55
Assink, E. M. H., 187, 189
Atkins, P., 178, 190, 340, 357
Azuma, T., 183, 189
B
Backman, J., 279, 290
Baddeley, S., 16, 18
Badecker, W., 47, 55, 56
Bailet, L. L., 187, 189
Baird, K., 100, 113
Ball, W., 261, 265
Baluch, B., 216, 218
Bargai, N., 199, 218, 221, 233
Baron, J., 221, 233
Barron, R. W., 195, 196, 218, 295, 315
Barry, C., 49, 55, 176, 182, 189, 195, 218, 295, 315, 330, 334
Basso, S., 40, 55
Baxter, D. M., 40, 41, 44, 46, 50, 55
Bean, W., 142, 148
Bear, D., 252, 253, 255, 269
Beaulieux, C., 16, 18
Beauvois, M.-F., 50, 55, 116, 127
Bchennec, D., 353, 358
Beck, I. L., 279, 293, 322, 336
Becker, W., 244, 265
Beers, C. S., 101, 113
Beers, J. W., 101, 113, 175, 191, 253, 265, 271, 272, 292
Behrmann, M., 54, 55
Bell, A., 129, 149
Bell, L.C., 209, 219, 279, 293, 322, 336
Bellugi, V., 100, 113
Benson, D. F., 39, 55
Bentin, S., 25, 38, 199, 202, 216, 218, 221, 233, 234
Benzecri, J. P., 301, 315
Berch, D., 68, 69, 80, 342, 359
Berent, I., 21, 179, 180, 183, 189, 192, 196, 197, 211, 216, 217, 218, 297, 317
Berko, J., 73, 78, 99, 100, 101, 103, 113, 158, 159, 170
Bereiter, C., 129, 150
Berman, R. A., 14, 18
Berninger, V.W., 284, 288, 291
Berset Fougerand, B., 145, 148
Bertelson, P., 225, 233, 278, 293, 322, 335, 336
Besner, D., 24, 37, 216, 218, 221, 233
Btrix Khler, D., 145, 148
Biedermann-Pasques, L., 16, 20
Bijeljac-Babic, R., 180, 193

Page 362
Biemiller, A., 274, 291
Bissex, G. L., 63, 78, 272, 273, 291
Blachman, B., 261, 265
Bled, B., 144, 148
Bloomfield, L., 3, 18
Boder, E., 116, 127, 320, 335
Boettcher, W., 65, 80
Bolla-Wilson, K., 41, 45, 55
Bonnet, P., 353, 356, 358
Bosman, A. M. T., 21, 30, 175, 177, 180, 184-186, 188, 189, 190, 191, 195, 218
Bottro, J., 9, 12, 18
Bouffler, C., 142, 148
Bourdin, B., 112, 113
Bowey, J. A., 69, 79, 155, 170, 252, 253, 265
Bowling, A., 126, 127
Bradley, L., 29, 31, 37, 69, 79, 174, 188, 190, 225, 226, 233, 234, 261, 262, 265, 277, 278, 279, 280, 286, 287, 288, 291, 292,
295, 297, 315, 322, 335
Brten, I., 188, 190
Breaux, A. M., 68, 80
Brittain, M. M., 99, 113
Brodin, D., 344, 358
Bronckart, J.-P., 132, 148
Brophy, J. E., 139, 148
Brown, A. L., 232, 233, 249, 265
Brown, A. S., 188, 190
Brown, G. D. A., 53, 55, 116, 127, 175, 178, 190, 290, 291, 320, 335
Brown, J. S., 132, 149
Brown, P., 221, 233
Brown, R., 99, 100, 113
Bruck, M., 125, 127, 128, 175, 190, 194, 226, 233, 250, 252, 264, 265, 269, 295, 315, 318
Bryant, P.E., 29, 31, 37, 69, 79, 152, 155, 170, 174, 177, 190, 225, 226, 230, 232, 233, 234, 261, 262, 265, 275, 277, 278, 279,
280, 286, 287, 288, 291, 292, 295, 296, 297, 315, 316, 322, 323, 335, 341, 342, 357
Bub, D., 43, 55
Bunce, F., 321, 333, 336
Burke, C., 253, 267
Butterworth, B., 51, 52, 55, 116, 127
Buttet, J., 40, 55
Byrne, B., 253, 261, 265, 266, 351, 357
C
Caesar, F. B., 186, 190
Caffarra, P., 48, 57
Campbell, R., 48, 51, 52, 55, 116, 127, 173, 175, 190, 230, 233, 252, 266, 295, 297, 315
apan, S., 14, 18
Capasso, R., 44, 48, 57
Caramazza, A., 41, 43, 44, 45, 46, 47, 48, 53, 54, 55, 56, 57, 178, 192, 195, 216, 217, 218, 219, 306, 315
Cardoso-Martins, C., 228, 233
Carello, C., 180, 190
Cary, I. L., 278, 293, 322, 336
Casalis, S., 231, 234, 298, 318, 353, 358
Cassar, M., 21, 52, 71, 72, 73, 74, 75, 76, 77, 79, 80
Castles, A., 321, 335
Catach, N., 5, 6, 7, 8, 9, 11, 16, 18, 97, 113, 140, 149, 340, 344, 357
Cataldo, S., 279, 283, 288, 291
Cazden, C. B., 99, 113
Champollion, J.-F., 5, 19
Chandler, P., 132, 150
Chervel, A., 97, 113, 138-140, 149
Chevallard, Y., 131, 149
Chiss, J. L., 4, 18, 129, 140, 149
Chomsky, C., 277, 291
Chukovsky, K., 279, 291
Clark, E. V., 279, 291
Clark, M. A., 143, 149
Clarke, L., 260, 266
Clesse, C., 300, 315
Coenen, M., 231, 233
Cohen, M., 6, 7, 19
Collins, A., 132, 149
Coltheart, M., 24, 37, 50, 55, 173, 175, 176, 178, 190, 274, 292, 298, 318, 321, 325, 326, 335, 336, 340, 341, 352, 355, 356,
357, 359
Coltheart, V., 252, 268
Content, A., 119, 127, 225, 226, 233, 278, 293, 312, 315, 322, 336
Cossu, G., 29, 31, 32, 37
Coulmas, F., 9, 12, 19
Covill, A., 133, 150
Cramer, B. B., 278, 279, 293
Croft, W., 13, 19
Cromer, R. G., 176, 190
Cronnell, B., 243, 266
Crossland, J., 225, 233, 322, 335
Cubelli, R., 48, 56, 57, 217, 218
Cummings, J. L., 39, 55
Cunningham, A. E., 278, 279, 280, 285, 286, 288, 289, 291, 293
Curtis, B., 178, 190, 340, 357
D
David, J., 140, 149
Davidovitch, C., 344, 358
Davidson, B., 238, 268
Davidson, J. A., 126, 128
Davies, W. V., 16, 19
Page 363
Davis, A., 143, 149
Davison, M., 116, 127
DeFrancis, J., 8, 11, 14, 19, 28, 37
de Groot, A. M. B., 175, 177, 180, 184, 285, 186, 189, 190
Delaney, S. M., 209, 219
Delattre, P., 340, 357
de Partz, M. P., 54, 56, 116, 296, 306, 320
de Ribaupierre, A., 301, 313, 316, 317
Drouesn, J., 40, 50, 55, 116, 127
Derwing, B. L., 67, 79
de Saussure, F., 3, 20
Desberg, P., 151, 170, 175, 192, 271, 274, 292, 297, 316
Desbordes, F., 17, 19
DeStafano, J., 129, 149
Deutsch, A., 202, 218
De Villiers, J., 99, 113
De Villiers, P., 99, 113
de Vries, R., 175, 184, 185, 191
Dewey, J., 132, 149
Ding, B., 279, 293
DiPardo, A., 129, 149
Dixon, E. M., 100, 114
Dixon, R., 244, 265
Djebbour, S., 144, 150
Doblhofer, E., 12, 19
Dodd, B., 176, 190
Douzenis, J. A., 39, 56
Dowker, A., 225, 233
Downing, J., 129, 149
Drake, D., 249, 266
Drum, P., 253, 268
Dubois, J., 97, 98, 113
Ducancel, G., 144, 149
Dugas, M., 126, 128
Duguid, P., 132, 149
Duighuisen, H. C. M., 175, 190, 193
Duncan, L. G., 327, 335
Durrell, D. D., 65, 79, 244, 269
E
Ehri, L.C., 21, 32, 37, 52, 64, 65, 70, 73, 77, 79, 117, 125, 127, 174, 176, 186, 188, 190, 192, 242, 243, 244, 254, 247, 249,
252, 253, 254, 256, 257, 258, 259, 260, 261, 266, 267, 268, 271, 272, 274, 282, 286, 287, 288, 291, 295, 297, 298, 299, 307,
313, 315, 316, 325, 326, 331, 335, 355 357
Elder, L., 275, 293, 298, 313, 318, 325, 336
Ellis, A. W., 43, 44, 56, 320, 321, 335
Ellis, N. C., 52, 116, 127, 175, 190, 277, 278, 279, 280, 283, 288, 290, 290, 291, 292
Ellis, W. E., 340, 357
Encrev, P., 340, 357
Englert, C. S., 129, 149
Esteve Serrano, A., 16, 19
Evans, H. M., 228, 234, 288, 293, 298, 318, 320, 321, 326, 327, 328, 331, 336, 348, 358
Evett, L. J., 340, 357
Exner, S., 39, 56
F
Fawcett, A. J., 112, 114, 174, 192
Fayol, M., 21, 36, 110, 111, 112, 113, 115, 133, 149, 299, 316
Feeman, D. J., 289
Feldman, L. B., 221, 233, 234
Ferguson, C. A., 99, 113
Fernald, J., 100, 113
Ferreiro, E., 297, 299, 316
Fvrier, J., 6, 7, 19
Fielding-Barnsley, R., 261, 266
Fields, H., 263, 268
Fijalkow, J., 299, 316
Fischer, F. W., 321, 335
Fitzgerald, J., 135, 149
Foltz, G., 238, 268
Foorman, B., 259, 267, 351, 357
Forster, K., 202, 218
Fountoukidis, D., 248, 267
Fowler, C. A., 217, 218, 321, 335
Fraca de Barrera, L., 221, 222, 234
Francis, D. J., 259, 267, 351, 357
Francis, J., 69, 79
Franklin, S., 50, 56
Fraser, C., 100, 113
Freebody, P., 351, 357
Freedman, S. W., 129, 149
Friedman, M. P., 151, 170, 175, 192, 271, 274, 292, 297, 316
Frith, C., 174, 191
Frith, U., 21, 24, 37, 52, 56, 116, 117, 125, 127, 151, 170, 174, 175, 191, 195, 218, 223, 229, 233, 253, 263, 267, 272, 274,
275, 276, 277, 278, 281, 282, 286, 288, 289, 292, 295, 296, 297, 298, 307, 309, 313, 314, 316, 326, 331, 335, 340-342, 346,
347, 352, 353, 355, 356, 357
Frost, J., 175, 191, 251, 261, 268, 287, 292, 322, 335
Frost, R., 21, 25, 38, 183, 196, 199, 202, 216, 218, 219, 221, 233, 234, 339, 357
Fry, E., 244, 248, 267
Funnell, E., 50, 55, 116, 127
G
Gak, V. G., 340, 344, 357
Gates, A., 351, 357
Page 364
Gaur, A., 7, 19
Geervliet-van der Hart, J. A., 185, 193
Geeves, E., 126, 127
Gelb, I. J., 5, 8, 9, 17, 19, 28, 38
Gentry, J. R., 64, 65, 70, 77, 79, 253, 267, 271, 272, 274, 292
Grard, C., 126, 128
Gernsbacher, M. A., 21, 38
Gibbs, A. L., 188, 191
Gibbs, P., 180, 182, 191
Gibson, E. J., 176, 191
Girolami-Boulimier, A., 97, 113
Glasspool, D. W., 53, 56, 178, 191
Gleitman, L. R., 115, 127
Glushko, R. J., 221, 234, 296, 316
Godart, L., 12, 19
Goldinger, S. D., 34, 38, 178, 193
Gombert, J. E., 155, 170, 221, 222, 234, 296
Gompel, M., 175, 184, 185, 191
Good, T. L., 139, 148
Goodman, K., 153, 170
Goodman, R. A., 41, 45, 46, 56
Goodman-Schulman, R. A., 45, 47, 56
Goswami, U., 83, 96, 152, 170, 221, 22, 225, 226, 227, 229, 230, 231, 232, 233, 234, 235, 252, 267, 269, 275, 278, 290, 292,
296, 316, 322, 323, 335, 339, 341, 342, 353, 357, 359
Gougenheim, G., 344, 357
Gough, P. B., 125, 127, 174, 176, 191, 243, 245, 248, 253, 256, 262, 267, 274, 282, 292, 295, 297, 316, 320, 331, 335, 342,
351, 357
Goulandris, N. K., 116, 125, 126, 127, 284, 292
Graham, S., 134, 135, 149
Gras, M., 16, 19
Graves, D. H., 129, 149
Graves, K., 329, 335
Graves, M. F., 99, 101, 113
Greenberg, D., 256, 263, 267
Griffith, P.L., 174, 191, 253, 256, 267, 282, 292, 295, 297, 316
Grossberg, S., 179, 191
Cruaz, C., 13, 19
Gugliotta, M., 31, 37
Guion, J., 16, 19, 138, 149
Guthrie, J., 262, 267
H
Hagge, C., 13, 19
Halford, G. S., 232, 234
Haller, M., 178, 190, 340, 357
Hanna, J. S., 136, 149, 244, 267
Hanna, P.R., 136, 149, 244, 267
Hansen, J., 252, 253, 265
Harris, M., 99, 100, 101, 114, 274, 292, 340, 341, 355, 356, 357
Harris, R., 4, 10, 19
Harris, W., 16, 19
Harste, J., 253, 267
Hatfield, F. M., 41, 43, 54, 56, 116, 127, 176, 191, 320, 335
Havelock, E. A., 9, 19
Hbrard, J., 300, 316
Hecaen, H., 39, 56
Heilman, K. M., 41, 57
Henderson, E. H., 21, 38, 62, 63, 64, 65, 70, 72, 73, 77, 79, 136, 149, 175, 191, 243, 253, 255, 265, 267, 271, 272, 292, 297,
316
Henderson, J. M., 126, 128
Henderson, L., 319, 335
Henry, M., 244, 267
Hill, A. A., 8, 19
Hill, S., 327, 335
Hillinger, M., 243, 267, 274, 292, 331, 335
Hillis, A., 44, 47, 49, 55, 56
Hindson, B. A., 67, 79
Hjelmslev, L., 4, 19
Hodges, R. E., 136, 149, 244, 267
Hogaboam, T., 125, 128
Hohn, W., 260, 267
Holligan, C., 175, 177, 191
Holmes, V. M., 175, 176, 191
Holt, L. K., 321, 335
Horn, E., 134, 135, 136, 149
Houghton, G., 53, 56, 178, 191
Howard, D., 50, 56
Hoyne, S. H., 133, 150
Hughes, C., 279, 293, 322, 336
Hulme, C., 231, 234, 268
Humblot, L., 299, 316
Hummer, P., 83, 96, 284, 294, 298, 318, 348, 353, 359
Humphreys, G. W., 340, 357
I
Inhelder, B., 232, 234
Istomina, Z. M., 289, 292
J
Jaffr, J.-P., 4, 14, 17, 19, 21, 28, 97, 113, 137, 140, 142, 144, 150
Janssens, J. M., 282, 293
Jarema, G., 112, 113
Jenkins, L., 188, 192, 351, 357
Johansen, L. S., 24, 38
Johnson, D., 252, 269
Johnson, K.E., 99, 114

Page 365
Johnston, R. S., 175, 177, 191
Jolivet, R., 40, 55
Jorm, A. F., 30, 38, 125, 127, 195, 219, 256, 259, 267, 268
Juel, C., 125, 127, 174, 191, 243, 253, 256, 267, 282, 288, 292, 295, 297, 316
Juilland, A., 344, 358
K
Karmiloff-Smith, A., 329, 335
Katz, L., 25, 38, 196, 199, 216, 218, 219, 221, 233, 234
Kavanagh, J. F., 319, 335
Kawamoto, A., 183, 191
Kaye, J. D., 340, 358
Keating, C., 282, 268
Keeney, T. J., 99, 100, 103, 113, 114
Kehayia, E., 112, 113
Kertesz, A., 43, 55
Kess, J. F., 99, 114
Kibel, M., 95, 96
Kilborn, K., 112, 114
King, P. V., 188, 190
Kirtley, C., 69, 79, 225, 234
Kliegl, R., 238, 268
Klima, E. S., 115, 127
Knapp, M. S., 132, 143, 144, 145, 150
Koriat, A., 216, 219
Kornfilt, J., 14, 19
Kort, J., 100, 114
Koziol, S., 99, 101, 113
Kramer, S. N., 15, 19
Kreiner, D. S., 176, 191, 195, 196, 219, 248, 267
Kremin, H., 50, 56
L
Landerl, K., 83, 96, 284, 294
Large, B., 277, 278, 279, 280, 292
Largy, P., 110, 111, 113, 133, 149
Lartigue, R., 144, 150
Laubstein, A. S., 67, 79
Laurent, J.-P., 129, 149
Lautrey, J., 301, 313, 316, 317
Laxon, V., 252, 268
Lemaire, P., 110, 111, 113
Lennox, C., 177, 191
Levy, Y., 99, 101, 114
Lewkowicz, N. K., 279, 292
Leybaert, J., 312, 315
Liberman, A. M., 321, 335
Liberman, D., 259, 267, 351, 357
Liberman, I. Y., 251, 268, 277, 279, 292, 293, 321, 335
Lie, A., 287, 288, 292
Link, K., 195, 219
Linortner, R., 83, 96, 284, 294
Liva, A., 299, 316
Lobrot, M., 118, 127
Logan, G. D., 112, 114
Longchamp, A., 299, 316
Loosemore, R. P. W., 53, 55, 178, 190, 320, 335
Lovegrove, W., 126, 127, 128
Lovell, K., 100, 114
Lowenstamm, J., 340, 358
Lukatela, G., 180, 190
Lundberg, I., 175, 191, 251, 261, 268, 278, 281, 287, 288, 292, 322, 335
Lyons, L., 4, 19
M
MacKay, I. R. A., 342, 358
Macken, M., 217, 219
MacLean, M., 69, 79, 225, 233, 234, 280, 292, 322, 335
Maclean, R., 259, 268
MacWhinney, B., 99, 114
Madden, N. A., 135, 150
Malus-Abramowitz, M., 250, 269
Manesse, D., 97, 113, 138, 139, 149
Manis, F. R., 321, 335
Mann, V. A., 271, 292
Manrique, A. M. B., 175, 192
Marcel, T., 75, 79, 175, 192
Marcie, P., 39, 56
Margolin, D. I., 43, 56
Marsh, G., 151, 170, 175, 192, 271, 274, 292, 297, 316
Marshall, J. C., 29, 31, 37, 40, 56
Martin, F., 126, 127
Mason, J. M., 65, 79, 254, 268
Masonheimer, P., 253, 268
Masterson, J., 175, 193, 221, 234
Matthews, R., 259, 268
Mattingly, I. G., 319, 335
McClelland, J. L., 53, 57, 178, 180, 192, 195, 219, 320, 328, 333, 336
McCloskey, M., 47, 48, 56
McConkie, G., 239, 268
McCutchen, D., 133, 150
McKenna, M. C., 142, 150
Mervis, C. B., 99, 114
Meyer, A., 300, 317
Meyer, J.-C., 129, 149
Miceli, G., 44, 45, 46, 47, 49, 55, 57, 216, 217, 218, 306, 315
Michalowski, P., 9, 10, 15, 19
Micha, R., 344, 357
Michel, Y., 145, 148
Mildes, K., 133, 150
Miles, T. R., 95, 96
Miller, J. W., 142, 150
Mitterer, J. O., 320, 335
Moats, L. C., 176, 177, 192

Page 366
Mommers, M.J.C., 174, 192, 282, 283, 285, 288, 293
Monsell, S., 50, 57
Morag, S., 201, 219
Morais, J., 225, 233, 278, 279, 290, 293, 321, 326, 334-336
Morris, D., 253, 262, 268, 271, 293, 298, 317
Morton, J., 41, 44, 57, 117, 127, 224, 234, 320, 331, 336, 340-342, 346, 347, 352, 353, 355, 356, 358
Mousty, P., 117, 118, 119, 122, 126, 127
Mowrey, R. A., 342, 358
Mullennix, J., 180, 193
Murphy, L. A., 126, 128
Muter, V., 227, 234
N
Nation, K., 231, 234
Naveh, J., 12, 15, 16, 19
Navon, D., 216, 219
Nearey, T. M., 67, 79
Needels, M. C., 132, 143-145, 150
Nelson, H. E., 175, 192
Nelson, K., 314, 317
Nelson, P., 126, 127
Nesdale, A. R., 155, 170
Newcombe, F., 40, 56
Newfield, M. K., 99, 114
Newman, S., 132, 149, 263, 268
Ng, E., 175, 176, 191
Nicolaci-da Costa, A., 99, 101, 114
Nicolson, R. I., 112, 114, 174, 192
Nie, H., 279, 293
Noel, R. W., 24, 38, 340, 358
Nolan, K. A., 43, 57
Novy, D., 259, 267, 351, 357
Nunes, T., 168, 170
O
Olofsson, A., 278, 292
Olson, A., 53, 57, 178, 192
Olson, B., 227, 235
Olson, R., 238, 263, 268, 320, 336
Ormrod, J. E., 188, 192
Oud, J. H., 282, 293
P
Paap, K. R., 24, 38, 340, 358
Patterson, K. E., 41, 44, 45, 56, 57, 116, 127, 176, 183, 191, 193, 320, 335
Peereman, R., 221, 234, 296, 317
Pennington, B. F., 178, 193, 196, 219, 320, 336
Perfetti, C. A., 28, 29, 38, 117, 125, 128, 152, 170, 180, 183, 189, 192, 196, 197, 209, 216, 217, 218, 219, 245, 268, 279, 280,
293, 295, 297, 313, 317, 319, 322, 326, 328, 336, 339, 355, 358
Perin, D., 256, 267, 277, 279, 293
Perney, J., 253, 259, 262, 268
Perregaux, C., 300, 317
Peterson, O. P., 175, 191, 251, 261, 268, 287, 292, 322, 335
Piaget, J., 232, 234, 313, 317
Pignot, E., 278, 290, 326, 334
Polk, J., 248, 267
Pollatsek, A., 126, 128
Porpodas, C. D., 174, 175, 192, 320, 336
Powell, B. B., 10, 17, 20
Puech, C., 4, 18
Pulgram, E., 8, 20
R
Rack, J., 263, 268, 320, 336
Radeau, M., 119, 127
Raphael, T. E., 129, 149
Rayner, K., 126, 128
Read, C., 69, 75, 79, 152, 170, 175-177, 192, 195, 219, 242, 255, 261, 268, 271, 279, 293
Rego, L. L. B., 155, 170
Reitsma, P., 242, 268
Reuchlin, M., 313, 317
Rhodes, L. K., 143, 149
Rich, G., 129, 149
Richmond-Welty, E. D., 64, 75, 80, 180, 193
Rieben, L., 52, 145, 150, 297, 298, 300, 301, 313, 316, 317, 353, 358
Rivenc, P., 344, 357
Robbins, C., 252, 254, 258, 266
Roberts, K. T., 186, 192, 261, 266, 268
Robins, R. H., 4, 5, 20
Robinson, R. D., 142, 150
Robinson, R. G., 41, 55
Roeltgen, D. P., 41, 45, 57
Roller, S., 138, 150
Romani, C., 46, 55
Romian, H., 129, 149
Rosch, E., 178, 194
Roth, S. F., 125, 128
Rouillard, P., 16, 19
Rozin, P., 115, 127
Rudorf, E., 136, 149, 244, 267
Rumelhart, D. E., 53, 57, 178, 180, 192
S
Saada-Robert, M., 52, 145, 148, 297, 298, 301, 317
Saenger, P., 16, 20
Saltmarsh, J., 242, 266

Page 367
Sampson, G., 5, 8, 11, 20, 28, 38
Sanders, K., 100, 114
Sanders, R. J., 46, 57
Sandon, J.-M., 144, 150
Saterdahl, K., 151, 170
Sauvageot, A., 344, 357
Savigny, M., 343, 358
Savin, H. B., 279, 293
Scardamalia, M., 129, 150
Schlanger, B. B., 99, 114
Schneuwly, B., 129, 132, 148, 149
Scholfield, P.J., 8, 20
Scott, J., 243, 253, 268
Scragg, D. G., 16, 20
Sebastin-Galls, N., 221, 234
Segarra, M., 16, 20
Segui, J., 296, 317
Seidenberg, M., 125, 128, 175, 177, 178, 183, 192-194, 196, 219, 295, 296, 318, 320, 328, 333, 336
Seron, X., 54, 56
Sevush, S., 41, 57
Seymour, P. H. K., 49, 51, 52, 55, 57, 174, 192, 228, 234, 275, 288, 293, 296, 298, 313, 318, 320, 321, 323, 325-331, 333,
334-336, 339, 341, 346, 348, 352, 356, 358
Shallice, T., 53, 56, 57, 116, 128, 178, 191, 320, 336
Shankweiler, D., 251, 268, 277, 279, 293, 321, 335
Share, D. L., 30, 38, 125, 127, 259, 268, 342, 351, 353, 358
Shepard, M. J., 185, 193
Sheperd, J., 260, 269
Shibatani, M., 14, 20
Shimron, J., 199, 202, 216, 219
Shurcliff, A., 176, 191
Shurtleff, H. A., 284, 291
Siegel, L. S., 175, 177, 191, 192, 348, 353, 358
Siegler, R. S., 107, 114, 298, 313, 318
Signorini, A., 175, 192
Simon, D. P., 183, 193, 249, 252, 268
Simon, H. A., 183, 193, 252, 268
Slaghuis, W., 126, 127, 128
Slavin, R. E., 135, 150
Slobin, D. I., 99, 113, 279, 293
Sloboda, J. A., 175, 193, 249, 268
Smith, F., 277, 293
Smith, L. B., 178, 193
Smith, M. R., 217, 218
Smith, N. D., 99, 113
Snowling, M. J., 95, 96, 116, 125, 127, 175, 193, 227, 234, 252, 263, 268, 277, 279, 293, 320, 336
Solomon, M., 99, 114
Speedie, L. J., 41, 55
Sprenger-Charolles, L., 231, 234, 298, 318, 348, 353, 356, 358
Stage, S. A., 68, 79
Stemberger, J. P., 67, 79
Stanovich, K.E., 30, 38, 95, 96, 125, 128, 263, 268, 279, 280, 285, 286, 288, 289, 291, 293, 320, 336
Stevens, R.J., 135, 150
Stone, G. O., 34, 38, 178-180, 182, 183, 191, 193, 196, 197, 205, 207, 217, 219
Strain, E., 183, 193
Stuart, M., 175, 193, 298, 318, 325, 326, 336, 355, 359
Studdert-Kennedy, M., 217, 219
Sulzby, E., 137, 150
Suzuki, H., 232, 234
Sweet, J., 298, 316
Sweller, J., 132, 150
Szeszulski, P. A., 321, 335
T
Taborelli, A., 40, 55
Taylor, S., 227, 234
Teixidor, J., 16, 19
Temple, C. M., 51, 52, 57
Templeton, S., 136, 137, 150, 252, 253, 255, 269
Thelen, E., 178, 193
Thompson, E., 177, 194
Thomson, M., 174, 193
Tincoff, R., 342, 359
Torgesen, J. K., 278, 294
Torneus, M., 286-288, 293
Torrey, J. W., 274, 293
Treiman, R., 21, 22, 38, 52, 67-77, 79, 80, 81, 82, 90, 95, 96, 174-176, 180, 190, 193, 195, 219, 226, 227, 233, 235, 243, 245,
252, 255, 261, 269, 322, 323, 336, 342, 359
Tromp, J., 184, 185, 191
Tunmer, W. E., 155, 170
Turco, G., 129, 150
Turvey, M. T., 180, 190, 221, 233
U
Uldall, H. J., 4, 20
Uhry, J. K., 185, 193, 260, 269
uit de Haag, I. J. C. A. F., 175, 193
Underwood, T. L., 187, 191
V
Vachek, J., 4, 6, 7, 9, 20
Valdois, S., 126, 128
Valtin, R., 175, 193, 279, 280, 293
Vanault, P., 126, 128
van Bon, W. H. J., 175, 190, 193

Page 368
van Daal, V. H. P., 185, 193
van der Leij, A., 185, 193
Van der Linden, M., 54, 56
van Doorn-van Eijsden, M., 185, 193
Vanhoy, M. D., 34, 38, 180, 193, 197, 205, 207, 219
van Leerdam, M., 175, 185, 186, 188, 189, 190
Van Leeuwe, J. F., 282, 293
Van Orden, G. C., 30, 34, 38, 177, 180, 183, 191, 193, 196, 197, 219
Varela, F. J., 177, 194
Varnhagan, C., 251, 269
Vellutino, F. R., 126, 128
Veltman, R., 15, 20
Venezky, R. L., 27, 38, 133, 150, 200, 219, 243, 244, 252, 255, 269
Venneri, A., 48, 57
Verhoeven, G., 187, 194
Vronis, J., 116, 119, 128
Vial, J., 138, 150
Vignolo, L. A., 40, 55
Villa, G., 45, 46, 55
Vygotsky, L. S., 133, 144, 150
W
Wagner, R. K., 68, 79, 278, 294
Wall, S., 278, 292
Walsh, M. A., 342, 351, 357
Wang, W.S.-Y., 28, 38
Warrington, E. K., 40, 41, 44, 46, 50, 55
Waters, G. S., 125, 127, 128, 175, 177, 181, 190, 194, 250, 252, 264, 265, 269, 295, 315, 318
Watteau, N., 358
Weatherston, S., 68, 69, 80, 342, 359
Weber, R. M., 274, 294
Welch, V., 151, 170, 175, 192, 271, 274, 292
West, R. F., 280, 293
Whitney Goodell, E., 217, 219
Wightman, J., 268
Wilce, L. S., 186, 191, 243, 245, 247, 253, 256, 258, 261, 266, 267, 277, 287, 288, 291, 297, 298, 316
Wimmer, H., 83, 95, 96, 233, 235, 284, 288, 294, 298, 318, 339, 348, 353, 359
Wingo, E. O., 12, 20
Winitz, H., 100, 114
Wise, B., 227, 235
Wolfe, J., 100, 103, 114
Woodward, V., 253, 267
Worden, P.E., 65, 80
Wright, A. D., 155, 170
Wright, S., 263, 268
Wylie, R., 144, 269
Y
Yee, A., 135, 150
Yonas, A., 176, 191
Young, A. W., 320, 335
Z
Zemblidge, J., 183, 191
Zesiger, P., 116, 296, 306, 320
Zhang, S., 28, 38, 180, 192, 297, 317, 355, 358
Zhang, Y., 279, 293
Ziegler, C., 13, 20
Zola, D., 239, 268
Zukowski, A., 67, 74, 75, 80
Zutell, J., 295, 318

Page 369
SUBJECT INDEX
A
Addressed route, 41, 44-45, 116-125
Agraphia, 39
Alphabetic stage of reading, 224, 254-256, 276, 296, 325-327, 341
Analogy and learning to read, 221-229, 232-233, 241, 252
Analogy and learning to spell, 229-233, 241, 252, 256
Aphasia, 39-40
Assembled route, 24, 41, 45-50, 116, 195-217, 296
C
Chinese, 11, 14, 23, 27-28
pin yin, 279
D
Decoding, 30
Dual-route model/theory, 24, 29, 32, 48-49, 296-297, 320, 340-342
Dutch, 184-188
Dysgraphia, 41, 43, 50-52, 116, 320
Dyslexia, 31, 50-51, 112, 117-126, 174-175, 321, 332-333
Dysphonia, 116, 320
E
English, 26-27, 36, 47, 81, 98-101, 168, 182-183, 320
spelling, 84-88, 93-94, 133-137
F
French, 11, 14, 26, 36, 97-113, 117-124, 130-133, 137-140, 144-145, 168-169, 339-340
orthography, 115-116
phonological mediation in, 345-347, 350-354
G
German, 81
spelling, 84-94
H
Hebrew, 14, 197-217
inhibition in, 203-217
I
Invented spellings, 152, 251-252, 254
Italian, 23, 25-27, 31-32, 36, 47
J
Japanese, 12, 23
L
Lexical route (see Addressed route)
Lexical spelling processes, 115-126
Linguistic awareness, 327-328, 333-334
Logographic scripts (see Writing systems, logographic)
Logographic stage of reading, 223-224, 230, 253, 275-276, 296-297, 324-325

< previous page page_370
Page 370
M
Memory use in spelling and reading, 251, 254-255
Models of reading, 29-30, 32-33, 37, 52, 178-182, 264-265, 274-278, 297-298, 320-321, 341-342, 346-348
Models of spelling, 32, 49, 52-54, 62-65, 77-78, 161-167, 178-182, 195-196, 201-207, 253-256, 264-265, 271-278, 313, 323-
332, 341-342, 346-348
Morphographic framework, 330-331
Morphological knowledge, 73-77
Morphology, 13
N
Number morphology, 97-113
O
Orthographic
buffer, 45-48
development, 323-332
framework, 328-330
knowledge, 70-73
neighborhoods, 221-223
stage of reading, 224, 297, 341
Orthography, 4, 6-8, 11, 15-17, 22, 25-28
deep, 11, 25, 36
shallow, 11, 24, 36
P
Phonological awareness, 30, 289, 321-323
and reading ability, 151-153, 278-285
and spelling ability, 289
training in, 286-287
Phonological processing, 321
Phonological route (see Assembled route)
Psycholinguistics, 21-22
S
Spelling, 22, 238
and connectionism, 53-54
development (see Models of spelling)
disabilities/disorders, 40, 51-53, 262-264
errors, 86-94, 159-160, 175-278, 187-188, 272-273
instructional approaches in, 83, 129-148, 186-189, 313-314, 334
and letter names, 65-70
and phonology, 175-176
processes, 40-43
and reading, 28-36, 174-175, 183-186, 238-265, 285-290, 295-298, 312-314, 346-356
and text production, 142-145
Syntactic awareness, 155-159
W
Word copying and writing, 298-315
Writing, 6-8
phonographic principle in, 9-11
semiographic principle in, 12-13
Writing systems, 7, 22-23
alphabetic, 23, 37, 243-244, 247-248
logographic, 11, 23, 319
syllabic, 23
< previous page page_370

Learning To Spell Research, Theory, and Practice Across PDF

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Learning To Spell Research, Theory, and Practice Across PDF

Enviado por

Direitos autorais:

Formatos disponíveis

cover next page >

title: Learning to Spell : Research, Theory, and Practice Across

cover next page >

intentionally left blank

LAWRENCE ERLBAUM ASSOCIATES, PUBLISHERS

< previous page page_iii next page >

< previous page page_iv next page >

< previous page page_v next page >

< previous page page_vi next page >

< previous page page_vii next page >

intentionally left blank

< previous page page_ix next page >

< previous page page_x next page >

< previous page page_xi next page >

< previous page page_xii next page >

< previous page page_xiii next page >

< previous page page_xiv next page >

< previous page page_xv next page >

< previous page page_1 next page >

intentionally left blank

Linguistics and Written Language

< previous page page_3 next page >

< previous page page_4 next page >

< previous page page_5 next page >

< previous page page_6 next page >

From Writing to Scripts

< previous page page_7 next page >

< previous page page_8 next page >

< previous page page_9 next page >

< previous page page_10 next page >

< previous page page_11 next page >

< previous page page_12 next page >

< previous page page_13 next page >

< previous page page_14 next page >

From Scripts to Orthography

< previous page page_15 next page >

< previous page page_16 next page >

< previous page page_17 next page >

< previous page page_18 next page >

< previous page page_19 next page >

< previous page page_20 next page >

< previous page page_21 next page >

Writing Systems and Languages

< previous page page_22 next page >

< previous page page_23 next page >

< previous page page_24 next page >

Orthographies Reflect Linguistic Structure

< previous page page_25 next page >

< previous page page_26 next page >

< previous page page_27 next page >

The Spelling-Reading Relationship Revisited

< previous page page_28 next page >

< previous page page_29 next page >

< previous page page_30 next page >

< previous page page_31 next page >

< previous page page_32 next page >

< previous page page_33 next page >

< previous page page_34 next page >

< previous page page_35 next page >

< previous page page_36 next page >

< previous page page_37 next page >

< previous page page_38 next page >

< previous page page_39 next page >