Escolar Documentos
Profissional Documentos
Cultura Documentos
Series Editors
Harald Clahsen
Lydia White
University of Essex
McGill University
Editorial Board
Melissa F. Bowerman
Luigi Rizzi
University of Siena
Katherine Demuth
Bonnie D. Schwartz
Brown University
Wolfgang U. Dressler
Antonella Sorace
Universitt Wien
University of Edinburgh
Nina Hyams
Karin Stromswold
Rutgers University
Jrgen M. Meisel
Jrgen Weissenborn
Universitt Hamburg
Universitt Potsdam
William OGrady
Frank Wijnen
University of Hawaii
Utrecht University
Mabel Rice
University of Kansas
Volume 30
The LexiconSyntax Interface in Second Language Aquisition
Edited by Roeland van Hout, Aafke Hulk, Folkert Kuiken and Richard Towell
Aafke Hulk
University of Amsterdam
Folkert Kuiken
University of Amsterdam
Richard Towell
University of Salford
TM
2003051906
<TARGET
<
R/ RE E F"toc"
F
DOCINFO AUTHOR ""TITLE "Table of contents"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150" VOFFSET "4">
"ack">
"tow">
"haw">
"cor">
"cra">
"duf">
"dyk">
"wil">
"sab">
"gre">
"hou">
Table of contents
Acknowledgments
vii
21
3. Perfect projections
Norbert Corver
45
69
97
129
151
175
197
219
< /R/TREARGET
E FF
"ni">
"si">
"toc">
vi
Table of contents
Name index
227
Subject index
229
<TARGET "ack"
</TARGET
"ack">DOCINFO AUTHOR ""TITLE "Acknowledgments"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Acknowledgments
February 2003,
The editors
<TARGET "tow" DOCINFO AUTHOR "Richard Towell"TITLE "Introduction"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 1
Introduction
Second language acquisition research
in search of an interface
Richard Towell
University of Salford
1.
Introduction
Richard Towell
In search of an interface
generative linguists have focused on syntax. That mainstream focus provides the
background for the articles by Hawkins and Liszka, Corver, Van de Craats and
Dueld in this volume.
However, during the same period, the study of language within psychology
also made rapid strides in exploring many aspects of psycholinguistics (see
Harley 2001). One of these has concentrated on the lexicon, as is demonstrated
in the chapters by Dijkstra and Williams in this volume; others have explored
issues of how language may be stored in the brain and have made use of
imaging techniques to enable us to begin to relate our theoretical analyses to
physical realities (Perani et al. 1996, Dewaele 2002). These are represented in
this volume by the chapters by Sabourin and Haverkort and by Green. There
has been the occasional fruitful interchange in some areas of the discipline from
time to time but there has been no real examination of how the two disciplines
have evolved with regard to SLA and whether there are more global reasons for
looking towards collaboration. There are now signs that both groups of
researchers are coming to an understanding that their particular view of the
world may not suce to account for the overall process and that each will have
to understand more about what the other knows.
Whilst this book does not pretend to complete that task, it will seek to
present current examples of the way linguists think about second language
acquisition and of the way researchers working within a more psychological
frame of reference think about the same subject in such a way as to show how
there is a degree of complementarity in the work being done, even if, at the
highest levels of argument, we are unlikely to see a swift return to the unity of
view of the 1950s (cf. Smith 1999: 174).
The belief that the time is right to seek such complementarity is encouraged
by two fairly recent developments. Within linguistics, the advent of the minimalist theory and its consequences for SLA has caused researchers to look again
at the relative roles of syntax and lexis. Under the minimalist view which, as
Corver demonstrates in chapter three, applies as much to interlanguages as to
any other natural languages, syntax is thought to be universal. It is constituted
of invariant principles with options restricted to functional elements and
general properties of the lexicon (Chomsky 1995: 170). The invariant nature of
the syntax is possible because the functional elements are now seen as part of
the lexicon: It is clear that the lexicon contains substantive elements (nouns,
verbs) And it is reasonably clear that it contains some functional categories. (Chomsky 1995: 240). More recently, SLA specialists (Hawkins 2001: 345,
Herschensohn 2000:80) have stated rather more rmly that, under minimalism,
Richard Towell
functional categories should be seen as part of the lexicon. We will argue below
that this change of emphasis Smith (1999) argues cogently that minimalism
is a natural evolution of the generativist enterprise rather than a revolution
may well modify the way in which linguists have to think about development in
SLA and indeed about the other central features of SLA research outlined above.
Within psychology, there has been a welcome renewal of interest in how
second languages are acquired. Since the mid 1980s, we have seen attempts to
account for second language learning drawing on a rich vein of inductive
research using computer modelling (Rumelhart and McClelland 1986). More
recently, a variety of non-intrusive ways of providing physical evidence of brain
processes have become available (Perani 1999). Both linguists and psychologists
have become interested in how knowledge is stored in the mind and how it is
retrieved from storage. Arguments have been put forward based on a distinction between declarative and procedural memory systems (Towell and Hawkins
1994, Ullman 2001) some of which suggest radical dierences in the way rst
and second languages are acquired and stored in the mind. Most recently,
arguments have been put forward to suggest that usage-based analyses can
account for all linguistic units: Psycholinguistic and cognitive linguistic
theories of language acquisition hold that all linguistic units are abstracted from
language use. In these usage based perspectives, the acquisition of grammar is
the piecemeal learning of many thousands of constructions and the frequencybiased abstraction of regularities within them (Ellis 2002: 144). All of this adds
to the view that a full account of second language acquisition will require
complementary input from both disciplines.
One of the main keys must lie in how we see the concept of development
and the psychological mechanisms which underlie development. The eorts of
syntacticians focus on describing the syntactic structure which lies behind the
interlanguage of the learner. This has always made it dicult for them to
account for development (see Gregg 1996): the placing of functional categories
within the lexicon makes this diculty more acute. The invariant syntactic
knowledge which learners have is a template present in the mind of the learner
which can be modied by the information inserted within it. There cannot be
a driving force for development in the syntax. It follows therefore that that
driving force really comes from the lexis. However, up to now the learning of
lexis has been thought of mainly in terms of one or other forms of associationist
learning theory, with connectionism being the most powerful. Theorists
pursuing this model have tended to argue that connectionist learning can
account for the totality of language learning, including the learning of syntax
In search of an interface
(see quotation from Ellis (2002) above). But syntacticians cannot accept that
the sophisticated structures which they observe and which provide no visible
clues on the surface structure of the language can be learnt in this empirical
fashion. They claim instead that innate knowledge (mediated or not by the L1)
must be guiding the learning (Hawkins 2001).
It is not clear that this argument can be resolved by theoretical debates
between nativists and non-nativists. We need to examine in detail the
evidence of how learners acquire a second language. This evidence must come
from a variety of sources using a variety of techniques. It will take us into
questions of what it is that learners acquire, how they acquire it and how that
process modies their linguistic capacity in both knowledge and use. Hawkins,
Corver and Van de Craats in this volume present clear empirical accounts of
how specic features of syntactic and lexical knowledge play a fundamental role
in second language development. Their accounts cannot, however, tell us
everything we need to know about the mental processes involved. Dueld asks
fundamental questions about the nature of the competence which is acquired.
Dijkstra provides an account of the way in which bilinguals store and access
their knowledge. Williams examines the way in which linguistic knowledge may be
built up on the basis of distributional evidence. Sabourin and Haverkort and then
Green look at the way in which the knowledge may be stored and used. In this way
we can see that a full account of the acquisition of a second language involves
the three systems outlined at the beginning of this chapter. We will argue that
none of the current arguments will suce alone to account for the total process
and that it is essential to attempt to integrate the sources of knowledge available
to us (see Jackendo (2002) for a similarly motivated position).
In this chapter we shall seek rst to outline the research context from the
generativist point of view, initially in general, and then specically with regard
to the contribution of minimalism. We will then examine in more detail the
contribution of psychological research and seek to show how a degree of
complementarity may well exist, if there is the will to look for it. In this way, we
are seeking to provide a context for the more detailed studies which gure in
the rest of the volume by means of which the value of their contribution can be
seen against the background of the evolving discipline of SLA research.
Richard Towell
In search of an interface
endowment is what enables the child to know more than the surface of the
language reveals: the surface forms act only as a trigger for the underlying
knowledge which the child already possesses.
It is important to highlight four signicant aspects of this theoretical
position as these will give rise to further comment below and will be dealt with
in subsequent chapters. First, a generativist approach involves idealisation of
the data to be examined. For the study of adult competence in the mother
tongue the researcher can frequently consult his or her own linguistic knowledge as a representative sample of the idealised speech community. This is not
possible in second language research. SLA research requires data gathering
methods which can isolate linguistic competence from performance factors.
Second, the primacy of syntax within the generative paradigm has led to a
separation between syntax and semantics. This is not without its problems as
more and more researchers are nding that semantic factors inuence syntactic
phenomena (Jus 1998, 2001). Dueld in this volume is concerned with how
competence can be successfully dened within this paradigm and proposes that
it is necessary to conceive of competence at two levels, one of which is more
related to the surface structures of language. Third, the conception of the
acquisition of syntactic knowledge through a process of triggering has given rise
to debate (Lightfoot 1993, Carroll 2000, Sakas and J. D. Fodor 2001). Hawkins,
Corver and Van de Craats, through an examination of the acquisition of
syntactic and lexical features, dene more clearly the nature of the features
which have to be learnt and discuss the role of the L1 in providing the initial
knowledge. This might provide a more satisfactory conceptual basis at least with
regard to the initial state. It does, however, leave open the question of how the
learners use empirical evidence to move to subsequent states. Fourth, the
generativist position for SLA has to adapt to the fact that a second language
acquirer has already learnt one language. The issue of how learners transfer or
access universal knowledge in cases where one language has already been learnt
is one which may have to be looked at again within the minimalist paradigm.
This is an issue for Hawkins and Liszka, for Corver, and for Van de Craats.
2.2 Generativist second language research
The particular strength of this approach has been in providing syntactic
analyses within a theoretical framework. This enabled SLA researchers to
predict in a precise way what learners needed to acquire in order to develop
their interlanguage system. During the 1990s, this was manifested mainly
Richard Towell
In search of an interface
10
Richard Towell
language faculty to another. Hawkins and Liszka feel that the evidence from
Chinese learners argues more in favour of a partial transfer perspective.
If Hawkins and Liszka are correct in their analysis, there are considerable
consequences for acquisition in a more general sense. If L2 learners really
cannot provide a feature based analysis for this part of their interlanguage
syntax, they will have to learn the required forms (and store them) in another
way. Whilst the discussion provided by Hawkins and Liszka remains within the
framework of the role of features within generative syntax, the discussion of the
alternative strategies available to learners provides food for thought as to
whether the learning of the non-integrated forms must then be carried out in a
dierent way e.g. stored as declarative as opposed to procedural knowledge as
discussed in the article by Green.
The notion of partial transfer also contrasts with the articles by Corver and
Van de Craats. They allow for the full transfer of both lexical and syntactic
features from the L1 to the L2. Their argument is that the full transfer of the L1
features into the L2 provides the starting point for learners beginning to acquire
an L2. This is combined with a conservative strategy i.e. one in which the
learners maintain those features unless and until they perceive the need to
modify them. The learners create a series of interlanguages, all of which remain
within the constraints of UG and thus provide checkable and interpretable
information for the internal and external interfaces. As they progressively
modify the features in response to the dierences which they perceive on the
basis of the input they receive from speakers of their target L2, their interlanguages move towards a point where their interlingual perfect systems
correspond more to the natural language used by the L2 speakers.
The above comments should serve to show that the generativist perspective
on second language acquisition has the power to create well-dened hypotheses
about the nature of language and to turn these into clearly dened investigative
strategies. The shift to minimalism means, however, that the learning of items
in the lexicon is potentially more signicant than it was previously. The
evidence from the Van de Craats article in particular shows how learners
progressively revise their featural specication of both functional and lexical
categories. As these are both part of the lexicon, it is clear that, from the
minimalist perspective, the way in which second language learners come to
revise their lexical forms especially those which are linked to functional
categories is now a more important issue. We have suggested above that it
provides the driving force for acquisition. The process of coming to know that
forms and features need to be revised, however, necessarily involves comparative
In search of an interface
perception of language forms. How the learners trigger their knowledge or how
they perceive the dierences is something which cannot easily be dealt with
within generative linguistics as currently dened, because it is more related to
performance than competence. As soon as we mention perception we need to
return to the psychological dimension and look again at second language
acquisition from that perspective.
11
12
Richard Towell
In search of an interface
terms and can operate in parallel (it is also known as parallel distributed
processing or PDP). The basic idea is to have (a large number of) processing
units which feed into one another at several levels (some of which are hidden)
in sequence. The units involved can be given levels of activation or weighting
prior to any simulation. These may be random or specied in relation to the
outcome envisaged. This will vary according to whether the model is being used
simply to replicate what is assumed to happen in certain forms of processing or
whether it is intended to model a developmental process (such as language
learning). In the former case, random levels will be initially assigned and it will
be intended that during the simulation the model will learn new levels of
activation. It is hoped that these levels of activation will represent eventually
some kind of reality. Put (over-)simply, the lowest level or input units are then
given dierent levels of activation and these levels feed through to the higher
levels. Where the interaction with higher levels is facilitated because certain
units at those levels already have positive activation (excitation), the signal
transmitted will be strengthened. Where the level of input activation encounters
negative activation at a higher level, it will be inhibited and, in the longer term,
weakened. Over very many trials, the simulation establishes a stable level of
output activation which is in essence derived through parallel processing from
the relative frequency of the input it has received. This is arrived at through
dierentiated interaction between signals at the intermediate levels on the basis
of the processes of strengthening and weakening.
In a simulated task learning context, models can be modied in such a way
as to include information about the extent to which the model is performing as
it should in relation to some target (back-propagation). This means that the
model can be trained to move nearer over time to a specied target. Researchers
can then see how the model responds to the information it has been given and
if and how it can progress towards the target. It should be noted that this
model only has one kind of unit which is linked to the signal strength of the
connections: there is no symbolic level within connectionist models. The
researchers can inspect the intermediate levels to see what activation levels the
model produced at dierent times, they will be aware of dierent stages of
learning and they will see the extent to which the desired outcome is obtained
and the relation with the input given.
There is no doubt that theorising alone cannot discover the eect of
enormous quantities of parallel processing and the computer is very ecient at
examining this eect. The sixty thousand dollar question, however, is whether
such modelling really reects anything which goes on in the human brain.
13
14
Richard Towell
Those who favour this kind of modelling argue that it parallels human learning,
that it can generalise beyond the data set which it is given and that it can
produce knowledge which is equivalent to what other approaches consider to
be abstract knowledge. It then follows naturally that proponents of this work
see no need for abstract symbolic categories in the mind because, from their
point of view, these are only the outcomes of the parallel processing described
above, not primitive units.
Numerous articles in the mid 1980s emanating from the PDP group led by
Rumelhart and McClelland (Rumelhart, McClelland and the PDP Research
Group 1986) made very strong claims about the way in which their network
could parallel human language learning. These drew a detailed response,
notably from Pinker and Prince (1988), in which the claims were thrown into
question. As Fodor and Pylyshyn put it: Pinker and Prince argue (in eect)
that more must be going on in learning past tense morphology than merely
estimating correlations since the statistical hypothesis provides neither a close
t to the ontogenic data nor a plausible account of the adult data on which the
ontogenic processes converge. It seems to us that Pinker and Prince have, by
quite a lot, the best of this argument. (Fodor and Pylyshyn 1988: 68). Connectionists have, however, gone on to rene their models and to enable them to call
on other sets of information through which they have renewed their claim to
model language.
These are the issues which are interestingly explored in the study by
Williams. He is critically interested in whether the learning which can be
demonstrated by computer modelling really models that of humans or not. To
investigate this, he combines computer-based simulations with experimental
investigations on humans. The context of learning is that of the specication of
abstract noun classes on the basis of a gender type classication. We already
know that there is considerable evidence available to show that second language
learners fail to assign gender in the consistent way that native speakers do.
There are two central issues. The rst is whether computer based inductive
processes can genuinely be said to have gone beyond exemplar based generalisations to the creation of abstract classes. The second is whether what computers
do and what humans do is in fact similar. Computers may very well demonstrate
a remarkable ability to generalise from distributional examples, but do humans
do the same? Do they rely on many examples to induce classes or do they
induce classes in other ways, such as by making use of other clues e.g. animacy?
Williams rst simulation of the training kind with feedback seemed to
show that the computer-based learning could generate productive knowledge
In search of an interface
of noun classes which enabled the network to generalise beyond the trained
exemplars. It could behave as if it had formed abstract representations. His
second simulation which did not involve feedback in the same way did not
learn as well. Humans who learnt to classify the data into noun classes at 66%
or above seemed to be using conscious explicit strategies, which clearly were not
available to the computers, and/or to be inuenced by prior knowledge of
gender languages, equally not true of the computers. The conclusions which
may be drawn are not simple: it seems as if some aspects of human behaviour
may be similar to the inductive generalising of computers but that humans
make use of other devices as well. Once feedback is introduced computer
learning is considerably more powerful, but then it is probably more powerful
than the human mind.
Green and Sabourin and Haverkort take us into yet other areas of psycholinguistic research. Their concern is with how linguistic knowledge may be
represented and stored in the mind. They share many reference points and to
some extent a methodology. They both discuss evidence based on aphasics, they
both rely on physiological evidence from ERPs to conrm or deny what other
sources have indicated. Their conclusions appear to be slightly dierent.
The broader argument which this research addresses concerns whether or
not L1 and L2 learners acquire and store language in the same way. Virtually all
researchers acknowledge the dierences in learning environment for many L2
learners: they are generally older (beyond the so-called critical age) with fully
developed memory and cognitive systems; they are often literate; they are
already in possession of a rst language and they are often exposed to explanations about language in classrooms as well as to language forms in more or less
authentic contexts of use. A key question has always been: do these dierences
mean that they will learn in a dierent way? Those who argue that they do
suggest that they rely more on explicit learning and that this will be reected in
the way their knowledge is stored in the mind. A separation is often made between
declarative knowledge, sometimes glossed as knowing that, the kind of knowledge which can be consciously accessed and articulated, such as a rule of grammar,
and procedural knowledge, sometime glossed as knowing how i.e. the kind of
knowledge which underlies skill activity, such as riding a bicycle, which cannot
be consciously accessed. The two forms of knowledge are said to be acquired in
dierent ways and stored in two dierent memories: a declarative memory and
a procedural memory each of which is accessed in a dierent way (Anderson
1983, 1993, 2000). Declarative knowledge is acquired explicitly, consciously
and quickly but cannot be used swiftly as a basis for any skill based action.
15
16
Richard Towell
In search of an interface
17
18
Richard Towell
Dijkstras article makes it very clear that when possessors of two languages
hear a given stimulus they activate all the relevant linguistic knowledge they
have without separating it out into L1 and L2. All the activated forms have the
possibility of giving rise to overt production: there is a selection process which
determines which will actually be produced. The level of activation relates very
much to frequency of use. This suggests that when second language learners
acquire new forms they will compete with the existing forms. There will be
dierences in activation depending on the context when dealing with words in
utterances (as opposed to the isolated words mostly studied) but the principle
will hold that the ability to use second language forms accurately in uent
utterances will depend at least to some degree on the frequency of use experienced by the user.
This raises interesting questions about how the activation of the interlingual
forms for the purposes of communication in the interlanguage permits modication of the interlanguage system. If the eect of use is merely to strengthen
the connections, as is implied by the non symbolic connectionist models
explored by Williams and implied in the work of Dijkstra, then how can use
give rise to modication of the forms? How also will they overcome the
competition of existing forms?
Sabourin and Haverkort suggest that whilst advanced L2 users may display
the same knowledge in grammaticality judgement tests, the knowledge that lies
behind their use of the language is not the same as that which lies behind the
use in language production. Williams in his discussion of the comparative
learning by humans and computers points out that humans appear to be
inuenced by factors other than the purely distributional. Is it possible that
second language learners do indeed have some dierential representation which
allows them not to have to rely purely on the distributional analysis? Or is there
another more symbolic version of implicit learning which could account for
modication rather than strengthening?
Green in his discussion of the storage of linguistic knowledge raises issues
about the relative contribution of declarative and procedural knowledge.
Within that area of reference, there are interesting questions about how learners
deal with those lexical and functional categories which are not fully integrated
within the syntax, as must be the case for interlingual systems which have not
yet fully developed the syntactic system. We have seen that Hawkins and Liszka
take the view that Chinese learners cannot integrate +/- past in the T category
of their interlanguage syntax. They nonetheless produce correct past tense
forms of regular verbs for at least some of the time. Where and how are these
In search of an interface
forms stored? If they are not generated by a function of the syntax, they must be
stored as separate lexical items. This immediately opens the door to the notion
that there must be a dierence between the proportion of knowledge which is
stored in the lexicon of the L1 and the L2 and in interlanguages: it would seem
probable that in the early stages of learning at least second language learners
must store a large proportion of the forms they have learnt as new lexical items
and only work out later how they may be part of the syntax.
Whilst there are no clear answers which immediately fall out of the existing
state of knowledge, it should be evident that a combination of the insights of
linguists and of psychologists will be required to answer these questions
properly.
References
Anderson J.R. 1983. The architecture of cognition. Cambridge, Mass.: Harvard University Press.
Anderson J. R. 1993. Rules of the mind. New Jersey: Lawrence Erlbaum.
Anderson J. R. 2000. Learning and memory. New York: John Wiley.
Bechtel, W. and Abrahamsen, A. 1991. Connectionism and the mind. Oxford: Blackwell.
Carroll, S. E. 2000. Input and evidence. Amsterdam: John Benjamins.
Chomsky, N. 1959. Review of Skinner 1957. Language 35: 2658.
Chomsky, N. 1986. Knowledge of language. New York: Praeger.
Chomsky, N. 1995. The minimalist program. Cambridge, Mass.: MIT Press.
Dewaele, J. M. 2002 Individual dierences in L2 uency: The eect of neurobiological
correlates. In Portraits of the L2 user, V. Cook (ed.), Clevedon: Multilingual Matters.
Ellis, N. 2002. Frequency eects in language processing: A review with implications for
theories of implicit and explicit language acquisition. Studies in Second Language
Acquisition 24 (2): 143189.
Flynn, S. 1987. A parameter-setting model of L2 acquisition. Dordrecht: Reidel.
Fodor, J. A. and Pylyshyn, Z. W. 1988. Connectionism and cognitive architecture. In
Connections and symbols, Special Edition of Cognition, S. Pinker and J. Mehler (eds), 373. Cambridge, Mass.: MIT Press.
Gass, S. and Selinker, L. 2001. Second language acquisition. New Jersey: Lawrence Erlbaum.
Gregg, K. 1996. The logical and developmental problems of second language acquisition.
In Handbook of second language acquisition, W. Ritchie and T. Bhatia (eds), San Diego:
Academic Press.
Harley, T. 2001. The psychology of language. Hove: Psychology Press.
Hawkins, R. 2001. Second language syntax. Oxford: Blackwell.
Hawkins, R., Towell, R. and Bazergui, N. 1993. Universal Grammar and the acquisition of
French verb movement by native speakers of English. Second Language Research 9: 189233.
Herschensohn, J. 1999. The second time around minimalism and L2 acquisition. Amsterdam:
John Benjamins.
19
</TARGET "tow">
20
Richard Towell
Hulstijn, J. 2002. Towards a unied account of the representation, processing and acquisition of second language knowledge. Second Language Research 18 (3): 193224.
Jackendo, R. 2002. Foundations of language. Oxford: Oxford University Press.
Jus, A. 1998. The Acquisition of semantics-syntax correspondences and verb frequencies
in ESL materials. Language Teaching Research 2: 93123.
Jus, A. 2001. Verb classes, event structure, and second language learners knowledge of
semantics-syntax correspondences. Studies in Second Language Acquisition 23: 305313.
Lightfoot, D. 1993. How to set parameters. Cambridge, Mass.: MIT Press.
Mitchell, R. and Myles, F. 1998. Second language acquisition theories. London: Arnold.
Paradis, M. 1997. The cognitive neuropsychology of bilingualism. In Tutorials in bilingualism.
Psycholinguistic perspectives, A.M.B. de Groot and J.F. Kroll (eds), New Jersey: Lawrence
Erlbaum.
Perani, D. 1999. The functional basis of memory: PET mapping of the memory systems in
humans. In Cognitive neuroscience of memory, L. G. Nilsson and H. J. Markovitsch
(eds), 5578. Seattle: Hogrefe and Huber.
Perani, D., Dehaene, S., Grassi, F., Cohen, L. Cappa, S. F. and Dupoux, E. 1996. Brain
processing of native and foreign languages. Neuroreport 7: 24392444.
Pinker, S. 1997. How the mind works. Harmondsworth: Penguin.
Pinker, S. and Prince, A. 1988. On language and connectionism. Analysis of a parallel
distributed processing model of language acquisition. Cognition 28: 73195.
Rumelhart, D. and McClelland, J. 1986. On learning the past tense of English verbs. In
Parallel distribued processing: Vol 1. Foundations, D. Rumelhart, J. McClelland and the
PDP Research Group 1986. Cambridge, Mass.: MIT Press.
Rumelhart, D., McClelland, J. and the PDP Research Group. 1986. Parallel distribued
processing: Vol 1. Foundations. Cambridge, Mass.: MIT Press.
Sakas, W. G. and Fodor, J. D. 2001. The structural triggers learner. In Language acquisition
and learnability, S. Bertolo (ed.), 172234. Cambridge: Cambridge University Press.
Schreuder, R. and Weltens, B. (eds). 1993. The bilingual lexicon. Amsterdam: John Benjamins.
Smith, N. 1999. Chomsky: Ideas and ideals. Cambridge: Cambridge University Press.
Skinner, B. F. 1957. Verbal behaviour. New York: Appleton Century Crofts.
Towell, R. and Hawkins, R. 1994. Approaches to second language acquisition. Clevedon:
Multilingual Matters.
Ullman, M. 2001. The neural basis of lexicon and grammar in rst and second language:
The declarative/procedural model. Bilingualism: Language and Cognition 4: 105122.
Waxman, S. 1996. The development of an appreciation of specic linkages between
linguistic and conceptual organisation. In The acquisition of the lexicon, L. Gleitman
and B. Landau (eds), Cambridge, Mass.: MIT Press.
White, L. 1989. Universal Grammar and second language acquisition. Amsterdam: John
Benjamins.
White, L. 1991. Adverb placement in second language acquisition: Some eects of negative
evidence in the classroom. Second Language Research 7: 13361.
Winograd, T. 1972. Understanding natural language. New York: Academic Press.
<LINK "haw-n*">
<TARGET "haw" DOCINFO AUTHOR "Roger Hawkins and Sarah Liszka"TITLE "Locating the source of defective past tense marking in advanced L2 English speakers"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 2
1.
Introduction
It is well-known that advanced L2 speakers of English from certain L1 backgrounds show persistent optionality in marking thematic verbs for simple past
tense in spontaneous oral production, as for example in The police caught the
man and take him away. Speakers whose L1 is Chinese appear to be such a
group. Bayley (1991, 1996) found the phenomenon suciently robust in
Chinese speakers to undertake a variationist analysis of the factors which might
be causing it. Wolfram and Hateld (1984) had found similar optionality in the
L2 English of Vietnamese speakers. More recently Lardiere (1998a, 1998b, 2000)
has reported remarkably consistent optionality in simple past tense marking in
a near-native speaker of English sampled with an eight-and-a-half year interval.
Patty, a native speaker of Mandarin and Hokkien, marked simple past tense on
thematic verbs around only one-third of the time in data collected from
spontaneous speech. Native speakers, by contrast, do not apparently show
endemic optionality of the same kind, although failure to inect a verb for past
tense is found sporadically in slips of the tongue (Fromkin 1988).1
An important question for theories of SLA which assume that the mental
grammars of individual L2 speakers are derived from Universal Grammar (UG)
is why such optionality might exist in advanced/near-native speakers. Given
that the morphophonology of forms in English provides clear positive evidence
that past tense is marked, it is unexpected that advanced/near-native speakers
should continue to have problems with it. Locating the source of the diculty
would be a small contribution to the broader goal of determining exactly how
the language faculty is involved in the construction of grammatical knowledge
by older L2 learners.
Beck (1997) has argued that the kind of optionality in question is unlikely
22
23
24
25
26
As a consequence, bare Vs in Chinese can be interpreted either as past or nonpast, depending on context:
(1) Zhangsan kan dianying.
Zhangsan see movie
Zhangsan is seeing OR saw a movie.
3. The study
To test the claim that the source of optionality in tense marking by L2 speakers
lies at the interface between the syntactic and morphological components of the
language faculty, we selected advanced L2 speakers of English from dierent L1
backgrounds, devised a test aimed at measuring informants knowledge of the
morphological processes involved in simple past tense marking, and collected
a sample of spontaneous production data including simple past tense verb
forms from the same informants.
3.1 Informants
Advanced L2 speakers of English were selected for this study on the basis of
their performance on an independent measure of general prociency. This
consisted of two components: (a) the written multiple-choice grammar test
component of the Oxford Placement Test (Allan 1992) which has 100 items
covering a range of the core morphosyntactic properties of English; (b) Nations
(1990) vocabulary levels test, which was designed as a language teachers aid
for giving help with vocabulary learning, but provides a rough notional measure
of the size of a speakers vocabulary up to the 10,000-word level. Using this
combined grammar/vocabulary test, we selected informants whose L1 was
Chinese, Japanese or German and whose mean scores broadly matched at the
upper end (over 80% correct). This produced an experimental set of informants
of two Chinese, ve Japanese and ve German speakers.3 Details of the prociency scores are given in Table 1.
3.2 Test of knowledge of morphology
If Becks (1997) ndings are generalisable, we expect to nd that the informants in
our study can productively manipulate morphological processes involved in past
Range (%)
86.6
85.7
90.7
83.489.7
83.487.0
85.896.0
tense marking in English. To test this we designed a task which required informants to inect both real and invented (nonce) verb stems for simple past tense.
Our reasoning was that if speakers know the verb morphology associated with past
tense, it should make no dierence whether real or nonce forms are involved.
3.2.1 Design
The task was adapted from one used by Prasada and Pinker (1993) with native
speakers (experiment 3 in their study). Prasada and Pinker were interested in
the extent to which natives would produce regular and irregular past tense
forms when presented with nonce verbs, and in particular whether novel
irregulars displaying the prototypical phonological shape of partially productive past tense irregulars (like string, sling, ing, cling strung, slung, ung,
clung) would elicit novel irregular past tense forms, such as spling splung. To
elicit such responses, informants were presented with nonce forms and an
invented denition for each six at the top of each page of a test questionnaire
followed by six sentences, each with a blank, where one of the nonce forms
belonged (Prasada and Pinker 1993: 24).
In our test we were interested simply in whether informants would inect
verbs appropriately for past tense in clear past tense contexts, and whether their
knowledge was generative in the sense of allowing them to inect verbs they had
never encountered before correctly. They were therefore presented with six
verbs at the top of each page of a test questionnaire with denitions (as in the
Prasada and Pinker study), but half were real and half invented (18 of them in
fact taken from the set of prototypical regular and irregular nonce verbs used by
Prasada and Pinker). Below each set of six verbs with their denitions were six
sentences where informants had to decide which verb to insert, and what its
form should be. These contexts required either a simple past or a present
perfect form. Thus informants could not simply produce a past tense form by
rote, but had to make a denite choice of tense appropriate to the context. A
partial illustration of the form of the test is given in (2):
27
28
(2) SPLING:
There were 120 contexts in all, 60 of which were expected to elicit simple past
tense verb forms, and of these 30 involved nonce verb stems (15 prototypical
regular types, like blark, and 15 prototypical irregular types, like spling). A
control group of native speakers (n = 5) also took the morphology test.
3.2.2 Results
Table 2 displays the frequencies of inected and uninected real and nonce
verbs in simple past tense contexts.
A 2 test comparing the total frequency of inected and uninected forms
produced by the non-native groups shows no signicant dierence between
Table 2.Frequencies of real and nonce verbs inected for simple past
L1
Chi (n = 2)
Jap (n = 5)
Ger (n = 5)
Eng (n = 5)
Verb type
Score (%)
Score (%)
Score (%)
Real reg
Yes
No
23
3
88.5
11.5
60
2
96.8
3.2
64
1
98.5
1.5
74
0
.100
.0
Real irreg
Yes
No
29
0
.100
.0
64
2
97.0
3.0
65
0
.100
.0
73
0
.100
.0
25
2
92.6
7.4
58
2
96.7
3.3
64
0
.100
.0
72
2
97.3
2.7
18
4
81.8
18.2
54
9
85.7
14.3
53
7
88.3
11.7
71
0
.100
.0
Total
95
9
91.3
8.7
236
15
94.0
6.0
246
8
96.9
3.1
290
2
99.3
0.7
Yes
No
NB: Cases where an informant selected a verb form other than simple past (e.g. past progressive or
perfect) are excluded from the table. Hence scores do not necessarily add up to the expected maximum.
29
30
Japanese (n = 5)
German (n = 5)
Score
(%)
Score
(%)
Score
(%)
Inected 25
Uninect. 15
62.5
37.5
137
12
91.9
8.1
52
2
96.3
3.7
40
.100
149
.100
54
.100
Irregular Inected 64
Ininect. 12
84.2
15.8
252
18
93.3
6.7
79
4
95.2
4.8
Total
.100
270
.100
83
.100
Verb type
Regular
Total
76
These results are problematic for the view that L2 speakers generally have
diculty mapping phonological forms (vocabulary items) with layers of
morphological features onto terminal nodes generated by the syntax. If this
were the case, we would expect all three groups to perform similarly on the
spontaneous production task, but the Chinese speakers are performing signicantly dierently from the Japanese and German speakers.4
Is the dierence the result of L1 phonology interfering with and depressing
the performance of the Chinese informants? This is unlikely. Recall that the
prediction was that if L1 phonology had such an eect, we would expect the
Chinese and Japanese to experience similar problems because the relevant
property (absence of word-nal consonant clusters) is present in both L1s.
However, to pursue this possibility further, we considered performance of the
non-native speakers on consonant clusters elsewhere in their spontaneous
production, specically in monomorphemic words like most, kind. If word-nal
clusters are problematic in production, some evidence for this should surface in
these forms. Table 4 compares informants retention of -t/-d in inecting
regular past tense verbs with their retention of -t/-d in monomorphemes.
Frequencies are small, but suggestive. Although Chinese speakers do drop
nal -t/-d in monomorphemes, they do not do so as frequently as in the simple
past. This result matches a similar nding in Bayley (1996), who in a group of
20 L1 Chinese speakers of intermediate and advanced prociency in L2 English
found a 65% retention of nal -t/-d in monomorphemes versus a 44% retention in simple past tense verb forms. This is in marked contrast with the pattern
Table 4.Absence of word-nal -t/-d in monomorphemes and regular simple past tense
forms compared
L1
Chinese
Japanese
German
Word type
-t/-d
Score
(%)
Score
(%)
Score
(%)
Monomorphemes
Present
Absent
9
2
82
18
27
1
96
4
48
0
100
0
Simple past
(Regular)
Present
Absent
25
15
63
37
137
12
92
8
52
2
96
4
31
32
Japanese
German
Word type
-t/-d
Score
(%)
Score
(%)
Score
(%)
Participles
Present
Absent
10
0
100
0
23
0
100
0
55
0
100
0
Monomorphemes
Present
Absent
9
2
82
18
27
1
96
4
48
0
100
0
Simple past
(Regular)
Present
Absent
25
15
63
37
137
12
92
8
52
2
96
4
both regular and irregular thematic verbs for past tense than the Japanese or
German speakers. The Chinese speakers retained word-nal -t/-d with monomorphemes more often than with regular past tense verb forms (suggesting the
problem is not caused by word-nal consonant clusters). The Chinese speakers
were perfect in inecting past participles, in contrast to simple past tense verb
forms, suggesting that performance pressures are unlikely as a source of
omission of inections with past tense verbs.
4. Discussion
We are interested in locating the source of optionality in the marking of
thematic verbs for simple past tense in oral production by certain groups of
high prociency L2 speakers of English, for example speakers of L1 Chinese. To
address this problem comparative data were collected from L1 speakers of
Chinese, Japanese and German matched for general prociency as advanced
speakers of English. An important caveat in discussing the results is that the
number of informants was small, therefore the generalisability of any conclusions drawn must be treated with caution. The data were elicited from a test of
knowledge of past tense morphology, and tasks allowing free oral production
(an anecdote and the retelling of a lm). Results of the morphology test suggest
that all three groups, including the Chinese speakers, know the morphological
properties of past tense marking in the context of real English verbs, and are
productively aware of the regular versus irregular distinction even with verbs
they have never encountered before (nonce forms). This is consistent with
earlier ndings of Beck (1997), who showed that the reaction times of nonnative speakers on a verb inection task were parallel to those of native speakers
on the same task. This led her to conclude that the source of optionality of past
tense marking is not located in the operation of the morphological component,
a claim with which we concur.
The comparative results from the free oral production of our three experimental groups show that the Chinese speakers are signicantly dierent from
the other two groups, producing uninected regular verbs in unambiguously
past tense contexts in over one third of cases. This was prima facie evidence
against the idea that L2 speakers in general have diculty mapping phonological forms (vocabulary items) onto syntactic terminal nodes where that
mapping implicates layers of morphological structure. If this were a general
problem for L2 speakers, and on the assumption that morphological properties
do not transfer from the L1 to the L2, we expected all three groups to show
evidence of it. By looking at the Chinese speakers performance on word-nal
consonant clusters in monomorphemic words, and their performance in
inecting past participles, we also established that optionality in producing nal
-t/-d was lesser in these contexts. This suggested that the source of the problem
with simple past tense was unlikely to be phonological in nature or the result of
performance pressure because similar optionality would be expected across
these contexts too.
Having eliminated the morphological component, and the mapping of
morphophonological forms onto syntactic representations as likely sources of
observed optionality in past tense marking in the informants studied, an
alternative explanation is needed. In this nal section we explore the consequences of assuming that optionality results from a failure of L2 learners to
include a syntactic feature for tense, that we are calling [past], among the
features which make up the lexical item T. Where this is the case, T enters
syntactic derivations without the feature which, for native speakers of English,
eventually forces the insertion of vocabulary items inected for past tense into
terminal nodes. Given such a hypothesis, we need to account for why optionality is characteristic of the Chinese speakers in our sample, but not the Japanese
or German speakers; why the Chinese speakers nevertheless have greater than
chance success in marking verbs for past tense in past tense contexts; and why
the Chinese speakers might be more successful in marking irregular past forms
and regular participles than in marking regular past forms. Central to this
account is an understanding of the relation between how tense is interpreted by
33
34
the semantic component, and the presence of a syntactic (formal) tense feature
in T. We therefore examine this relationship more closely.
Languages appear to dier in whether they grammaticalise tense through
special morphophonological forms or not. Chinese is standardly assumed to
lack such morphophonological forms, although it does have verbal aspect
markers like -le, -guo and -zhe (Li and Thompson 1981, Li 1990, Packard 2000).
For example, as already noted in Section 2, the verb kan see can be freely
interpreted as present or past depending on context:
(3) a.
Given such specications, only walked can be inserted into a terminal node with
the features [+nite, +past], even though walk is unspecied for those features
and is in principle available for insertion. The reason is that although both
forms have feature specications which are non-distinct from the terminal
node, walked is the more highly specied item.
With such an account of the interaction between syntax and morphology,
it is easy to see how the proposal that adult L2 speakers have diculty with the
35
36
with simple past tense forms because they involve a syntactic feature missing
from T, but not participle forms because they do not involve T. In the case of
irregular past tense forms, one possibility is that the Chinese speakers have
acquired them as items independent from the equivalent bare V forms; i.e. ran
is not the past tense exponent of run as it is in native grammars, but an independently acquired word form, in the same way that, say, amble and saunter are
independent word forms for native speakers. While this is speculative, there is
some limited evidence consistent with it in the transcripts of the spontaneous
production data. There are some cases of doubly inected verb forms such as
in (5a). The same speaker who produced (5a) also used ran in clearly non-past
environments ((5b) and (5c)):
(5) a. The girl ranned not far away.
b. You should ran away together.
c. She could not ran any more.
If ran were an independent word form it could be inected for past tense and be
used in non-past contexts. Observe also that (5b) and (5c) involve the use of ran
as a non-nite verb. Given the model of vocabulary insertion assumed, ran
could not be an inected variant of run specied for the feature [+past]. If it
were, its feature specication would clash with the feature specication of the
non-nite terminal nodes into which it is inserted in (5b) and (5c). Insertion
should simply be impossible.
If past participles and irregular forms like ran have a dierent morphological status in the L2 English grammars of Chinese speakers from regular past
tense verb forms like walked, this would be one kind of answer to the question:
why are Chinese speakers more successful in marking irregular past tense verbs
and regular participles than regular past tense verbs? They are not treating them
as past tense verb forms at all; rather they have independent lexical item status
for Chinese speakers. However, given this account, we have little idea currently
of what these forms might mean for Chinese speakers.
This leaves the need to give an account of optionality in the marking of
regular thematic verbs for past tense. We have no real answer to give in this
case. Firstly, we have to assume that there is some reason why Chinese speakers
do not treat regular verbs as independent vocabulary items as in the case of
irregulars. One possibility is that it is the very regularity of the inection and the
frequency of such forms in the input which forces a morphological analysis of
them as rule-based variants of the bare V. The problem then is to explain why
some forms, but not others, are inected in spontaneous production. To give a
37
38
avour of the problem, consider the following short extract from the transcript
of one of the Chinese informants in our sample:
When I saw the lm Lonely and Hungry and it reminded me of the old time
when life was very hard. Some people they were very hungry and they have no
work to do. They really dont want to steal. But they had no other choice and
when they become so hungry and they really want to just get the food and just
want to eat the food so they stole or they just do something bad. But its
understandable. It was not their mistake And I watch it maybe 20 years ago.
But it I didnt remember clearly about what it talk about. I just laugh a lot.
But now when we see the lm, I think. It gave me much imagination.
What might be prompting this speakers use of the bare verb forms want, watch,
talk and laugh in a context where past tense reference is often clearly intended?
We have little idea yet of what the answer might be, but some possibilities
appear unlikely. One explanation for selective verb inection that has been
advanced in studies of the early stages of L2 acquisition is the Aspect Hypothesis. This maintains that English past markers in early grammars are associated preferentially with verbs/predicates which have telicity as part of their
meaning (achievements and accomplishments in the terminology of Vendler
(1967)). See Bardovi-Harlig (1999) for a review of work on this topic. Does the
same pattern obtain in the high prociency Chinese speakers investigated here?
That is, are they treating the regular past tense form as a marker of inherent
verbal/predicate aspect? A breakdown of inected regular forms by verb type is
given in Table 6.
Table 6.Inected past tense regular verbs by aspectual type: Chinese speakers
Inected tokens
%
Statives
Activities
0/4
0%
10/14
71%
Accomplishments Achievements
1/2
50%
14/20
70%
The results are not conclusive. While statives are not inected at all, which
would be consistent with the Aspect Hypothesis, activities (which are atelic) are
inected to the same degree as achievements (which are telic). Interestingly,
Lardiere (2002) has also analysed Pattys past tense verb forms in terms of
whether there is a correlation with the telicity of the verb/predicate, and found
that 40% of all telic verbs (112/277, covering both regulars and irregulars) were
marked for past tense, while 35% of all atelic verbs (130/371) were past-marked.
The surface ordering of clitic clusters in French also appears to be the eect of a
post-syntactic output condition (Perlmutter 1971). If two third person clitics cooccur, an accusative form precedes a dative, but if a rst or second person and
a third person clitic co-occur, a dative form precedes an accusative (as in (7)):
(7) a.
Elle le
lui
donne.
she it-acc him-dat gives
She is giving it to him.
39
40
b. Elle me
le
donne.
she me-dat it-acc gives
She is giving it to me.
This suggests that the language faculty allows some monitoring of surface
strings. Perhaps in the normal case for Chinese speakers the morphology inserts
bare verb vocabulary items into T-V strings. But because output checking is a
possibility available in the grammar, Chinese speakers monitor the ambient
discourse for pastness and insert V-ed forms when they are able to detect it.
So, as in French where ordering of object clitics is determined by the person of
the forms in question (rst and second person must precede third person), for
the Chinese speakers selection of a V-ed form is determined by pastness. The
dierence between the two cases is that whereas [person] is a feature present in
the specication of the French pronoun vocabulary items, pastness has to be
determined on the basis of context.
Because context is involved, the monitoring process is unstable and gives
rise to the kind of apparently random use of bare and inected verb forms
illustrated in the sample of informant speech given above. This is consistent
with two observations about the marking of regular verbs for past tense by
Chinese speakers. First, individual speakers of high prociency can dier
markedly in the extent to which they are successful in inecting verbs. Our
informants are apparently more successful in spontaneous oral production than
Patty, for example. Secondly, dierent modalities allow a greater degree of
success in past tense marking by the same individual. So our informants were
more successful on the morphology test than in oral production, and Lardiere
(2002) reports that Pattys written output (in the form of e-mail messages)
shows higher proportions of past tense marking than her spoken output.
5. Conclusion: The interface between syntax and the lexicon in adult SLA
In a comparison of the performance of three groups of highly procient L2
speakers of English (with Chinese, Japanese and German as L1s), it was found
that the marking of past tense was optional in the spoken English of the Chinese
speakers, but not in the case of the Japanese and German speakers. With the
caveat that caution is required in generalising from small numbers of informants,
we argued that the Chinese speakers knowledge of English morphological
processes is intact (as Beck (1997) had already argued), and that phonological
properties (i.e. -t/-d deletion) did not appear to be a factor for our speakers.
Furthermore, we found that past participle and irregular past tense forms were
inected more consistently than regulars.
Although previous studies have argued that L2 speakers can fully acquire
the syntactic features of lexical items like T (Lardiere 1998a, 1998b, 2000,
Prvost and White 2000), we explored the alternative possibility that Chinese
speakers cannot establish [past] on T in English precisely because this feature
is absent in their L1. By contrast [past] is present on T both in Japanese and
German. If Chinese speakers do not have access to such a feature in the construction of L2 knowledge, but Japanese and German speakers do, this would
explain the observed dierence between them in inecting thematic verbs for
past tense in spontaneous production.
We linked this decit in the Chinese speakers grammars to the idea that
optional syntactic features, when selected, block the free application of
semantic operations (Chierchia 1998, Takeda 1999). Our account assumes that
these optional properties are subject to a critical period (an idea which originates in the work of Tsimpli and Roussou (1991) and Smith and Tsimpli
(1995)). We then explored, in highly speculative mode, reasons why Chinese
speakers might be successful at all in marking thematic verbs for past tense.
This line of enquiry raises an interesting possibility for the interface
between the syntactic component and the lexicon in adult SLA. Within the
spirit of minimalist enquiry, it might be proposed that all of the resources and
computational procedures of the language faculty LF, syntax, morphology
are intact and operative in adult SLA. However, optional (parametrised)
syntactic features which play a major role in tying morphosyntactic structure to
semantic operations in the L1 acquisition of specic languages are unavailable
if not activated in early life. In parsing L2 data, older L2 speakers can use all of
the resources except for these features. Phonological forms (vocabulary items)
which are selected by such features in native grammars are not selected by those
features in the grammars of L2 learners beyond the critical period. Highly
procient speakers must nevertheless nd some motivation for them. We
rejected some of the conceivable ways in which Chinese speakers might
motivate past tense verb forms in English (the Aspect Hypothesis and the
Discourse Hypothesis), and we suggested that they might be operating in
terms of an output condition which monitors the T-V string in relation to the
pastness of the discourse.
41
<DEST "haw-n*">
42
Notes
* Parts of this work have been presented to audiences at McGill (2000) and GASLA (2000).
We are grateful for the comments of those present in both cases. We have also beneted
greatly from the comments of Donna Lardiere on an earlier draft discussion of the ideas
outlined here. She will not agree with our conclusions, but her views have helped us sharpen
our thinking on a number of issues.
1. The phenomenon of -t/-d deletion from word-nal consonant clusters attested widely in
informal varieties of native English (Labov 1989), is strongly disfavoured in the context of the
regular simple past tense (Bayley 1996: 109). We return to this below, since it appears that the
pattern is the reverse in non-native speakers: -t/-d is more likely to be missing from the
regular past tense forms of verbs than in other contexts.
2. The assumption underlying the predictions made in this paragraph is that the morphological features of vocabulary items do not transfer from the L1 to the L2. Donna Lardiere (p.c.
and 2000: 116117) points out that the assumption is not self-evident. If the L1 exhibits the
same (or higher) degree of [morphological] complexity [L2 learners] may know what to look
for. Whether L2 speakers do or do not transfer morphological properties from their L1 is
obviously an empirical question. If it turns out that they do, the observations reported in this
article will need to be reinterpreted. Given that evidence bearing on the empirical question
is currently lacking (although see Jarvis and Odlin (2000) for some relevant discussion), for
the purposes of this article it will be assumed that transfer of verbal morphological properties
from L1 to L2 is not a factor.
3. The small size of the Chinese group is a reection of the diculty we had in locating L1
Chinese informants who can achieve a score of 80% or above on the prociency test (rather
than a diculty in nding L1 Chinese speakers of L2 English per se). Since, however, for the
purposes of the present investigation, the frequency of past tense marking in non-nativespeaker English is the focus of interest, the variation in group size is relatively unimportant.
In further work it would be desirable to have larger samples of advanced English speakers of
the L1s in question.
4. Unless, of course, L2 speakers whose L1s are morphologically complex know what to look
for. See footnote 2.
5. However, it should be noted that the present perfect ich habe gekauft predominates
in everyday spoken German in what would be simple past contexts in English (Comrie 1976).
6. A question that might be raised here is whether this shouldnt predict random usage by
the Chinese speakers of all three forms (Donna Lardiere p.c.). This would be the case only if
L2 speakers assumed that morphological variants of a lexical item can occur in free variation.
However, we argue subsequently that Chinese speakers attempt to establish representations
which distinguish morphologically distinct forms. The problem for them is that they have to
do so without the benet of the syntactic feature [past].
7. In practice the criteria seem to allow for considerable variation between raters in deciding
which events are in the foreground and which in the background. We had diculty applying
them to our own sample, and do not report the results here.
References
Allan, D. 1992. Oxford placement test. Oxford: Oxford University Press.
Bardovi-Harlig, K. 1995. A narrative perspective on the development of the tense/aspect system
in second language acquisition. Studies in Second Language Acquisition 17: 263291.
Bardovi-Harlig, K. 1999. From morpheme studies to temporal semantics: tense-aspect
research in SLA. Studies in Second Language Acquisition 21: 341382.
Bayley, R. 1991. Variation theory and second language learning: Linguistic and social constraints on
interlanguage tense marking. Unpublished doctoral dissertation, Stanford University.
Bayley, R. 1996. Competing constraints on variation in the speech of adult Chinese learners
of English. In Second language acquisition and linguistic variation, R. Bayley and D.
Preston (eds), 97120. Amsterdam: John Benjamins.
Beck, M-L. 1997. Regular verbs, past tense and frequency: tracking down a potential source
of NS/NNS competence dierences. Second Language Research 13: 93115.
Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6:
339405.
Chomsky, N. 1998. Minimalist inquiries: the framework. MIT Working Papers in Linguistics 15: 156.
Chomsky, N. 1999. Derivation by phase. Ms. MIT.
Chomsky, N. 2001. Beyond explanatory adequacy. Ms. MIT.
Comrie, B. 1976. Aspect. Cambridge: Cambridge University Press.
Embick, D. and Noyer, R. 2001. Movement operations after syntax. Linguistic Inquiry 32:
555595.
Fromkin, V. 1988. The grammatical aspects of speech errors. In Linguistics: The Cambridge
survey, Vol. II, F. Newmeyer (ed.), 117138. Cambridge: Cambridge University Press.
Halle, M. and Marantz, A. 1993. Distributed morphology and the pieces of inection. In
The view from building 20: Essays in linguistics in honor of Sylvain Bromberger, K. Hale
and S. J. Keyser (eds), 111176. Cambridge, MA: MIT Press.
Hansen, J. 2001. Linguistic constraints on the acquisition of English syllable codas by native
speakers of Mandarin Chinese. Applied Linguistics 22: 338365.
Jarvis, S. and Odlin, T. 2000: Morphological type, spatial reference, and language transfer.
Studies in Second Language Acquisition 22: 535556.
Labov, W. 1989. The child as linguistic historian. Language Variation and Change 1: 8598.
Lardiere, D. 1998a. Case and tense in the fossilized steady state. Second Language Research
14: 126.
Lardiere, D. 1998b. Dissociating syntax from morphology in a divergent L2 end-state grammar. Second Language Research 14: 359375.
Lardiere, D. 2000. Mapping features and forms in second language acquisition. In Second
language acquisition and linguistic theory, J. Archibald (ed.), 102129. Malden, MA:
Blackwell.
Lardiere, D. 2002. Second language knowledge of [past] vs. [nite]. Paper presented at
GASLA 6, University of Ottawa.
Lasnik, H. 1999. Minimalist analysis. Malden, MA: Blackwell.
Li, A. Y-H. 1990. Order and constituency in Mandarin Chinese. Dordrecht: Kluwer.
43
</TARGET "haw">
44
<LINK "cor-n*">
<TARGET "cor" DOCINFO AUTHOR "Norbert Corver"TITLE "Perfect projections"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 3
Perfect projections*
Norbert Corver
Utrecht University
1.
46
Norbert Corver
knowledge of language L if he is able to form linguistic expressions (i.e. soundmeaning pairs) that are fully-interpretable at the interface levels.
This requirement that the linguistic expressions (PF-LF pairs) generated by the
computational system be legible to the external systems plausibly holds for any
state of L1-knowledge (both interlanguage knowledge and nal state grammatical
knowledge).1 In all stages of language acquisition, the language system (i.e the
grammar) must interact with the external systems. If it does not, the linguistic
knowledge is not usable at all and consequently there would not be any output
products. Thus, from this interface perspective we could say that the L1-products
in any stage of language acquisition are perfect linguistic objects (PF-LF pairs)
in the sense that they fully consist of interface-interpretable properties.
What does the interface perspective on linguistic knowledge contribute to
our view on second language knowledge (and second language acquisition)? In
fact, the conclusion seems inescapable that L2-expressions are also perfect
grammatical objects. If they were not, the L2-objects generated by the computational system would not be legible and usable at all by the external systems
which interact with the L2-grammar. There simply would not be any output
(i.e. utterances). The conclusion must be that L2-products are interpretable
both on the meaning side and the sound side. And just like it does for the
various L1-knowledge states, this conclusion should also hold for the various
knowledge states of the L2-grammar (initial state, interlanguage states and
target state).2
Of course, the conclusion that L2-representations are natural language
objects, in the sense of being objects that fall within the bounds of Universal
Grammar, is not new. Over the last two decades, various researchers interested
in investigating the linguistic competence of L2-learners have argued and tried
to show that the (mental) L2-representations generated by the interlanguage
grammar are constrained by principles of UG (and consequently can be
analysed in the same way as other (e.g. L1 linguistic data; cf. Bley-Vroman 1990,
Schachter 1989, White 1988, 2000). The importance of the interface perspective
on linguistic expressions is that the conclusion of UG-consistency seems
inescapable: if interlanguage representations did not obey the bare output
conditions (i.e. the interface legibility requirements), these representations
would be illegible and inaccessible to the external cognitive systems with which
the interlanguage grammar interacts. The linguistic expressions generated by
the interlanguage grammar would simply not contain the right instructions to
the performance systems, and consequently, they would not be able to put it to
use. And as a consequence of that, there would not be any output.
Perfect projections
Thus, the L2-system, just like the L1-system, is a perfect system in the sense
of being a system that is optimally designed to meet external conditions (bare
output conditions) imposed by other cognitive systems that the language
faculty interacts with. From a dierent perspective, though, linguistic expressions produced by L2-learners seem to be highly imperfect. They very often
deviate, to a greater or lesser extent, from the linguistic expressions produced by
adult mother tongue learners of the language that is acquired. Some L2-expression E with meaning representation LFx very often diers from the equivalent
L1-expression generated by a mother tongue speaker of the language.
Consider, for example, the imperfect L2-expressions in (1a)(4a), which
are produced by Turkish L2-learners of Dutch.3,4 The b-examples represent the
corresponding target expressions.
(1) a.
(2) a.
Altijd uh alles
woonte Klirsehir van Turkije.
always uh all/everything live(d) Klirsehir of Turkey
Everyone still lives in Klirsehir in Turkey. (L2-expression)
b. Nog altijd woont iedereen in Klirsehir in Turkije.
still always lives everyone in Klirsehir in Turkey
Everyone still lives in Klirsehir in Turkey. (target expression)
(3) a.
Ik gaan school.
I go-inf school
I go to school. (L2-expression)
b. Ik ga
naar school.
I go-1sg to school
I go to school. (target expression)
(4) a.
47
48
Norbert Corver
Let me briey dwell on the example in (5), which from a supercial perspective
looks like a perfect target Dutch pattern (surface perfection). In possessive
structures of native speakers of Dutch, the element van is an adpositional (i.e.
prepositional) marker that is interpreted as the spell-out of the abstract genitive
case feature associated with the possessor DP in postnominal position (like in:
Perfect projections
dat boek van Jan, that book of Jan). A Dutch-based analysis of the sequence
examen van tolk is highly unlikely, however; under such an analysis, in which
tolk is the complement to the noun examen, the entire noun phrase needs to be
interpreted as the examen taken by the interpreter. This is not the reading it
has, which is the interpreter at/of the exam; examen acts as the possessor.
Given this semantic interpretation, Van de Craats, Corver and Van Hout reach
the conclusion that the syntactic structure associated with the linear sequence
in (5) is a structure transferred (i.e. conserved) from the rst language (i.e.
Turkish). This amounts to an analysis according to which van is an inectional
sux attached to the possessed noun. It is the equivalent of the inectional
element -nin in Turkish expressions like Ayse-nin araba-si (Ayse-gen car-3sg,
Ayses car). The Turkish syntactic structure which underlies the Dutch surface
sequence in (5) is then the one in (6a). (6b) represents the syntactic structure of
the Turkish sequence Ayse-nin araba-si.
(6) a. [DP [AgrP [examen-van]i [Agr [NP ti tolk] Agr]] D]
b. [DP [AgrP [Ayse-nin]i [Agr [NP ti araba] si]] D]
49
50
Norbert Corver
(8) a.
tolk
examen van
interpreter exam of
the interpreter of/at the exam (unattested)
b. zn lamp auto
his light car
the cars light (unattested)
Perfect projections
For the lexicon-syntax interface, this implies that each lexical item (i.e. a
constellation of lexical features) must be legible to the computational system
(merge and attract/move) that accesses these objects (i.e. lexical expressions)
and builds more complex expressions (i.e. syntactic expressions) from those
lexical expressions. A question which then arises is: What makes a lexical item
legible to the computational system (i.e. the rules of grammar: e.g. merge,
move/attract, agree)?
To answer this question, let us rst address the question of what a lexical
item (LI) is. In line with De Saussures conception of words, a LI is typically
dened as a sound-meaning pair (i.e. a PF-LF pair). In a sense, a LI is a structured object with a sound representation (the phonological matrix; sound
properties) and a meaning representation (semantic properties).
Presumably it is not the phonological and purely semantic properties which
make a lexical item legible to the computational system. If a LI were just a
sound-meaning pair one could wonder what the syntax (i.e. the recursive
procedure) should do with it. That is, what would make such a sound-meaning
pair legible to the computational system? Phonetic and purely semantic features
do not seem to be accessed by the recursive syntactic procedures. In short, these
PF-LF-pairs would remain illegible to the computational system.
So, what makes these sound-meaning pairs legible to the computational
rules, which combine these pairs into more complex sound-meaning constructs? The answer is: formal (i.e. syntactic) features. Suppose a formal feature
must be added (merged) to the sound-meaning pair (i.e. the lexical item) for
the LI to be legible to the computational processes that generate larger structures. In other words, merger of a categorial feature with the sound-meaning
pair turns the lexical item into an object that is visible at the lexicon-syntax
interface (cf. Marantz 1997, Chomsky 2000b).
Thus, what you have at the clausal level (syntax as a mediating representation between sound and meaning) is also what you have at the word level:
(9) meaning
phonology
syntactic-formal feature
51
52
Norbert Corver
Surface imperfections in the L2 derivational output can now be due to miscategorisation (i.e. mis- from the perspective of the target-language): an
incorrect categorial feature is associated with some sound-meaning pair. In
(10)(13), some examples of miscategorisation by L2-learners are given:
(10) Hier komt weg. Ik beetje momentes.
here comes away I a-bit moment-infl
Here he goes away. I wait a bit.
(11) Ik komt huis en Slenol wegt
huis Stokhasselt.
I come home and Slenol away-3sg house Stokhasselt
I go home and Slenol went to his house in Stokhasselt.
(12) A: In de buit
ligt die.
in the outside lie those
Outside lay these.
I: Hm?
lack of understanding
A: In de buiten. Buiten ook heeft de steen.
in the outside outside also has the stone
There are also stones (at the) outside.
(13) a.
In (10) and (11), moment and weg are treated as roots (i.e. sound-meaning
pairs) that receive a verbal character after merger of a verbal categorial feature.
Schematically (order irrelevant):
v
(14)
moment/weg
After attachment of the verbal categorial feature to the root, the lexical item
displays verbal behavior: it carries, for example, the verbal inection -t (present
tense, third person singular). From the perspective of the (grammar of the)
L2-learner, there is nothing odd about these lexical items: they each represent
a root (a sound-meaning pair), which carries a categorial feature that turns it
into a verbal form that is legible at the lexicon-syntax interface, in the sense that
Perfect projections
(16)
buit(en)/verzeker
The verbal analysis of items like momentes, wegt and bint and the nominal
analysis of buit(en) and verzeker are imperfect from the perspective (of the
grammar) of the mother tongue speaker of Dutch. For him, binnen (meaning:
inside) and buiten (meaning: outside) are both prepositions; moment
(meaning: moment) is a noun and verzeker (meaning: to insure) is a verbal
form. In short, the wrong categorial feature is associated with the root in these
L2-expressions. From the perspective of the L2-learners interlanguage grammar, however, these non-target-words are perfect projections: assignment of a
categorial value makes these items legible at the lexicon-syntax interface and
accessible to the computational rules that take these objects as their input. The
noun buit, for instance, is merged with the determiner de.
53
54
Norbert Corver
Perfect projections
(cf. (17a))
(cf. (17b))
(cf. (17c))
(20) a.
Alles mensen #
toerist ja.
all people (were) tourists yes
b. Ken je in Nederlands alles stad?
know you in Netherlands every city
c. Alles ja uh kinderen niet Turks spreken.
all yes uh children not Turkish speak
All children dont speak any Turkish.
55
56
Norbert Corver
In (19) and (20), the form alles is used instead of the target form allen and alle,
respectively. The L2-learner has identied the universal quanticational
meaning of alles, but he has not discovered yet that alles has a non-human
interpretation and that it cannot occur as a quanticational determiner. Or to
put it dierently, the L2-learner has identied the quantier feature associated
with alles, but other lexical features, like its categorical feature (e.g. alle being a
quanticational determiner and alles being a noun-like expression), do not
seem to have been identied yet.
Importantly, the quanticational expressions in (19) and (20) are LF-interpretable: they receive a restrictive reading (i.e. a set of objects is dened over
which they range) and bind a variable at LF. Thus, an L2-expression like (19c)
has the following LF-structure:
(21) For all xi, [xi: people], xi come here today.
Ja nu alleen mag.
yes now alleen may
Yes now, everything is permitted.
b. Ik wil alleen leren.
I want alleen learn
I want to learn everything.
(23) a.
The item alleen in (22) and (23) receives a clearly (universal) quanticational
interpretation. In (22), it receives the interpretation everything; in (23), it is
interpreted as everyone. Also in these examples, then, the L2-learner has
identied the universal quanticational element al.
Perfect projections
Contrary to the L2-expression alleen in (22) and (23), the target language item
alleen can never occur in an argument position:
(25) *Ik kende alleen.
I knew alone
57
58
Norbert Corver
distribution and interpretation (e.g. nonhuman versus human) may dier from
that of the target lexical items alles and alleen, the conclusion seems inescapable
that these quanticational expressions are fully legible at the LF-interface: they
receive a restrictive reading and bind a variable at LF.
(28) a.
Ik gaan school.
I go school
I go to school.
b. Kom maar mijn huis.
come just my house
Come to my house.
c. Ja wij moet altijd moskee gaan een dag vijf keer moskee gaan.
yes we must always mosque go a day ve time mosque go
Yes, we must go to the mosque ve times every day.
As shown by the following target Dutch equivalents of (27a) and (28a), respectively, Dutch requires the presence of a prepositional element in those syntactic
constructs that express the abstract property of location.
(29) Hij woont ook in Kirslehir.
he lives also in Kirslehir
(30) Ik ga naar school.
I go to school
Perfect projections
In (27), the elements Kirslehir, Ankara and Istanbul indicate a static location, i.e.
place. In (28), the elements school, huis and moskee have a path interpretation. It is obvious that abstract meaning properties like place or path are not
directly related to these nouns. These nominal elements dont have an inherent
locative meaning, as is clear from sentences in which they full a non-locative
role, as in (31):
(31) a.
It is more likely that the locative properties place (static location) and path
(dynamic location) are associated with the category P(reposition). In the
(target) Dutch examples in (29)(30), this prepositional element expressing the
abstract meaning property space or path is phonetically realized; in the
L2-expressions in (27)(28) it is not. The non-visibility (i.e. phonetic absence)
of the prepositional element does not imply that the category P (and its
projection) is absent in expressions like Kirslehir and school. In fact, the preposition may very well be empty:
(32)
PP
Pspace
(33)
(cf. (27a))
NP
Kirslehir
PP
Ppath
(cf. (28a))
NP
school
59
60
Norbert Corver
Kitap masa-da.
book table-loc
The book is on the table.
b. Hasan Ankara-ya git-ti.
Hasan Ankara-dat go-past
Hasan went to Ankara.
c. Hasan Ankara-dan gel-di.
Hasan Ankara-abl come-past
Hasan came from Ankara.
(35)
Pspace
NP
masa-DA (-DA as alternative realization of P-feature)
Perfect projections
(38)
Pspace
NP
VAN-Tilburg
(VAN as a case-aYx)
In conclusion, both the prepositional structures in (32)(33) and the prepositional structure in (38) are LF-interpretable objects. The empty preposition
(and its projection) carries the interpretable property space or path. The
target imperfection is simply a surface phenomenon that relates to the phonetic
spell out of the prepositional position: the L2-learner initially leaves the
prepositional head empty and tries to realize the prepositional feature by means of
a case-marking (i.e. alternative feature realization). Given the lack of clear case
markings in Dutch, this alternative realization remains empty initially. At a certain
stage in the acquisition process, the prepositional feature gets alternatively
realized on the NP-complement by the semantically empty preposition van.
61
62
Norbert Corver
person and number; it is not the subject-DP that is dependent on the verb for
agreement. And an adjective is said to agree in number and gender with the
noun it modies; it is not the noun that is valued for certain phi-features under
agreement with the modifying adjective. In recent generative studies (cf. e.g.
Chomsky 1995), this asymmetry of the agreement relationship is captured in terms
of the notion interpretable (formal) feature. An interpretable (formal) feature is
a feature that has a semantic contribution at the LF-interface (i.e. it is interpretable
at LF). A non-interpretable (formal) feature has no interpretation at LF (or for that
matter: PF). Structural case for nouns and phi-features for categories that agree
with nouns are core examples of uninterpretable formal properties.
The asymmetry in the agreement relationship, as observed in traditional
grammatical studies, has been reinterpreted in terms of the notions [+interpretable] versus [interpretable]. It is the element carrying the [+interpretable]
feature that agrees with the element carrying the [interpretable] feature. This
agreement relationship involves feature matching and elimination of the
uninterpretable feature that is associated with the matching constituent. In
(39a), for example, it is the verb zagen that enters into an agreement relationship with the plural subject-noun phrase de mannen. Plurality (more than one)
and singularity (one) is an interpretable property of nouns. In a way, the
plurality feature on the verb zagen is redundant; it does not contribute any
semantics and, as such, can be characterized as uninterpretable. The plurality
marking on the verb is just a formal marker of the agreement relationship;
semantically, it does not contribute anything to the linguistic expression.
Therefore, in a certain intuitive sense, it is easier to imagine that a plural
interpretation gets associated with the ill-formed expression (39b) than with the
ill-formed expression in (39c). In (39b), plurality is morphologically specied
on the noun, i.e. the category that carries the number feature as a [+interpretable] property, and the verb is not morphologically specied for plurality. Even
though there is a morphological mis-match, there is a tendency to assign this
sentence a plural interpretation: that is, the interpretation the men-pl saw-pl
me is much more likely than the interpretation the man-sg saw-sg me. In
other words, it is the plurality marking on the noun that most strongly determines the semantic interpretation of this expression which does not satisfy
agreement at the level of morphological expression. Consider next (39c), where
we have the reverse situation: the subject noun phrase does not bear any
marking of plurality; it is the verb zagen which is plural morphologically. In
spite of the plural marking on the verb, it is intuitively more dicult to get a
plural reading of the noun. As a matter of fact, it is the singularity of the noun
Perfect projections
which seems to be dominant again for the interpretation. In short, given the fact
that in subject-verb agreement relations, it is the noun that determines plural or
singular interpretation, it is expected that morphological marking of singularity
or plurality is more likely to be realized on the noun than on the verb.
(39) a.
In these examples, the phi-feature number carries the value plural. This
number feature is correctly spelled out on the noun of the subject-noun phrase
that enters into an agreement relation with the verb. The verb carries what looks
like a singular inection. Arguably, this verbal form is unanalysed morphologically or, alternatively, the marking -t is underspecied for number.
What is important is that, even though the L2-expressions in (40) may be
regarded as imperfect from the perspective of the target language, they are
63
64
Norbert Corver
perfect from the perspective of the LF-interface: i.e. they are expressions that are
fully legible semantically.
Let me close o this section with another illustration of an LF-interpretable
structure in which the agreement properties are morphologically spelled out
asymmetrically. The relevant example comes from noun-phrase internal
agreement between a numeral and a noun. As noted among others in Emonds
(1985), plurality is a property of numerals: i.e. numbers above one are inherently marked for the property [+plural]. This formal property is interpretable,
since plurality plays a role in the semantic interpretation of a linguistic expression. As illustrated by the following L2-expressions, plurality is not always
marked on the agreeing noun:
(41) a.
In (41a), the noun minuut does not carry plural morphology. The target pattern
would be: twee minuten. Also in this case, then, the L2-learner chooses the
strategy of not morphologically expressing plurality if this semantic feature is
already specied in the projected structure. In other words, redundant marking
of the plurality feature on the noun is avoided.
Again, it is important to stress that the numeral+noun-patterns in (41) are
fully legible at the LF-interface. From a target language perspective, however,
these structures are imperfect: in Dutch, plurality is (redundantly) marked on
the noun, when it combines with a numeral that is inherently specied for
plurality (i.e. more than one). In this respect, Dutch diers, by the way, from
Turkish. As noted in Kornlt (1996: 225), there are syntactic contexts in
Turkish where, despite plural semantics of the noun phrase, the head noun
cannot be marked for plurality. When the noun is preceded by a numeral or
certain quantiers, the plural sux cannot occur:
Perfect projections
(42) a.
bes ocuk(*-lar)
ve child(*ren)
ve children
b. birok ocuk(*-lar)
many child(*ren)
many children
6. Conclusion
From the perspective of the target language, L2-expressions often seem highly
imperfect. At the surface, these L2-expressions (e.g. Dutch L2-products of
Turkish learners) seem to dier greatly from those produced by mother tongue
speakers. From a dierent perspective, though, there does not seem to be much
wrong with those L2-expressions: they are perfect expressions, in the sense that
they meet conditions imposed by other cognitive systems that the language
faculty interacts with (external requirements). That is, any L2 (interlanguage)
grammar provides (grammatical) information that is legible to the cognitive
systems with which it interacts. I have tried to illustrate the interface-legibility
of L2-expressions by means of four types of phenomena: (a) categorial labeling
of words, (b) quanticational expressions, (c) the expression of location in
prepositional structures, and (d) the morphological expression of certain
agreement patterns. As regards the categorial labeling of words, it was noted
that certain L2-words (e.g. an L2-verb like wegt (goes)) that do not exist in the
target-language, are perfect lexical constructs from a (lexicon-syntax) interface
perspective: it is a sound-meaning pair that through assignment of a categorial
value (i.e. V) becomes accessible to the computational system (merge, move,
morphological rules) of the interlanguage grammar. I further argued that
certain L2 quanticational expressions (e.g. alles mensen) that are imperfect
from a target language perspective are fully legitimate from the perspective of
LF-legibility: they are normal quanticational expressions in the sense of
65
<DEST "cor-n*">
66
Norbert Corver
Notes
* I thank the participants at the workshop for their comments on the talk. I would also like
to thank an anonymous reviewer for helpful comments and suggestions.
1. In a recent interview with Adriana Belletti and Luigi Rizzi, Chomsky states the following
(cf. Chomsky, Belletti and Rizzi 1999 (rev. 2000:17)): Every language meets minimalist
standards. Now, that means that not only the language faculty, but every state that it can attain
yields an innite number of interpretable expressions. That essentially amounts to saying that
there are no dead ends in language acquisition. He further states: The minimalist thesis
would say that all states have to satisfy the condition of innite legibility at the interface.
2. A reviewer raises the following question: What can a generative-minimalist theory account
for with respect to interlanguage expressions that other theories (e.g. a GB-based approach
using presumed UG-notions like government) cannot account for? Although such a
comparison of approaches may be useful in certain respects, it is not always easy to evaluate
the benets of one specic analysis of interlanguage data over another one, especially if the
analytic tools are dierent. Important, though, is the dierent perspective that the minimalist
approach towards language design provides: linguistic properties are not so much considered
from an intra-grammatical perspective, but rather from an interface perspective (lexicon-
Perfect projections
References
Bley-Vroman, R. 1990. The logical problem of foreign language learning. Linguistic Analysis
20: 349.
Chomsky, N. 1991. Some notes on economy of derivation and representation. In Principles
and parameters in comparative grammar, R. Freidin (ed.), 417454. Cambridge, MA:
MIT Press.
Chomsky, N. 1995. The minimalist program. Cambridge MA: MIT Press.
Chomsky, N. 2000a. New horizons in the study of language and mind. Cambridge, UK:
Cambridge University Press.
Chomsky, N. 2000b. Minimalist inquiries (MI). In Step by step: Essays in minimalist syntax
in honor of Howard Lasnik, Martin et al. (eds). Cambridge MA: MIT Press.
Chomsky, N., Belletti, A. and Rizzi, L. 1999 (rev. 2000). An interview on minimalism.
University of Siena.
Craats, I. van de 2000. Conservation in the acquisition of possessive constructions. Doctoral
Dissertation, Tilburg University.
67
</TARGET "cor">
68
Norbert Corver
Craats, I. van de, Corver, N. and Hout, R. van 2000. Conservation of grammatical knowledge: on the acquisition of possessive noun phrases by Turkish and Moroccan learners
of Dutch. Linguistics 38 (2): 221314.
Emonds, J. 1985. A unied theory of syntactic categories. Dordrecht: Foris.
Emonds, J. 2000. Lexicon and grammar: The English syntacticon. Berlin/New York: Mouton
de Gruyter.
Kornlt, J. 1996. Turkish. New York: Routledge.
Marantz, A. 1997. No escape from syntax: Dont try morphological analysis in the privacy of
your own lexicon. University of Pennsylvania Working Papers in Linguistics 4 (2): 201225.
Paardekooper, P. C. 1974. Beknopte ABN-syntaxis. Den Bosch: Malmberg.
Perdue, C. (ed.) 1993. Adult language acquisition: Cross-linguistic perspectives, vol. I: Field
methods. Cambridge, UK: Cambridge University Press.
Schachter, J. 1989. Testing a proposed universal. In Linguistic perspectives on second language
acquisition, S. Gass and J. Schachter (eds), 7388. Cambridge: Cambridge University Press.
Schenning, S. 1998. Learning to talk about space. The acquisition of Dutch as a second language
by Moroccan and Turkish adults. Doctoral Dissertation, Tilburg University.
Verkuyl, H. 1981. Numerals and quantiers in X-Bar syntax and their semantic interpretation. In Formal methods in the study of language, J. Groenendijk, T. Janssen and M.
Stokhof (eds), 567599, Amsterdam: Mathematic Centre.
White, L. 1988. Island eects in second language acquisition. In Linguistic theory in second
language acquisition, S. Flynn and W. ONeill (eds): 144172. Dordrecht: Reidel.
White, L. 2000. Second language acquisition: From initial state to nal state. In Second
language acquisition and linguistic theory, J. Archibald (ed.): 130155. Oxford: Blackwell.
Zwarts, J. 1992. X-Syntax X-Semantics. On the interpretation of functional and lexical
heads. Doctoral Dissertation. Research Institute for Language and Speech OTS.
Utrecht University.
<TARGET "cra" DOCINFO AUTHOR "Ineke van de Craats"TITLE "L1 features in the L2 output"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 4
1.
Introduction
In current generative syntax (e.g. Chomsky 1995), the role of the lexicon has
become more prominent in the generation of syntactic expressions. Under this
view, the lexicon is not limited to vocabulary, but also contains important
grammatical information. Lexical knowledge consists of grammatical properties
as dened by language-particular knowledge of functional categories (parameter settings), language-specic knowledge of lexical items and their features (the
vocabulary) and morphological knowledge. By means of lexical items selected
from the lexicon, the computational system of human language can build
phrases and sentences. A syntactic object a clause or a phrase is considered to be the structural projection of a series of linguistic properties associated
with a lexical item. Those formal features (e.g. singular, accusative, human,
+V, N) are stored in the vocabulary, together with the semantics of that
specic lexical item. Lexical items, like nouns and verbs, are base-generated as
a unit, including case morphology and inectional morphology like person,
number, tense, under lexical heads. Functional heads, on the other hand, do not
dominate inectional morphology, they dominate bundles of abstract features.
These features have to be eliminated or erased in the course of the derivation,
which is done by feature checking. This feature checking is a matching of the
features (e.g. case morphology is checked by its case assigner) and is done by
adjoining the inected N or V to the relevant functional head. So, morphology
which is associated with a verb or a noun has to be checked by the abstract
features dominated by a functional head (e.g. Agr or T for verbs, and Agr and
D for nouns). What features are dominated by a functional head and whether
these features are strong or weak is lexical knowledge which is necessary for the
generation of a syntactic object, but is not part of the vocabulary. So, in the
minimalist approach, a parameter is related to a feature of a functional head
70
that attracts an identical feature of a lexical item at some point in the derivation,
and so, is essentially linked to the lexicon.
If, in recent linguistic theorizing, the formal features of a lexical item and the
specication of functional heads, can be seen as the seeds for building a syntactic
structure they must play the same role in the acquisition of a new language,
assuming that we are dealing with a natural language. Applying generative theory
to second language acquisition is not new of course. Before the work of White
(1982, 1985), Flynn (1986) and many others, Adjmian (1976) was the rst to
adopt a Chomskyan approach to interlanguage development. He considered
grammatical interlanguage systems to be natural languages but dierent from
L1 grammars only in their permeability to aspects of the L1 system.
In this chapter, the current view of generative syntax will be applied to the
analysis of naturalistic second language data. Although L2 expressions are the
syntactic products of the interaction between the computational system and a
changing lexicon, the focus will be on how the grammatical knowledge of an L2
learner is encoded in the lexicon, not on derivation and syntactic representations. The question is more: what is the nature of this grammatical knowledge
at the L2-initial state and how does it change? Through examples produced by
L2 learners and by outlining the longitudinal development of some lexical items,
it will be shown how features of a lexical item may change in the course of the
acquisition process, giving rise to new syntactic structures. For a detailed discussion of the syntactic representations of nominal possessive constructions in which
van (Subsection 5.1) and the realisation of the personal pronoun (Subsection 5.2)
are involved, the reader is referred to Van de Craats, Corver and Van Hout
(2000), for representations of clausal possessive constructions in which heeft is
involved (Section 6) to Van de Craats, Corver and Van Hout (2002).
output cannot show all L1 properties because of a strongly limited L2 vocabulary. With the developing vocabulary, (more) L1 properties related to free and
bound functional morphemes become manifest gradually, as we will see for
genitive markers (Subsection 5.1) and copular forms (Subsection 6), which,
initially, are not found in learners data but appear gradually.
The following aspects of lexical knowledge may be conserved at the L2-initial state, when all learning starts:
71
72
They were followed over 2.5 years from the time they entered primary school.
At the time of the rst recording the childrens age ranged from 6;4 to 7;9 years.
phonological matrix
semantics
formal features
intrinsic
intrinsic
intrinsic
optional
optional
Turkish
Dutch
English
/bisiklet/
bicycle
/ets/
bicycle
/bike/
bicycle
[+N,V]
[human]
[3 person]
[singular]
[nominative]
[+N,V]
[human]
[3 person]
[singular]
[nominative]
[+N,V]
[human]
[3 person]
[singular]
[nominative]
phonological matrix
semantics
formal features
L1 item
(Turkish)
Interlanguage
L2 item
(Dutch)
/bisiklet/
bicycle
[+N,V]
[3 person]
[singular]
[nominative]
//
bicycle
[+N,V]
[3 person]
[singular]
[nominative]
/ets/
bicycle
[+N,V]
[3 person]
[singular]
[nominative]
73
74
matching operation without making any errors. For most content words, such
a perfect match is possible as far as formal features are involved. We do not
expect a dierent set of formal features in two languages because entities are
typically nouns ([+N,V]), actions are typically verbs ([N,+V]), and qualities
are typically adjectives, each of those with its own categorial values (N, V, A)
and the formal features typically related to those categories. As for the semanticconceptual aspects, dierences between L1 and L2 are to be expected, however.
The rst example of imperfect matching involves the domain of semantics.
A meaning representation of a lexical item may have dierent aspects. The
Turkish verb imek, for instance, diers minimally from the Dutch verb drinken
(to drink). The formal features are the same, but the basic meaning of the verb
/imek/ is to put something in something else. This general meaning has
several more specic meaning aspects, viz. to drink and to smoke. Consider
the L2 expression in (1).
(1) Als ik Marlboro drinken.
when I Marlboro drink
When I smoke a Marlboro.
(2) a.
(3) a.
Politie
zegt: Jij geld geven.
policeman says you money give
The policeman says: You must pay.
Mahmut, III-9
The examples in (3) make clear what Mahmut meant to say in (2). We can
assume that instead of betalen to pay he intended to say geld money.4 From
the perspective of the Conservation Hypothesis, the argument runs as follows.
A Turkish learner expects to nd the verb at the end of the sentence, as Turkish
is basically a language with an SOV sentence structure. Hence, the Turkish
learner in (2a) considers geven to be a verb. Since the L1 item demek (to pay)
has an internal argument, this L2 learner places this argument in the object
position that normally precedes the verb, e.g. in (2a). Hence, betalen must be
the argument, and a noun. In (2b) betalen might be meant as a verb, but (3b)
makes that unlikely, the more so because, in cycle II, Mahmut is not yet able to
produce possessive clauses in which hebben (to have) occurs.
As can be inferred from the sentences in (2) and (3), the categorial value of
betalen in (2) is that of a noun: [+N,V], and not that of a verb [N,+V]. Only
in (3), the phonological matrix /betalen/ is replaced by /geld/, as represented in
Table 3. In this table, the subcategorization frame of to pay has been integrated
in order to underline the verbal character of the L2 item.
Table 3.A learner variant of the lexical item /betalen/ compared to the relevant items
in source and target languages; deviances from the target language are in italics
L1 item
(Turkish)
phonological matrix
semantics
formal features
/para/
money
[+N,V]
[human]
[3 person]
[singular]
[accusative]
subcategorization frame
Learner
variant
L2 item
(Dutch)
/betalen/
money
[+N,V]
[human]
[3 person]
[singular]
[accusative]
/betalen/
to pay
[N,+V]
[DP2 DP1 ]
Mahmut, I-5
75
76
b. Ik meisje baby .
I girl baby
My girlfriend is expecting.
Vijf maanden baby terug.
ve months baby back
In ve months the baby will come.
Mahmut, I-6
In (4a), Mahmut tells about his wedding. Before the wedding, he used to go to
his girlfriends house, far away, but after the wedding his wife came to his
house. In this early stage of acquisition, Mahmut had two options for expressing
the possessive pronoun rst person singular: the target variant mijn (my) +
possessee and the learner variant ik (I) + possessee. He used them both in this
sentence. The particle terug expresses the action of coming. In Dutch, the
particle terug is the separable part of the compound verb terugkomen (to come
back/to return). In matrix clauses, the nite part of the verb appears in (more
precisely: is moved to) the second position in the sentence, while the particle
remains at the end. This is probably the cause of the misinterpretation by the
Turkish learner, who expects the nite verb in end position, where he nds
terug (instead of geliyor comes). In Dutch, as in English, it is even possible to
leave out the past participle gekomen in a perfect tense and to say, as the result
of the action terugkomen, ik ben terug (I am back), which may be interpreted
by this learner as: I have come.
In (4b) the directional element is not so evident. Mahmut took part in a
role playing task. He was asked to explain to a housing ocer why he needed a
house. The informant was given the information that his girlfriend was pregnant. The introducing sentence (ik meisje baby) cannot have another meaning
than that his girlfriend is expecting. In Dutch, the internal argument (a baby)
of the verb verwachten (to be expecting) must be expressed overtly. Note that we
are dealing here with two arguments that are strongly suggestive of the predicate
verwachten (to expect), so that it makes sense to assume a predicate with an
empty phonological matrix, viz., verwachten. The second sentence of (4b)
confused the housing ocer. He understood that the baby was already born
and that he or she would come back after a stay in Turkey or somewhere else.
The cause of this misunderstanding is that Mahmut maps the L2 phonological matrix /terug/onto a L1 feature bundle linked to the verb /geliyor/. In that
way, the particle terug can act as a verb and has the same argument structure as
the verb to come.5 This is represented in Table 4.
The examples above have shown that beginning L2 learners may have
diculty in discerning the grammatical properties of L2 content words, which
Table 4.Lexical items of the concept comes; deviance from the target language is in italics
L1 item
(Turkish)
phonological matrix
semantics
L1 formal features
/geliyor/
comes
[N,+V]
[3 person]
[singular]
[present tense]
subcategorization frame [DP]
Learner
variant
L2 item
(Dutch)
/terug/
back
[N,+V]
[3 person]
[singular]
[present tense]
[DP]
/komt/
comes
[N,+V]
[3 person]
[singular]
[present tense]
[DP]
are lexical elements with a relatively high salience in the environmental input.
For the understanding of functional elements this must be still harder.
77
78
The possessor noun phrase (DP) carries a genitive case feature (-nin or -nun,
choice determined by vowel harmony) and the possessed noun a genitive case
agreement feature (-s), as represented in Table 5. This possessive relationship
is characterized by agreement, overtly realized both on possessor and possessee.
In line with Chomsky (1986, 1995), it is assumed that the genitive case agreement feature of the possessee must check o the genitive case feature associated
with the possessor DP. The required structural conguration is AgrP where the
possessor DP and possessee N enter in SpecHead conguration because of the
strength properties of the Agr head (cf. Van de Craats et al. 2000 for details).
Table 5.Formal feature complex of two lexical items and the functional head Agr in the
Turkish possessive construction
phonological matrix
semantics
formal features
Possessor
Possessee
Agr
/Ayse-nin/
of Ayse
[+N,V]
[3 person]
[singular]
[genitive]
/araba-s/
car
[+N,V]
[3 person]
[singular]
[+genitive case assignment]
//
[+N,V,+D]
strong
[+N,V]
weak
Unlike Turkish, Dutch has several ways for expressing a possessive relationship. Two of them look, supercially, like Turkish (6a, 6b), the other (analytic)
construction (7) does not.
(6) a.
Ayse s auto
Ayse s car
Ayses car
b. Ayse dr auto
Ayse her car
Ayses car
(Saxon genitive)
(Doubling possessive)
(Analytic construction)
79
80
Table 6.Formal feature complex of two lexical items and the functional head Agr for
three possessive constructions in Dutch
Possessor
phonol. matrix
semantics
formal features
/Ayse/
Ayse
[+N,V]
[3 person]
[singular]
[genitive]
Possessee
/auto/
car
[+N,V]
[3 person]
[singular]
[+gen. case
assignment]
Agr
Saxon
genitive
Doubling
possessive
Analytic
construct.
/-s/
[+N,V,+D]
strong
/clitic/
Agr not
(e.g. zn)
projected
[+N,V,+D]
strong
[+N,V]
weak
[+N,V]
weak
Table 7.Lexical items of the concept possessor of ; deviances from the target language
are in italics
L1 item
(Turkish)
Learner
variant 1
Learner
variant 2
L2 item
(Dutch)
/-(n)In/
possessor
[ax gen]
//
possessor
[ax gen]
/van/
possessor
[ax gen]
[N]
[N]
/van/
possessor
preposition
[N,V]
[DP]
phonological matrix
semantics
categorial value
L2 learners who map the phonological matrix of the L2, van (of), onto the
grammatical properties of their L1, produce such nominal phrases as in (9)
and (10).
(9) pronominal possessor
a. [die van] auto
[that of car
his car
b. [onze van] broer
[our of brother
our brother
(10) full noun possessor
a. [examen van] tolk
[exam of interpreter
the interpreter at the exam
Ergn, III-4
Ergn, III-5
Osman, III-2
In the examples in (9) and (10), the preposition van (of ) should not be
considered an element of an analytic construction, as in (7), but the genitive
case marker of the preceding possessor noun phrase as in Turkish (cf. example
(5). Notice that both child learners and adult learners produce these constructions based on the L1 bundle of formal features as presented in Table 7.
It is a complicating factor that L2 learners do not show this genitive case
marker from the earliest stage of L2 acquisition. L2 learners, in general, produce
only a few functional elements in the beginning of the acquisition process and
it is questionable whether they perceive functional elements in the L2 input at
all. Nevertheless, they build syntactic constructions like the L2 expressions in
(11), which are not simple two word utterances but can be extensive phrases.
tante dochter7
Osman, I-5
aunt daughter
my aunts daughter (= cousin)
b. tante zoon auto
Mahmut, I-5
aunt son car
my aunts sons car
c. mijn vrouw oma
andere man dochter
Mahmut, II-8
my wife grandmother other man daughter
my wifes grandmothers second husbands daughter
(11) a.
81
82
van (of), and by the fact that only Turkish learners exhibit possessive constructions overtly marked for genitive case. Moroccan learners do not. One may
object that the productions exemplied in (11) proceed directly from UG. But
why would Turkish learners have access to UG and Moroccan learners not?
In addition to possessive noun phrases in which the possessor precedes van,
as in (5) and (6), L2 learners with a Turkish language background produce
possessive noun phrases where van precedes the possessor, as in (12) and (13).
(12) pronominal possessor
a. [van hem] moeder
Ergn, II-9
[of him mother
his mother
(target: zijn moeder)
b. [van ons] die fabriek
Ergn, III-4
[of us that factory
our factory
(target: onze fabriek)
c. [van ons] buurman heeft goed gemaakt child learner (number T41)
[of our neighbour has good made
our neighbour has repaired it
(target: onze buurman)
(13) full noun possessor
a. [van Ergn] auto
[of Ergn auto
Ergns car
(target: Ergn zn/s auto)
b. [van schoenen] die touwtje8
[of shoes
that rope
the shoelaces
(target: de veters van de schoenen; de schoenveters)
Abdullah, II-7
Osman III-6
The examples in (12) and (13) exemplify a new stage of lexical development,
viz., the stage in which van is no longer a genitive case sux, but an adposition
preceding the possessor, like van in analytic constructions as in (7). A second
alteration vis-a-vis the former stage is that van precedes a full DP instead of
being a sux on a noun.
In addition to these examples, there are some rare learner variants of the
Saxon genitive and the doubling possessive constructions in which L1 and L2
elements are mixed. These are given in (14). In (14a), we see a target doubling
Osman, III-2
Likewise, the complete development of any lexical item can be represented from
the L2-initial state, viz. the nal state of the L1 lexicon, to the state where the
lexical item of the target language is attained. The formal features required for
building target syntactic objects, viz. those dominated by the functional head
Agr and those of possessor and possessee, were already presented in Table 6.
Table 8.Developmental stages of the L2 lexical item /van/ in constructions with full DP
possessors; changes with regard to the previous stage are in italics
L1 agreement pattern
phonological matrix
semantics
categorial value
subcategorization frame
example
L2 construct pattern
phonological matrix
semantics
categorial value
subcategorization frame
example
L2 analytic pattern
phonological matrix
semantics
categorial value
subcategorization frame
example
Stage 1
Stage 2
Stage 3
Target
//
possessor
[ax gen]
[N]
Ayse- auto
/van/
possessor
[ax gen]
[N]
[Ayse-van]
auto
/van/
d.n.a.
possessor
[adposition]
[DP possor]
[van Ayse] auto
/van/
possessor
[ax gen]
[N]
Ayse-van dr
auto
/van/
d.n.a.
possessor
[adposition]
[DP possor]
van Ayses auto Ayse dr auto
Ayses auto
/van/
possessor
[preposition]
[N,V]
[DP possee]
auto [van Ayse]
/van/
possessor
[preposition]
[N,V]
[DP possee]
auto [van Ayse]
/van/
possessor
[preposition]
[N,V]
[DP possee]
auto [van Ayse]
83
84
Table 9.Distribution of variants of the lexical possessive item /van/ produced by four
Turkish learners in constructions with full DP possessors
Mahmut
Cycles
Ergn
Abdullah
Osman
II
III
II
III
II
III
II
III
L1 agreement pattern
Possessor
63
Possessor ax
Possor -adposition
55
42
15
11
2
7
2
3
7
4
1
9
12
2
3
1
1
2
1
L2 construct pattern
Possor +van + clitic
Possor + clitic L2
2
L2 analytic construct.
Van-preposition
2
8
23
12
2
14
25
In this way, lexical item learning in all aspects can be made visible. The vocabulary-internal development of the L2 lexical item /van/ can be represented as in
Table 8 for constructions with full DP possessors. Three developmental stages
are distinguished, each of them is characterized by one or more changes
compared to the previous stage. The order of the stages is determined by the
rst emergence of the changed value. In stage 1, the possessive L2 expressions
are based on an L1 pattern and the genitive case is not overtly realized. Stage 2 is
characterized by the new (analytic) L2 pattern in which the genitive case is realized
by van, both as a preposition and an ax. Likely, the new construction has impact
on the agreement pattern from the L1 because learners interpret van correctly
as the morphological realization of genitive case, but they mis-assess its categorial value. In stage 3, van is used as a preposition and adposition as well.
In Table 9, the distribution of these lexical stages over the four Turkish
informants is given. The time course represents three successive cycles, each
representing ten months of data collection. The informants are arranged from
the slowest learner on the left of the table to the faster learners on the right.
From the combination of the data in the Tables 8 and 9, it can be inferred
that Mahmut just entered stage 2, that Ergn attained this stage somewhat
earlier, and that Abdullah and Osman have reached stage 3 at the beginning of
the data collection. Osman is the only informant who produced a few doubling
possessive constructions. Table 9, also shows a considerable overlap between the
stages. In cycle III, Osman and Abdullah have abandoned stage 1, Mahmut and
Ergn have not yet done so.
(pro) geli-yorum
come-pres.1sg
b. ben
geli-yorum
I (stressed) come-pres.1sg
I come
(non-focused)
(focused)non-
The same holds for the pronoun in a possessive noun phrase like his car. The
presence of a lexical pronoun yields a reading in which the possessor is emphasized, e.g., in (16b), his car as opposed to your car. In neutral, non-focused
readings, the pronoun is not lexically expressed but contains number and
person features, as in (16a).
(16) a.
(pro) araba-s
car-3sg
his car
b. o-nun
araba-s
he/that-gen car-3sg
his car
(non-focused)
(focused)non-
So, in Turkish two dierent phonological matrices, /ben/ and //, can be linked
to the same formal feature bundle, as illustrated in the columns 2 and 3 of
Table 10. Notice here that empty phonological matrices are not restricted to
learner variants as in (11) and in the Tables 7 and 8. Dutch does not have the
possibility of pro-drop and therefore can only use a lexically lled pronominal
ik (I). According to recent analyses (see, among others, Corver and Deltto
1999) Dutch pronouns should be considered transitive determiners (D). It is
assumed that their complement is a pro which stands for the phrasal projection
NP and not just for the head noun: [DP D [NP pro]]. The name prodeterminer
seems to be more appropriate for these determiners than pronoun. Turkish
pronouns, however, seem to be pronominals in the true sense of this word; i.e.
they seem to be of the lexical category type N rather than of the functional
85
86
phonol. matrix
semantics
formal features
/ben/
I focused
[+N,V]
[1 person]
[singular]
[nominative]
L2 item
(Dutch)
//
I
[+N,V]
[1 person]
[singular]
[nominative]
/ik/ stressed
I focused
[+D]
[1 person]
[singular]
[nominative]
categorial type D (see Kornlt 1997: 300). Support for this comes, for example,
from their behaviour with respect to various morphological rules; like common
nouns, pronominals in Turkish function as stems to which case and plural
morphology can be attached. The two types of pronominals in Turkish and
Dutch are compared in Table 10.
The dierences between the two pronominal systems consist in (i) the
categorial value of the pronoun, and (ii) the use of overt pronouns as a means
of focalising. Given the Conservation Hypothesis, Turkish L2 learners match a
phonological matrix of the L2 with an L1 feature bundle. The question arises
which combination they will make. As can be inferred from the L2 data,
Turkish learners link the non-focused item (an empty pro), which is more
frequently used in Turkish than the focused one and can be considered the
default form, to the Dutch full pronoun. That is not surprising because, in
initial stages, adult learners turn out to perceive full pronouns before clitic
pronominals (cf. Broeder 1991), although the clitic pronouns occur more
frequently in the spoken L2 input. However, the question remains for L2
learners how to express the focused pronoun in their L2 Dutch. Assuming that
Dutch pronominals are of the categorial type N as well, their projections may
contain a slot for demonstrative determiners. So these L2 learners may attach a
demonstrative die or deze (that) to the pronoun in emphatic contexts, as in
(17a), in which Mahmut explains why his daughter must learn Turkish, and in
(17b), in which Ergn talks about his brother.
(17) a.
Mahmut, I-4
L2 item
(Dutch)
Learner
variant 2
L2 item
(Dutch)
/ik/ (stressed)
I (focused)
[+D]
[1 person]
[singular]
[nominative]
/ik/
I
[+N,V]
[1 person]
[singular]
[nominative]
87
88
(18) a.
Some languages also show a non-incorporated variant like French (ce livre est
moi this book is to me). Moroccan Arabic, however, does not have a variant in
which the locative preposition is incorporated in the be-copula, viz. no equivalent of the verb to have. Moroccan Arabic uses only a locative sentence with the
PP at+clitic (i.e. Eend-u at+him), both for a present tense sentence (19a) and for
a past tense sentence (19b) (cf. Harrell 1970).
Abder, Eend-u dar
kbira.
(present tense)
Abder at-him house-f big-f
Abder has a big house
[CP Abder, [CP [AgrP [Agr [Eend-u]i [SC dar kbira [PP ti]]]]]]
b. Abder, kanet
Eend-u dar
kbira.
(past tense)
Abder cop.past.3sg.f at-him house-f big-f
Abder had a big house.
[CP Abder, [CP [AgrP [Agr kanetk [TP [Eend-u]i [T tk [SC dar kbira
ti]]]]]]]
(19) a.
Notice that, in (19a), a copular form is lacking in the present tense and the
preposition Eend-u occupies the position of the verb (in Agr), whereas in (19b),
the position of the verb in Agr head is taken by the past tense copula kanet and
Eend-u is in SpecTP. This may have interesting consequences for L2 learners
who are in search of equivalents of their L1 grammar in the environmental
input of the L2. Let us make rst a comparison of the concept has in the
source and the target languages (see Table 12).
Just as was the case for the genitive case marker in Section 5.1, L2 learners
start with an empty phonological matrix linked to the L1 feature bundle, as can
be seen in the following L2 expressions produced by adult Moroccan learners.
In (20a), Fatima is asked about her work in Morocco: she had a small shop with
some knitting machines. In (20b), Mohamed describes his situation in an
interview with a housing ocial.
Table 12.Lexical items of the concept have compared for source and target languages
L1 item
(Moroccan Arabic)
L2 item
(Dutch)
/Eend- /
with/at
[N,V]
/heeft/
has
[Agr +T +P]
3sg +PRES +LOC
[DP]
phonological matrix
semantics
categorial value
(20) a.
pronominal possessor
Ik klein winkel. (= ik klein winkel)
Fatima, I-3
I small shop
I had a small shop.
b. nominal possessor
Mijn vrouw ook klein huis. (= mijn vrouw ook klein huis)
my wife also small house
My wife has also a small house.
Mohamed, I-6
Slow learners often show a stage in which they produce a preposition between
the possessor and the possessee, e.g. in (21). The preposition appears in the
same position where it would appear in the L1, between the strong pronoun
(not the clitic) or the full noun phrase in the dislocated position at the beginning of the sentence (cf. the position of Abder in (19a, b) and second noun
phrase. In (21a), Fatima talks again about her work and, in (21b), about her
relatives on a photograph: three of them are half sisters and brothers.
(21) a.
pronominal possessor
Ik met klein winkel.
I with small shop
I had a small shop.
[CP ik [CP [AgrP metj [SC klein winkel [PP tj]]]]]
b. nominal possessor
Fatiha Mustafa Khiliye met andere moeder.
Fatiha Mustafa Khiliye with other mother
Fatiha, Mustafa and Khiliye have another mother.
Fatima, I-3
Fatima, II-4
89
90
SpecAgrP, or (22c) in which the locative preposition is overtly spelled out and
incorporated in heeft as well.
Met kind een jaar.11
with child one year
The child is one year old.
b. Bij hem kief.
at him hashish
He has hashish.
c. Met Soumiya heeft veel pijn.
with Soumiya has much pain
Soumiya suers very much.
(22) a.
Fatima, II-7
HassanK, III-2
HassanM, III-5
Except for Fatima, such locative constructions instead of the verb have are rare
in the data produced by the Moroccan informants. When their have-constructions are considered more closely, however, we become more sceptical about
the character of the have-forms. To start with Fatima, she uses the form heeft
(3sg) not only for all person roles, but she alternatively uses met and heeft with
the same meaning for more than 13 months. Another informant produces
instances of heeft in which it can be a synonym for met (23a). The sentence in
(23b) is even a direct copy of Moroccan Arabic in which heeft is followed by a
clitic as if it were a locative preposition.
Die heeft geel haar mag binnen.
that has yellow hair may inside
The boy with blond hair is allowed to go inside.
b. Die meisje, heef-ze een oom.
that girl
has-she an uncle
That girl has an uncle.
(23) a.
HassanK, II-2
HassanK, II-2
The situation becomes crucial when this informant wants to express a possessive clause in the past tense. Recall from (19b) that, in Moroccan Arabic, the
past tense of a have-clause is expressed by a be-copula in the past tense
followed by the locative preposition and see what HassanK produced in (24).
Die was
heeft (-pro)
30 jaar.
HassanK, II-3
he was-cop.past.3sg has [=Eend+3sg] 30 years
He was 30 years.
b. Die meisje was
nooit heeft (-pro)
verkering. II-4
that girl was-cop.past.3sg never has [=Eend+3sg] relationship
That girl was never in a relationship.
(24) a.
c.
Dan was
heeft (-pro)
een huis.
then was-cop.past.3sg has [=Eend+3sg] a house
Then he had a house.
HassanK, II-9
From the examples above we infer that HassanK and Fatima base their syntactic
have-structures on the feature bundle of the Moroccan Arabic preposition
Eend (at, with). In this developmental stage, they still match the L1 feature
bundle with an L2 phonological matrix. Heeft is in fact a prepositional heeft.
The other two informants do not produce past tense forms of the verb hebben;
that is already sucient reason to be sceptical about the real identity of their
have-forms. The question can be raised how and when we can be sure that
heeft is a real copula and no longer a prepositional form. It is not easy to decide
when we are dealing with a copular verb, but it can safely be claimed that it is
no longer a preposition when past tense forms of hebben are expressed in one
verbal form (had had). HassanK succeeds in producing several past tense
forms and Mohamed does only once (see Table 14). A second indication is that
the verb hebben has extended from a possessive copula to an auxiliary of the
perfect tense, e.g. hij heeft gezien (he has seen). In this light, it is relevant to
signal that all Moroccan learners produce have-copulas six months or more
before they start using the auxiliary have (see Van de Craats 2000).
Finally, the development of the lexical item /heeft/ is represented in Table 13, from the L2-initial state (i.e. the empty phonological matrix with the L1
feature bundle) via various developmental stages to the state in which the item
has attained the complete feature constellation it has in the target language.
Tables 13 focuses on the properties of the lexical item heeft, which is the spell
out of the features [agreement], [tense] and [N,V]. This already suggests a
process of adjunction of several functional heads. In Table 13, we abstract away
from the formal features dominated by T and Agr triggering the movement of
P(P) (see Van de Craats et al. 2002: 156 for the derivation).
Table 14 shows how the stages are distributed over the data of the four
Moroccan learners. Slow learners exhibit more tokens of the developmental
stages than fast learners. The order of the stages is determined again by the rst
emergence of a changed value. For the slow learners, a considerable overlap of
the stages can be observed. The overlap is smaller for learners like Mohamed
and HassanM. At the end of the data collection, stage 3 is attained by all four
learners; only two of them can form past tense forms according to target
language standards, however.
91
92
Table 13.Stages in the development of the L2 lexical item /heeft/; dierences with
regard to the previous stage are in italics
phon.matrix
semantics
categorial value
subcategorization frame
example
Stage 1
Stage 2
Stage 3
Target state
//
with, at
[N,V]
[DP clitic]
Abder boek
/met, bij/
with, at
[N,V]
[DP clitic]
Abder met
boek
/heeft/
with, at
[N,V]
[DP clitic]
Abder was
heeft boek
/heeft/
has
[+Agr,+T,+P]
[DP]
Abder heeft
boek
Fatima
I
Mohamed
II
III
II
L1 construction
Stage 1
48
Stage 2prep.
8
Stage 3prep.
10
Stage 3heeft
62
2
26
21
76
17
1
44
Target state
Hebben present
Hebben past
13
HassanK
III
II
10 1
1
1
77
138 150
1
HassanM
III
II
III
7 4
1 1
173 1
7
52
1
1
1
82
2
15
143 137
7. Conclusions
From the examples discussed above, we can infer that, in general, L2 learners
are more aware of the fact that words dier from language to language, i.e., that
phonological matrices dier, than that they are aware of the fact that other
lexical properties dier. Adult L2 learners are inclined to map or to attach a new
phonological matrix from the L2 to a bundle of semantic and formal features
from the L1. Only if those features bundles dier, can we get an insight in how
L2 learners proceed in their learning of lexical items: they start by assuming an
L1 feature constellation and they gradually change the features of the bundle
one by one, as we saw in Tables 8 and 13 for the acquisition of van and heeft. In
this way, the L2 learners output becomes more and more target-like; not only
the surface form (= the phonological matrix) alters, but also the features.
In this light, it can be explained why functional elements are more dicult
to acquire than lexical elements (i.e. content words). It is not only the case that
they are less salient in the environmental input, but they also dier more in the
structure of their feature bundle. Therefore, it takes more time to discover each
of the composing features. The formal features, however, are of crucial importance for the development of the L2 syntax: they are the input for the computational system and, hence, decisive for the result of derivation: the syntactic
objects. They are the interface itself, so to say.12 Until a Moroccan learner of
Dutch, for instance, has not discovered each of the formal features of heeft
(has), he cannot attain the formation of the verb with subject-verb agreement,
he cannot form the past tense of that verb, nor the perfect aspect by means of
the auxiliary hebben. This shows the direct link between lexicon and syntax.
Why is there progress? The syntactic objects (i.e. the product of the computational system and the selection of lexical elements, including formal features)
in interlanguages seem to provide L2 learners with a reason to change their L2
output, because it is not suciently understood by native speakers, incomplete
(recall, for instance, the empty phonological matrices in the output of beginners
in example (11)), etc. Therefore, L2 learners are forced to constantly reanalyse
their L2 output and to change the underlying formal features and/or the
parametric values. This process of restructuring seems to occur in a continuous
interplay between syntax and formal features as the seeds of syntax and as a part
of the lexicon as well.
Notes
1. In Chomsky (1994) it is argued that only the idiosyncratic formal features are part of the
lexical item. Optional formal features are added to the lexical item when it is selected from
the lexicon. Here, we will abstract away from that distinction.
2. Except for the case in which L1 and L2 are so much similar that an L2 learner decides to
use an L1 lexical item (with the L1 phonological matrix) directly in the L2 output.
3. It is certainly not the case that learners keep on using all meaning aspects of a lexical item
in their L2: the more transparent and concrete a specic semantic aspect is, the earlier it is
used in L2 expressions (cf. the 17 dierent aspects of breken (to break) the acceptability of
which Kellerman (1979) asked his Dutch subjects to judge).
4. A reviewer proposed that betalen geven would be a serial verb construction, or geven to
give a light verb; geven can function as a light verb in Dutch. Although this account is not
quite impossible, the reviewer is arguing too much from a target language perspective.
Moreover, there is no light verb geven in (2b), and the example in (3b) makes clear that it is
not the verb betalen that is meant.
93
94
5. A reviewer advanced that Adjmian (1983) already found that L2 learners tend to transfer
lexical patterns from their L1 to their L2, and assume that verbs take the same kinds of
subject and object in their L2 as they do in their L1. Mapping an L2 phonological matrix
onto L1 lexical properties goes even further.
6. Following insights from Longobardi (1996), we take the position that the Saxon genitive
and the doubling construction are in essence hidden construct noun phrases, because they
share certain properties with the Construct-State construction (as in Arabic and Hebrew).
The Saxon genitive and the possessive doubling have the same distribution as a denite
article, viz. they block the presence of a denite article in D0.
7. The English translation of these examples suggests that the word order is not uncommon,
but possessors DPs in the Saxon genitive are restricted to proper names, and to human
beings in the doubling possessive construction. Moreover, recursion is odd in those
constructions. A native speaker of Dutch would only use the analytic construction for
recursive patterns due to processing limitations.
(i) de dochter van mijn tante
(ii) de auto van de zoon van mijn tante
(iii) de dochter van de tweede man van de oma van mijn vrouw
8. Note that in Dutch, the order Possessor-Possessee is restricted to human possessors; the
Saxon Genitive (e.g. Jans ets Johns bike) is even restricted to proper names and nouns
equivalent to proper names, e.g. tante aunt. This points all the more to the L1 grammar in
example (11b).
9. See also Van de Craats (2000) and Van de Craats, Corver and Van Hout (2002) for details
on this construction in Dutch and Moroccan Arabic.
10. In this example, a subordinated clause is given rst because it shows the basic position of
the verb in Dutch (SOV language).
11. In Moroccan Arabic, ones age is also expressed by a have-construction.
12. Although the lexicon plays an important role in this view on adult L2 acquisition, it is not
identical to the Lexical Learning Hypothesis (e.g. Clahsen, Eisenbeiss and Penke 1996) nor
to other recent views on L1 acquisition (e.g. Radford 2000, Roeper 1996) in which the
acquisition of (L1) syntax is seen as gradual structure building.
References
Adjmian, C. 1976. On the nature of interlanguage systems. Language Learning 16: 297320.
Adjmian, C. 1983. The transferability of lexical properties. In Language transfer in
language learning. S. Gass & L. Selinker (eds), 250268. Rowley, MA: Newbury House.
Benveniste, E. 1966. Etre et avoir dans leur fonctions linguistiques. Bulletin de la Socit
de la linguistique de Paris 55 (1): 113134. Reprinted in Problmes de linguistique
gnrale. Paris: Gallimard.
Broeder. P. 1991. Talking about people. A multiple case study on adult second language
acquisition. Amsterdam/Lisse: Swets & Zeitlinger.
Chomsky, N. 1986. Barriers. Cambridge, Massachusetts: MIT Press.
</TARGET "cra">
Chomsky, N. 1994. Bare phrase structure. In MIT occasional papers in linguistics. Cambridge, Massachusetts: MIT Press.
Chomsky, N. 1995. The minimalist program. Cambridge, Massachusetts: MIT Press.
Clahsen, H., Eissenbeiss, S. and Penke, M. 1996. Lexical learning in early syntactic development. In Generative perspectives on language acquisition, H. Clahsen (ed.), 129159.
Amsterdam/Philadelphia: John Benjamins.
Corver, N. 1990. The syntax of left branch extractions. Doctoral dissertation. Tilburg University.
Corver, N. and Deltto, D. 1999. On the nature of pronoun movement. In Clitics in the
Language of Europe, H. van Riemsdijk (ed.), 799861. Berlin: Mouton & de Gruyter.
Flynn, S. 1986. A parameter-setting model of second language acquisition. Reidel, Dordrecht.
Freeze, R. 1992. Existentials and other locatives. Language 68: 553595.
Harrell, R. 1970. A short reference grammar of Moroccan Arabic. Washingon, D. C.: Georgetown University Press.
Kellerman, E. 1979. Transfer and non-transfer: Where we are now. Studies in Second
Language Acquisition 2: 3757.
Kornlt, J. 1997. Turkish. London, New York: Routledge.
Longobardi, G. 1996. The syntax of N raising: A minimalist theory. Manuscript, Utrecht
University.
Miller, P. 1991. Clitics and constituents in phrase structure grammar. Doctoral dissertation,
University of Utrecht.
Moro, A. 1997. Predicative noun phrases and the theory of clause structure. Cambridge:
Cambridge University Press.
Perdue, C. 1993. Adult language acquisition: Cross-linguistic perspectives. Volumes I and II.
Cambridge: Cambridge University Press.
Radford, A. 2000. Children in search of perfection: towards a minimalist model of acquisition. In Essex research reports in linguistics, 34.
Roeper, T. 1996. The role of merger theory and formal features in acquisition. In Generative perspectives on language acquisition, H. Clahsen (ed.), 415449. Amsterdam/
Philadelphia: John Benjamins.
Schwartz, B. D. and Sprouse, R. 1996. L2 Cognitive states and the full transfer/full access
model. Second Language Research 12: 4072.
Van de Craats, I. 2000. Conservation in the acquisition of possessive constructions. A study of
second language acquisition by Turkish and Moroccan learners of Dutch. Doctoral
dissertation, Tilburg University.
Van de Craats, I., Corver, N. and Van Hout, R. 2000. Conservation of grammatical
knowledge: on the acquisition of possessive noun phrases by Turkish and Moroccan
learners of Dutch. Linguistics 38 (2): 221314.
Van de Craats, I., Corver, N. and Van Hout, R. 2002. The acquisition of possessive haveclauses by Turkish and Moroccan learners of Dutch. Bilingualism: Language and
Cognition 5 (2): 147174.
Vermeer, A. 1986. Tempo en struktuur van tweede-taalverwerving bij Turkse en Marokkaanse
kinderen. Doctoral dissertation, Tilburg University.
White, L. 1982. Grammatical theory and language acquisition. Dordrecht: Foris.
White, L. 1985. The pro-drop parameter in second language acquisition. Language
Learning 35: 4762.
95
<LINK "duf-n*">
<TARGET "duf" DOCINFO AUTHOR "Nigel Dueld"TITLE "Measures of competent gradience"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 5
1.
98
Nigel Dueld
While one might take issue with the reference to experts intuitive judgments,
the rest of this paragraph is unexceptionable. Allen and Seidenberg further
point out that misgivings about the basis of grammaticality judgments are to be
found within the generativist camp. The authors cite a telling paragraph in
Schtze (1996: 20):
It is conceivable that competence in this sense of a statically represented
knowledge does not exist. It could be that a given string is generated or its
status computed when necessary, and that the demands of the particular
situation determine how the computation is carried out, e.g., by some sort of
comparison to prototypical sentence structure stored in memory. Since such
a scenario would demand a major re-thinking of the goals of the eld of
linguistics, I will not deal with it further.
Allen and Seidenberg claim that this sidesteps an important issue. Their
response is to try to derive the eects of grammaticality from an input-driven
system lacking any autonomous syntactic representations: that is, without any
reference to grammatical competence as theoretical linguists would construe it.
For various reasons, including both the results of their own simulation and
the results of other experiments to be discussed below, I believe that this
alternative approach to grammaticality is mistaken, and that something like the
competenceperformance distinction is empirically more correct. At the same
time, their work demonstrates the urgent need for a clearer articulation of the
nature of the distinction, and for a more explicit statement of the content of
grammaticality judgments. While I will not be able to oer that statement here,
I hope to sketch out the direction in which it might be found.
A second reason for reconsidering the competenceperformance distinction
is that I agree with those who claim that the line between competence and
performance is presently drawn in the wrong place: that the current idealization
to competence presents too narrow a view of scope of competence. This is a
view held not only by connectionists, but also by some generativists, most
notably Culicover (1998, 2000), though cf. Fodor (2001).
The specic notion that I would like to include within a revised version of
competence is that of syntactic gradience. As the title of this article implies, I
will argue that part of competence consists, not in correctly making categorical
judgments about grammaticality, but in correctly making gradient ones. To cite
one brief example, to be a competent knower of English is to know, not merely
that the sentences in (1) and (2) below are all less than perfect, but also that
they are rather precisely ordered in acceptability, each example being slightly
less acceptable than the one that immediately precedes it. These examples, cited
in Kluender (1992), are originally due to Chung and McCloskey (1983) (where
the symbol > indicates more acceptable than):
(1) a.
99
(2) a.
This is a paper that you really need to nd someone that you can
intimidate with. >
b. Which paper do you really need to nd someone that you can
intimidate with. >
c. How many papers do you really need to nd someone that you can
intimidate with. >
d. What do you really need to nd someone that you can intimidate with.
101
Surface
competence
Underlying
competence
Figure 1.Full convergence (SC and UC generate the same set of grammatical sentences,
NS and NNS converge on this set).
However, several legitimate alternatives exist. These are cases where either
the performance judgments on the data presented are a reex of only one type
of competence (UC or SC), or the judgments of two subject groups (L1 versus
L2) reect dierent competences. These logical alternatives are schematised in
Figures 25 below.
Common set of judgments
(shared by NSs & NNs)
Surface
competence
Underlying
competence
Surface
competence
Underlying
competence
Native
Speaker
Judgments
Non-Native
Speaker
Judgments
Surface
competence
Underlying
competence
Surface
competence
Native
Speaker
Judgments
Underlying
competence
Notice that all of these scenarios assume convergence between NS and NNS
competences; that is to say, I am ignoring those cases where L2 learners have
internalised a surface or underlying competence that is dierent from that of
native-speakers; cf. Sorace (1993, 1996). What this model is intended to account
for are cases in which the L2 learner has attained the target grammar, but where
his/her judgment patterns nevertheless diverge systematically from those of the
native-speaker.
Clearly, hypothesizing two dierent types of competence complicates any
theory of the relationship between competence and performance. Parsimony
requires that this complication be shown to be empirically necessary. The
purpose of the rest of this article, therefore, is to provide such motivation. The
organization is as follows. The following section provides some experimental
evidence motivating two types of competence. Next, I consider two types of
gradient eect, both of which I will argue are properties of surface, rather than
underlying, competence. Then, I discuss two instances of principled mismatch
between NS and NNS judgments, motivating both Figures 4 and 5 above.
Finally, I examine a special case of Figure 2 above, where NS and NNS apparently converge on the same SC, but for dierent reasons.
One piece of empirical evidence supporting this structural contrast comes from
the alleged (in)ability of these intransitive verbs to form transitive counterparts:
it has been claimed that ECSs, but not ICSs, permit transitive alternants. So, for
example, there is claimed to be a categorical contrast in the acceptability of (4a)
versus (4b):
(4) a. The sunshine reddened Amys cheeks.
b. *The sunshine bloomed the tulips.
McKoon and MacFarland tested the validity of this empirical claim through a
number of dierent studies. First, they examined a corpus of approximately 180
million words of written and spoken English. The results of that investigation,
summarized in Table 1, below demonstrate that ICSs considered as a class are
in fact just as likely to transitivise as ECSs.
Table 1.Probability of transitive use for ECS and ICS verbs (adapted from McKoon and
MacFarland 2000: Table 2)
SentenceType
External Cause
Internal Cause
verb
prob. yes
verb
prob. yes
atrophy
awake
crumble
abate
.03
.05
.05
.10
bloom
deteriorate
germinate
rot
.00
.01
.06
.08
mean
.06
mean
.06
redden
dissipate
fray
fossilize
.24
.41
.52
.60
blister
ferment
corrode
erode
.22
.54
.63
.67
mean
.48
mean
.45
in the acceptability ratings. Overall, the acceptability rate for both classes of
verbs was extremely high: even those verbs judged least acceptable had a mean
acceptance rate of over 80%.
The results of the rst two studies speak against the theoretical claim that
these two types of verbs are represented dierently. On the contrary, subjects
close convergence on a scale of relative acceptability (for the transitive forms of
individual predicates) suggests that a ne-grained, lexically determined, and
inherently gradient, type of competence informs native-speakers performance.
Had McKoon and MacFarland stopped at this point, their work would
provide support for the usage-based, probabilistic models of language performance favoured by connectionists and others (cf. especially MacDonald et al.
(1994), Barlow and Kemmer (2000)). A nal experiment, however, suggested
a dierent conclusion. In this latter, implicit experiment, McKoon and
MacFarland measured the response latencies involved in reading ECSs versus
ICSs that had previously been matched in terms of length and oine acceptability. In direct contrast to the previous results, this implicit measure revealed
a reliable contrast in reading time between ECS and ICS verbs; ECS verbs
whether presented in intransitive or transitive frames took signicantly
longer to read than their matched ICS counterparts. The main results are
reproduced in Table 2 below (from McKoon and MacFarland: Tables 7 and 8:
848, 851). That is to say, the results of these last experiments support the idea of
distinct representations for these verbs on the basis of the hypothesized verb
classes.
McKoon and MacFarland argue that the results of the nal experiment
speaks against probabilistic models of lexical representation, and conrm the
psycholinguistic reality of the theoretical model. Both conclusions are too
strong, I think. On the one hand, whether or not there is a categorical distinction in the representation of the two classes of verbs, it is still necessary to
account for subjects ability to converge on a scale of gradient judgments for
individual predicates. As discussed earlier, a speaker whose judgments of the
acceptability of such predicates was the inverse of all other subjects (in Experiment 2) could reasonably be judged less competent than one whose relative
judgments were in accordance with other native speakers (at least by the second
denition of competence), even if their reading times in Experiment 3 were
comparable.
Moreover, the fact that subjects performance in an implicit task yields a
statistically discrete result does not prove that the competence underlying this
behaviour is itself categorical, and certainly is no more than consistent with the
Table 2.Mean reading times for ECS and ICS verbs (from McKoon and MacFarland
2000: Tables 7 & 8)
Sentence Type
External Cause
Internal Cause
Intransitive frames
JTime (ms)
prob. yes
JTime (ms)
prob. yes
1551
1561
1538
.91
.92
.90
1400
1392
1413
.96
.96
.96
Intransitive frames
External Cause
Internal Cause
JTime (ms)
prob. yes
JTime (ms)
prob. yes
2220
2230
2210
.86
.81
.93
2069
2131
1963
.96
.96
.96
(5) a.
Je veux le voir.
I want 3sg to.see
I want to see him.
b. *Je le veux voir.
c. Je le fais chanter.
I 3sg make to.sing
I make him sing.
d. *Je fais le chanter.
The phenomenon of interest, illustrated in (5) above, is the contrast in pronominal clitic placement between so-called restructuring verbs (such as pouvoir,
vouloir) and the causative verb faire. The distinction is of theoretical interest in
that the two structures are assumed to involve distinct syntactic representations.
Briey, many theoretical analyses assume that restructuring verbs as the
name suggests restructure the syntax of the lower clause, creating a monoclausal structure. By contrast, sentences involving causatives are assumed to
remain fundamentally bi-clausal: at an abstract level of representation, the
pronominal clitic is still syntactically associated with the lower verb; see
Dueld et al. (2002), for more detailed discussion.
Notice that this analysis, derived on the basis of cross-linguistic comparison
with other Romance varieties, is precisely the opposite of that suggested by a
nave inspection of the French facts. In French, the pronominal clitic stays close
to the verb of which it is an argument in restructuring contexts, and is displaced
from it in causative contexts: hence, one might expect that restructuring would
be required for the interpretation of clitics with causatives, rather than with
verbs like vouloir.
The theoretical analysis, however, predicts a distinction between the
acceptability of two types of ungrammatical sentence, namely, between (5b)
and (5d) above: whereas (5b) is predicted to be ungrammatical at all levels of
representation, (5d), in which the clitic is attached to the verb with which it is
thematically associated, is predicted to be ungrammatical only at surface level.
This predicts that if a task could be found that taps underlying, rather than
surface, grammaticality, then sentences such as (5d) should pattern with other
grammatical, as opposed to ungrammatical, sentences.
In Dueld et al. (2002), we employed just such a task to elicit implicit
grammaticality judgements. This is the Sentence-Matching (SM) paradigm
introduced by Freedman and Forster (1985), later developed for SLA research
by Bley-Vroman and Masterson (1989); see also Eubank (1993), Eubank and
Grace (1988). In this task, subjects are asked to determine whether or not two
110
Nigel Dueld
In our own experiments on clitic placement (see Dueld et al. 2002), we determined that in SM native-speakers treat ungrammatical causatives dierently
from ungrammatical restructuring sentences. As predicted by the theoretical
analysis, French native speakers treat cases such as (5d) on a par with other
grammatical sentences: there was no signicant dierence in response latency
between (5c) pairs and (5d) pairs; by contrast, it took subjects signicantly
longer to match (5b) pairs than their grammatical counterparts (5a). Crucially,
in all other conditions involving incorrect clitic placement, French nativespeakers were signicantly slower to match sentences compared to the corresponding correct placements of these clitics. Thus, it was not the case that SM
overall was insensitive to constraints on clitic placement; quite the contrary.
Instead, SM was selectively insensitive to surface violations of grammaticality:
sentences that were underlyingly grammatical were accepted as grammatical,
even though the surface string was ungrammatical.
In oine grammatical judgment tasks, on the other hand, French nativespeakers consistently treat surface ungrammatical sentences equally: oine,
(5d) is considered no more grammatical than (5b). Thus, there is once again a
divergence between the results obtained by implicit versus explicit methodologies. In contrast to the McKoon and MacFarland experiments discussed in the
previous section, here it is the selective absence of a specic result from the
111
112
Nigel Dueld
Change of Location
Change of State
Continuation of a pre-existing state
Existence of State
Uncontrolled process
Controlled process (motional)
Controlled process (non-motional)
(9) a.
(10) a.
If only the poles of the continuum are considered, this contrast gives the
appearance of being categorical; however, once one considers the middle range,
it becomes clear that auxiliary selection is a lexically gradient phenomenon.13
2.2.2 Syntactic/constructional gradience
In other constructions, however, gradient eects are observed that are independent of the particular lexical items involved. That is to say, certain sentences are
regularly judged to be less than perfectly acceptable without being deemed
wholly unacceptable. In the theoretical literature, such sentences are typically
designated as marginal, a status denoted by one or two question-marks (?/??).14
However, as noted above, since standard theoretical models have no way of
representing such judgments, marginal sentences are usually re-classied as
grammatical or ungrammatical ad hoc, depending on the analysis that is being
pursued. Such reclassication immediately obscures an essential feature of most
acceptability judgments, namely, their syntactic gradience.
One example of this type of gradience has already been mentioned, viz., the
inuence of referential specicity in determining the relative strength of
syntactic island eects: Kluender (1992), see also Kluender and Kutas (1993).
A dierent example of syntactic gradience is provided by my work with
Ayumi Matsuo on VP-ellipsis constructions in English. Ellipsis constructions,
113
114
Nigel Dueld
and the constraints pertaining to them, have provided core data for generative
analyses for several decades, their importance rst brought to general attention
in Sags dissertation (1976). The aspect of ellipsis constructions relevant to the
present discussion is a constraint on structural parallelism, which the
theoretical literature claims requires the VP of the antecedent clause to be
syntactically parallel to that of the understood ellipsis. This structural parallelism constraint is used to explain the contrast between (a) versus (b) examples
in (11) and (12) below: examples (11a) and (12a) show VP-ellipsis with
parallel active/verbal antecedent clauses; those in (11b) and (12b) illustrate
two types of non-parallel antecedent, passive VPs and nominal antecedents,
respectively. The examples in (11/12c) and (11/12d) are intended to show that
this parallelism constraint fails to apply or, at least, does not apply so
strongly if the ellipsis (VPE) is replaced with the semantically equivalent
VP-anaphora (VPA) clause.
(11) a.
(VPE)
(VPA)
(12) a.
The structural parallelism eect is interesting for at least two reasons. First, for
native-speakers, the parallelism constraint has generally gradient, rather than
categorical eects. That is to say, native-speakers typically disprefer, but do not
necessarily exclude, violations of structural parallelism with VPE (the (b)
examples above). This has been demonstrated experimentally in Tanenhaus and
Carlson (1990), as well as in our own work (Dueld and Matsuo 2001, 2002).
The availability of non-parallel ellipsis, in contrast to some other kinds of
Hence, it seems fair to claim that such sentences have a dierent status from
those that native-speakers quite generally reject as unacceptable.
A second point to observe about non-parallel ellipsis is that constructiontype seems to be a factor in determining acceptability. Once again, experimental
evidence just cited conrms the intuition that non-parallel ellipsis where the
antecedent is a derived nominal (12b) is signicantly less acceptable than nonparallel ellipsis where the antecedent is a passive VP (11b) (though it still remains
signicantly more acceptable than some other kinds of ungrammatical sentence).
Standard theoretical analyses of ellipsis have no way to represent either the
gradient eects of the parallelism constraint overall, or the dierential eects of
construction type. Hence, in the theoretical literature, the relative acceptability
judgments just described get recoded categorically, with non-parallel ellipsis
being considered uniformly ungrammatical (*), irrespective of the particular
type of antecedent, and non-parallel anaphora ((11d)/(12d)) deemed perfectly
acceptable, native-speakers intuitions notwithstanding. This contrast is
represented schematically in Table 3.
It should be stressed that this type of gradience is orthogonal to lexical
gradience: for each condition tested in our experiments, dierent verbs and
dierent auxiliaries were deemed more or less acceptable in ellipsis contexts;
crucially, though, all of the verbs were accepted some of the time in nonparallel contexts.
The main statistical ndings for native speakers were as follows. First, in
both experiments, there was a reliable main eect of parallelism: in particular,
VP-ellipsis following non-parallel antecedents was signicantly less acceptable
than following parallel antecedents, irrespective of construction type. Second,
115
116
Nigel Dueld
Table 3.Designated vs. actual acceptability judgments for parallel vs. non-parallel
antecedents (VPE and VPA completions). RH column shows acceptance rates as
percentages for trials in Dueld & Matsuo (2001), (2002), respectively.
Antecedent-ellipsis
type
Designated
grammaticality
judgment
Active-VPE
Passive-VPE
90
52
88
48
Verbal-VPE
Nominal-VPE
??
89
39
93
57*
Active-VPA
Passive-VPA
96
91
87
84
Verbal-VPA
Nominal-VPA
97
74
88
76
* The relatively high acceptance rate for VPE following nominal antecedents is due to the dierent
balance of nite and non-nite ellipsis sentences in the latter experiment. See 2.4 below for further
discussion.
two clear interactions were observed: between ellipsis type and parallelism (VPE
following non-parallel antecedents is reliably less acceptable than VPA in the
same context), and between construction type and parallelism (VPE following
passive antecedents is signicantly more acceptable than following nominal
antecedents, though still less acceptable than following active antecedents).
Third, VPA also shows a reliable parallelism eect with nominal antecedents,
albeit a smaller one compared to the VPE eect.
Before discussing the second language learners, let us consider how we might
model the native-speaker results, rst, given the traditional competenceperformance model, and then within the dual competence model proposed here.
As far as I can determine, it is simply impossible to model these gradient
eects in the traditional framework without arbitrarily recoding them as
categorical eects. One could, for example, model the main eect of parallelism
by reclassifying the circa 50% acceptance rate for VPE in passive contexts as
equivalent to categorically unacceptable sentences, say, those with less than 5%
acceptance ratings by native speakers. One could also gloss over the statistically
reliable dierence between construction types: since the principles and parameters approach allows no construction-specic rules, it cannot allow construction-specic eects to bear on grammaticality. Finally, one could dismiss the
117
118
Nigel Dueld
speakers: in this sense, these intermediate learners were more competent rather,
closer to underlying competence than the others. On the other hand, they were
clearly less competent than the advanced learners in converging on the judgments, and by extension, on the overall competence, of native-speakers, since
in this case target competence is lexically-constrained, gradient behaviour.
Although this contrast is not exactly comparable to the other phenomena
discussed in this paper, since it is purely lexical, rather than syntactic, the same
logic applies: some grammatical phenomena are categorical and autonomous
properties, others show lexically-specic, gradient eects; both need to be
accommodated.
2.3.2 Type 2 disjoint convergence
The converse behaviour where second language learners results reect
surface competence when native-speakers results show the inuence of UC
can be seen in the sentence-matching experiments on clitic placement discussed
earlier. Recall that the claim was that native-speakers failure to show a grammaticality eect (in the online task) in French causative constructions in
clear contrast to the grammaticality eect they exhibited for restructuring verbs
was due to the underlying grammaticality of the clitic placement in
sentences such as (5d) above.
If this is the correct explanation of the observed asymmetry, then the
predictions for the second language learners on this task are somewhat paradoxical, raising the possibility that L2 learners implicit acceptability judgments
might fail to match those of native speakers by outperforming them with
respect to the presumed theoretical target. In our experiment, this is precisely
what happened: both English and Spanish speaking L2 learners of French
showed a grammaticality eect for both restructuring and causative contexts in
the online task, in contrast to the French native speakers.
Once again, the standard model provides no satisfactory account of these
results: either one is forced to exclude the causative condition altogether on the
grounds that it did not work for the native-speaker controls, or one accepts
(paradoxically) that the L2 learners have achieved native-speaker competence
in this condition, as measured by approximation to the theoretical target,
although their implicit judgments are wholly distinct from those of the nativespeakers. Once more, something signicant is lost.
By contrast, this pattern of results can be accommodated directly by the
dual competence model, as schematised in Figure 5 above. The necessary
assumption would be that whereas native-speakers analyse syntactic structures
119
the antecedent clause in the active-passive experiment (see above), such that the
linguistic information necessary to (re)-construct and interpret the ellipsis
clause was more or less recoverable from the antecedent. Specically, we
hypothesized that (conceptual) recoverability of non-parallel passive antecedents would be enhanced by the presence of a by-phrase, as for example in (14a)
versus (14b) below.16
(14) a.
The other property manipulated was the niteness of the ellipsis clause.
Standard theoretical accounts do not distinguish between non-nite ellipsis
involving to as in (11) above, on the one hand, and nite ellipsis involving
do, or some other auxiliary verb as in (12) above, on the other. That is to say, the
parallelism eect is generally claimed to constrain nite and non-nite ellipsis
equally. Intuitively, however, the parallelism eect is considerably weaker with
non-nite ellipsis. These experiments tested that intuition experimentally.
Detailed results and discussion of these experiments are reported in
Dueld and Matsuo (in preparation). Here, it suces to report the main
ndings, which were as follows (see also Dueld and Matsuo 2002). First,
contrary to standard theoretical assumptions, our experiments show that the
parallelism eect in ellipsis is not uniquely due to the structural properties of
the antecedent clause: other lexical and conceptual factors interact to determine
the strength of the eect. Of these factors, niteness is crucial: non-nite ellipsis
displays signicantly weaker parallelism eects in non-parallel contexts than
nite ellipsis. Second, conceptual recoverability does have an eect on the
acceptability of non-parallel antecedents, but only for native speakers, at
least in interaction with niteness: recoverability weakens the parallelism
eect only in non-nite contexts.
The comparison between native-speakers and L2 learners performance was
also revealing. Overall, L2 learners performance parallels that of nativespeakers: both groups exhibit a signicantly lower acceptance of ellipsis in nonparallel contexts; and, for both groups, this eect is gradient, rather than
categorical (just as in the previous experiments). This clearly indicates that
gradient eects can be successfully acquired in SLA.
On the other hand, the constraint ranking underlying native speakers and
L2 learners common overall results appear to be quite dierent. Whereas
121
<DEST "duf-n*">
niteness shows a robust main eect for both groups, for L2 learners conceptual recoverability has an ameliorating eect even in nite clauses (which was not
the case for native speakers). This suggests that this type of conceptual information plays a larger role in determining L2 learners judgments than it does for
native speakers, who rely more on purely formal information.
Whatever the nal interpretation of these results should be, it should be
clear that such ndings are only attainable in principle if attention is paid to the
details of gradient eects: a categorical model can neither describe nor accommodate them.
3. Conclusion
The purpose of this paper has been to draw attention to various types of
gradient eects, both lexical and syntactic. As suggested in the title, I have argued
that these eects form an essential part of our implicit grammatical competence.
A revised model of competence was proposed that accommodates these eects,
but which maintains the generativist assumption that certain core aspects of
grammatical knowledge are still categorical, and autonomous of the lexicon.
The proposed model was also shown to oer an explanation for apparent
paradoxes that arise in SLA whenever second language learners outperform
native-speakers: by distinguishing two types of implicit knowledge, it is possible
to oer a principled account for certain systematic mismatches between nativespeakers and second language learners. Finally, this model may ultimately allow
us to bridge the gap between those who argue for strong continuity in SLA and
those who advocate fundamental dierences. There is considerable empirical
evidence for both positions; it could be that both are correct.
Notes
* I am grateful to two anonymous reviewers for constructive comments and suggestions. I
would also like to thank those people who commented on previous drafts of this paper,
including David Birdsong, Jonathan Bobaljik, Tom Roeper and Lydia White. I am especially
grateful to Jonathan Bobaljik for clarifying many misunderstandings on my part, and for
oering a persuasive defense of the standard approach. Unfortunately, due to time constraints, I have not been able to integrate all suggested revisions into this paper. No-one other
than myself is responsible for remaining errors and inconsistencies.
1. These assumptions were not always held: as noted by Levelt et al. (1977), theoretical
researchers in the sixties and early seventies developed theories on degrees of ungrammaticality (Levelt et al. cite Chomsky 1964, Katz 1964, Zi 1964, and Lako 1971).
2. I postpone any discussion of the status of marginal (?) sentences, since most theoretical
analyses actually end up designating such sentences either as grammatical or ungrammatical,
usually the former; their relative deviance or amelioration relative to the other
sentences of their designated type being attributed to peripheral, extra-syntactic, factors. For
further discussion, see Schtze (1996: especially pp. 41).
3. In this sense, there has been comparatively little progress from a notion of grammaticality
dened in terms of extensional (innite) sets: in this model, particular sentences are either
generated by the grammar, or they are not. Yet most generative linguists would claim,
following Chomsky (1986), that I-language, rather than E-language, is the proper object of
study (see also Hoekstra 1990).
4. Of course, denying the competenceperformance distinction is nothing new; in the past,
though, its detractors often misunderstood or vastly underestimated the intricacy and
complexity of grammatical knowledge.
5. The paradox arises most clearly in L2 research simply because L2 researchers are, of necessity,
more acutely aware than are theoretical linguists of the metalinguistic nature of linguistic
judgments, and of the methodological and analytic problems of data collection and comparison.
6. The comparison involved here may be direct, where L2 subjects judgments are compared
with the judgments of a control group of native-speakers, or indirect, where L2 judgments
are compared with a pre-established set of judgments (perhaps taken from a theoretical paper),
which native-speakers presumably would agree on; either way, L2 learners are judged competent
if their judgments match those of native-speakers in some statistically reliable way.
7. It might be objected that such a conict only arises where native-speakers judgments are
gradient. However, the contention here is that almost all apparently categorical judgments
are in fact gradient (when properly analyzed); hence, there is a real problem here.
8. This is a possible move, provided that there is something interesting left to a judgment
once the gradient properties have been factored out of the equation. Often, though, it seems
as if there is nothing left, no interesting residue that UG could explain. This again echoes
Culicover (1998: 48):
Chomsky has argued consistently that this perspective about linguistic theory [including
the notion of UG as an idealized characterization of linguistic competence: NGD] is
rational and scientic, virtually indisputable. In fact, it cannot reasonably be disputed
given the presumptions that: (i) a language faculty exists that contains specic syntactic
knowledge; (ii) what is left after stripping away the dynamical aspects of language is
something that really exists, in some sense, in the mind/brain (emphasis mine: NGD).
9. This latter assumption may of course be incorrect. The guiding intuition here is that the
introspection involved in explicit tasks is inevitably mediated (in adult native speakers) by
lexical knowledge, a product of surface competence. My speculation is that direct introspection of the computational system is impossible.
123
10. Since McKoon and MacFarland did not test L2 learners, this interpretation is intended
to apply to native-speakers only.
11. Here, I make simplifying assumption that these are independent factors. Obviously, this
is not always the case: if, for example, inherent semantic constraints restrict or reduce the
occurrence of a verb in a transitive frame (see immediately below), this will aect the token
frequency for that item, which in turn may further inhibit its acceptability.
12. Numbers in square brackets designate Soraces original example numbers. Example 9b
[11] above is originally due to van Hout (1993: 7).
13. Similar remarks would seem to apply to other unaccusative eects: for example,
there-insertion in English (see Levin & Rappaport Hovav 1995).
14. Crucially, the marginal status of these sentences emerges from a uni-modal pattern of
acceptances: all speakers accept these sentences sometimes, and reject them on other
occasions; see Avrutin and Wexler (1992), for a relevant discussion of uni-modal vs. bimodal patterns of acceptance, and their proper interpretation.
15. In response to a reviewers query, the expression constraint ranking is not intended here
to imply a treatment in terms of Optimality Theory necessarily. It is not obvious that
standard OT models capture gradient eects any better than mainstream generative models,
since violable constraints do not yield gradient judgments (in most models, at any rate).
Rather, the term is intended to refer to dierences in the relative weighting of various lexical
and syntactic factors that determine the judgment. How these should relate to a particular
theoretical description is an independent question.
16. In the case of non-parallel nominal antecedents, we manipulated syntactic, rather than
conceptual, recoverability. Here, we contrasted zero-derived versus non-zero-derived
alternations (e.g., visit versus discussion), since it has been argued that the former (zeroderived nominals) are more easily reconstructable as verb-phrases in VPE contexts. No eect
was found for this type of syntactic recoverability (see Dueld and Matsuo in preparation).
References
Allen, J. and Seidenberg, M. S. 1999. The emergence of grammaticality in connectionist
networks. In The emergence of language, B. Macwhinney (ed.), 115152. Mahwah, NJ:
Erlbaum.
Avrutin, S. and Wexler, K. 1992. Development of principle B in Russian: Co-indexation at
LF and coreference. Language Acquisition 2 (4): 259306.
Barlow, M. and Kemmer, S. 2000. Usage-based models of language. Stanford: Center for the
Study of Language and Information.
Bever, T. G. 1970. The cognitive basis for linguistic structures. In The development of
language, J. R. Hays (ed.), 279362. New York: John Wiley & Sons.
Birdsong, D. 1989. Metalinguistic performance and interlinguistic competence. New York:
Springer-Verlag.
125
Eubank, L. and Grace, S. 1988. V-to-I and inection in non-native grammars. In Morphology
and its interface in L2 knowledge, M.-L. Beck (ed.), 6988. Amsterdam: John Benjamins.
Fodor, J. D. 2001. Parameters and the periphery: Reections on syntactic nuts. Journal of
Linguistics 37 (2): 367392.
Freedman, S. E. and Forster, K. I. 1985. The psychological status of overgenerated sentences. Cognition 19: 101131.
Greenbaum, S. 1977. Acceptability in language. The Hague: Mouton.
Hardt, D. 1993. Verb phrase ellipsis: Form, meaning and processing. Computer and Information Science, University of Pennsylvania: Ph.D. dissertation.
Hedgcock, J. 1993. Well-formed vs. ill-formed strings in L2 metalingual tasks: Specifying
features of grammaticality judgments. Second Language Research 91: 121.
Hoekstra, T. 1990. Markedness and growth. In Logical issues in language acquisition, I. Roca
(ed.), 6383. Dordrecht: Foris.
Katz, J. J. 1964. Semi-Sentences. In The structure of language: Readings in the philosophy of
language, J. A. Fodor and J. J. Katz (eds), 400416. Englewood Clis: Prentice Hall.
Kluender, R. 1992. Deriving island constraints from principles of predication. In Island
constraints: Theory, acquisition and processing, H. Goodluck and M. Rochemont (eds),
195222. Dordrecht & Boston: Kluwer.
Kluender, R. and Kutas, M. 1993. Subjacency as a processing phenomenon. Language and
cognitive processes 8 (4): 573640.
Lako, G. 1971. Presuppositions and wellformedness. In Semantics, D. D. Steinberg, and
L. A. Jakobovitz (eds). London: Cambridge University Press.
Levelt, W. J. M., Van Gent, J. A. W. M., Haans, A. F. J. and Meijers, A. J. 1977. Grammaticality, paraphrase, imagery. In Acceptability in language, S. Greenbaum (ed.), 87101. The
Hague: Mouton.
Levin, B. and Rappaport Hovav, M. 1995. Unaccusativity: At the syntax-lexical semantics
interface. Vol. 26. Linguistic Inquiry Monograph Series. Cambridge, Mass.: MIT Press.
Macdonald, M.-E. C., Pearlmutter, N. J. and Seidenberg, M. A. 1994. Syntactic ambiguity
resolution as lexical ambiguity resolution. In Perspectives on sentence processing, C.
Clifton Jr., L. Frazier and K. Rayner (eds), 123154. Hillsdale, NJ: Erlbaum.
Mandell, P. B. 1999. On the reliability of grammaticality judgment tests in second language
acquisition research. Second Language Acquisition 15 (1): 7399.
Marslen-Wilson, W., Tyler, L. K., Waksler, R. and Older, L. 1994. Morphology and meaning
in the English mental lexicon. Psychological Review 101 (1): 333.
Martohardjono, G. 1998. Measuring competence in L2 acquisition: Commentary on part I.
In The generative study of second language acquisition, S. Flynn, G. Martohardjono and
W. ONeil (eds), 151157. Mahwah, NJ: Lawrence Erlbaum Associates.
McKoon, G. and MacFarland, T. 2000. Externally and internally cause change of state
verbs. Language 76 (4): 833858.
Sag, I. 1976. Deletion and logical form. MIT: Doctoral dissertation.
Schtze, C. 1996. The empirical base of linguistics. Chicago: University of Chicago Press.
Sorace, A. 1993. Incomplete and divergent representations of unaccusativity in non-native
grammars of Italian. Second Language Research 9: 2248.
</TARGET "duf">
<LINK "dyk-n*">
<TARGET "dyk" DOCINFO AUTHOR "Ton Dijkstra"TITLE "Lexical storage and retrieval in bilinguals"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 6
1.
Introduction
namely the language of the preceding words, which might provide an independent source of constraint on bilingual word recognition.
Because the recognition of words in sentence context is so complex, it is no
wonder that most studies in this area during the last decades have focussed on
the bilinguals recognition of isolated words. This process already requires a
distinction of dierent types of structural representations for words (for
instance, orthographic, semantic, and phonological; a storage issue), and is
inextricably bound up with how words are retrieved (a processing issue), for
which purposes they are retrieved (issues with respect to task demands and
instruction), and in which non-linguistic context their retrieval takes place (e.g.
is the word positioned within a stimulus list containing words from the same or
a dierent language?).
The major part of this chapter is specically concerned with the issue of
isolated word recognition, and it addresses the following questions:
1. Structure: which representations are activated during bilingual word
recognition?
2. Process: what is the time-course of activation for words of dierent languages?
3. Contextual constraints: how do non-linguistic experimental factors, such as
participant expectancies and instruction, aect lexical selection in bilinguals?
In the short second part of this chapter, we will discuss a few recent studies that
have examined eects of linguistic context (syntactic, semantic, and lexical) on
the recognition of words in sentence context. At present, only a handful of
reaction time (RT) studies has been done, investigating quite diverse linguistic
questions within the perspective of dierent theoretical frameworks. Given this
sorry state of aairs, we will focus on the more coherent studies that have
investigated the bilinguals brain activity during the processing of syntactic and
semantically incorrect sentences in terms of Event-Related Potentials (ERPs).
We will argue that while semantic processing may be quantitatively dierent
between monolinguals and bilinguals, syntactic parsing may be both quantitatively and qualitatively dierent, in complex ways that depend, for instance, on
the L2 prociency of the bilinguals involved.
131
132
Ton Dijkstra
belonging to one of two languages. One button must be pressed when a word
belongs to one language (e.g. English), and another if it belongs to the other
language (e.g. Dutch). In the go/no-go task, only words from two languages are
presented as well. However, participants react only when they identify a word
from the target language, for instance, English (English go/no-go), but they do
nothing if a word of the non-target language (for instance, Dutch) is presented.
In progressive demasking, participants identify a word that is presented in an
alternating sequence with a pattern mask, for instance a checkerboard or a row
of hash marks. Across alternations, the pattern mask decreases in duration,
while stimulus presentation increases, until the target word is recognized. In
word naming, the participants must read aloud words in the target language, for
instance, in their mother tongue (e.g. Dutch) or in their second language (e.g.
English). Finally, in word association, participants respond to a presented word
by producing the rst word that comes to their mind in the target language.
2.1 Structure: Which representations are activated during bilingual
word recognition?
Research has shown that in the initial stages of monolingual word recognition,
an input letter string leads to the activation of multiple word candidates in the
mental lexicon that closely match the input (see Grainger and Dijkstra 1996).
For instance, when the letter string word is presented to English monolinguals,
the stored representations for words like word, cord, ward, wood, and work will
initially become active (such candidates are called neighbours). When the word
identication process proceeds, inappropriate candidates will gradually be
reduced in activation and no longer be considered as a possible input word.
Finally, only the lexical representation corresponding to the presented word
remains active and becomes recognized. For bilinguals, the interesting question
arises if word candidates from dierent languages are activated if they overlap
suciently with the input letter string. For instance, are both the English and
the Dutch readings of the interlingual homograph angel activated in parallel if
Dutch-English bilinguals read the word in an English book?
According to the language-selective access view on bilingual word recognition, only lexical candidates of the task-relevant language (in this example
English) are activated (Gerard and Scarborough 1989, Macnamara and Kushnir
1971). Thus, when the word word is presented, Dutch word candidates that are
similar to word, like bord and wond would not become activated. In contrast,
according to the language non-selective access view, word candidates from
both languages (here English and Dutch) become active (Altenberg and Cairns
1983, Grainger and Beauvillain 1987, Van Heuven, Dijkstra and Grainger 1998).
As we shall see later, the majority of recent studies support the language nonselective access hypothesis. However, formulating the lexical access views in this
general way ignores at least two important points. First, the issue of (non)selective access should be dierentiated with respect to dierent types of lexical
representations: e.g. orthographic, phonological, and semantic representations.
Second, the answer to the question might depend on whether one is processing
in ones native language (L1) or in a second language (L2).
Dijkstra, Grainger and Van Heuven (1999) found evidence of crosslinguistic competition between words of dierent languages that are similar in
form and/or meaning. They investigated whether Dutch-English bilinguals
recognized interlingual homographs faster or slower than matched onelanguage control items in English lexical decision and progressive demasking
tasks. The English stimulus words varied in their degree of orthographic (O),
phonological (P), and semantic (S) overlap with Dutch words. Examples of
items in their six test conditions are sport (overlap in S, O, and P codes), wild
(SO), wheel (SP), pink (OP), angel (O), and pace (P). The rst two conditions
(SOP and SO conditions) consist of what are usually called cognates, while the
last three conditions contain interlingual homographs (OP and O conditions)
or interlingual homophones (P condition).
Participants were faster to make a lexical decision to the target words with
cross-linguistic overlap than to exclusively English control words if the overlap
was orthographic and/or semantic in nature (e.g. in the SO and O conditions).
In contrast, cross-linguistic phonological overlap produced inhibitory eects.
Responses to test items of the P condition, for example, were slower than to
matched purely English words.
To show that the observed result pattern did not arise because the test and
control items were not well matched in some unknown aspects, the lexical decision
experiment was replicated with American-English monolinguals. For these
participants, no RT or error dierences were found between test items and their
matched controls. Furthermore, to show that the results were not restricted to
lexical decision, the experimental materials were also included in a progressive
demasking task. The result pattern obtained with this paradigm was strikingly
similar to that in lexical decision, indicating that it was not the task that induced
the facilitation and inhibition eects for homographs relative to controls.
In sum, a presented word form in L2 appears to initially activate orthographic, phonological, and semantic lexical representations in both L2 and L1.
133
by her) are still facilitated but signicantly less so than identical cognates.
Furthermore, she has shown that the amount of facilitation that is observed
depends on the position of the deviating letter in the word. Neighbour cognates
with the dierent letter at the end of the word (e.g. French texte Spanish texto)
are facilitated more than neighbour cognates with the dierent letter inside (e.g.
French usuel Spanish usual). In fact, facilitatory eects for the latter type of
cognate disappeared and eects tended towards inhibition when such cognates
were of low frequency in both languages. Similar patterns of results were found
in both L1 and L2 processing.
These results make it likely that the size of RT eects observed for cognates
and interlingual homographs depends on their degree of cross-linguistic
overlap (also cf. Cristoanini, Kirsner and Milech 1986). Note that it follows
logically that across language pairs that do not share orthography at all (e.g.
Chinese and English), no orthographically similar word candidates can be
activated, while eects of phonological similarity might still occur (depending
on, for instance, the way tonal information aects the establishment of the set
of lexical candidates).
2.2 Process: What is the time-course of activation for words
from dierent languages?
The issue of language (non)selective access can also be examined from a
processing point of view by considering the time-course of lexical activation and
selection in bilingual word recognition. As a rst question, we may consider the
rate of code activation in L1 and L2: how fast do orthographic, phonological, and
semantic codes from the two languages become active? From the monolingual
domain, we know that high frequency words are generally recognized faster
than low frequency words, and, because the words of L2 must have a lower
subjective frequency than those of L1 (simply because the former have been
encountered less often), it seems likely that L2 codes become available slower
than L1 codes. A comparison of the study by Dijkstra et al. (1999), discussed in
the previous section, with a study by Lemhfer and Dijkstra (submitted)
provides information that supports this viewpoint.
Dijkstra et al. (1999) showed that in a lexical decision task where L2
(English) was the target language of the bilingual participants, cross-linguistic
eects arose for L1-L2 (Dutch-English) homographs with respect to all three
codes. Because English was the target language in this task, task execution
implied the verication of the English language membership of possible word
135
candidates, even when Dutch codes would be available faster than English ones.
In other words, Dutch codes had time to establish themselves and exert eects
on later available English targets that were necessary for responding.
Lemhfer and Dijkstra (submitted) presented the same stimulus materials
to Dutch-English bilinguals in a generalized lexical decision task. In this task,
participants responded with yes to both English and Dutch words, but with
no to non-words. In contrast to English lexical decision, participants in this
task can use both Dutch and English lexical representations as a reliable basis
for response. Thus, in this task cross-linguistic eects will arise only to the
extent that L1 and L2 codes can aect each other before the fastest codes
(usually Dutch ones, we assume) are retrieved and responded to. The results of
this study were quite clear: no facilitation eects arose for interlingual homographs, while cognates were facilitated relative to control words. The pattern of
results for homographs indicates that responses were based upon the fastest
available code, usually the Dutch orthographic code, while cross-linguistic
overlap with respect to semantics in the case of cognates apparently can be used
to speed up the response.
In sum, even though L1 and L2 codes become active in parallel, L2 codes
are often activated more slowly than L1 codes, probably because of dierences
in subjective frequency between languages. As a consequence, the development
of cross-linguistic eects depends on the target language in the experiment (L1
or L2) and on other temporal characteristics of the task involved.
There is a dierent way of approaching the issue of the time-course of
lexical selection in bilinguals. Rather than asking how fast lexical representations from dierent languages become active, we might wonder how long they
remain active. Even if there is an initial activation of various codes from
dierent languages, lexical selection might be relatively fast or slow afterwards.
This issue has been investigated in experimental studies by varying the frequency ratio of the two readings of interlingual homographs (e.g. angel is relatively
more frequent in English than in Dutch).
Dijkstra, Timmermans and Schriefers (2000) examined how long the two
readings of an interlingual homograph compete for selection and whether
language information provided by the item can be used to facilitate the selection of one of these readings. In three experiments, each with a dierent
instruction, bilingual participants processed the same set of homographs
embedded in identical mixed-language lists. Homographs of three types were
used: high-frequent in English and low-frequent in Dutch; low-frequent in
English and high-frequent in Dutch; and low-frequent in both languages. In the
rst experiment (involving language decision), one button was pressed when an
English word was presented and another button for a Dutch word. In the
second and third experiments participants reacted only when they identied
either an English word (English go/no-go) or a Dutch word (Dutch go/no-go),
but they did not respond if a word of the non-target language (Dutch or
English, respectively) was presented.
In all three tasks, clear inhibition eects arose for homographs relative to
one-language controls. Even in the Dutch go/no-go task for Dutch-English
bilinguals performing in their native language, participants were unable to
completely exclude eects from the non-target language on homograph
identication. More important for the present discussion, however, is the
nding that target-language homographs were often overlooked, especially if
the frequency of their other-language competitor was high. The relative
frequency of the two readings of the interlingual homograph was found to aect
both RTs and error rates. In the Dutch go/no-go task, participants did not
respond to low-frequency items belonging to their native language in about 25
percent of the cases!
Inspection of cumulative distributions showed that if they did not respond
after about 15001600 ms, they did not respond anymore within the time
window of two seconds. The observed attening of the cumulative distribution
towards an asymptotic value suggests that recognition of the homograph
reading from the non-target language in some way prohibits the subsequent
recognition of the target language reading (e.g. after recognition, all other
lexical candidates may be suppressed). Thus, selection of one of the readings of
the interlexical homographs takes place rather late during processing. The
results suggest that until that time both readings of a presented homograph are
involved in a (frequency-dependent) race to recognition that is won by the
fastest candidate.
It is clear that the system must at some time arrive at a selection of one
lexical item only, but apparently the role played by the language of that item in
aiding selection is only minor. In fact, determination of the language of the item
may depend on lexical selection having taken place. In addition, it does not
seem possible to discard the homograph reading from the non-target language
and to focus on the target reading only on the basis of the instruction that just
the target language needs to be responded to.
Similar results were found when the target language was English (L2) and
when it was Dutch (L1), even though fewer target words were overlooked in the
second case. Again, this nding points to activation from both readings of the
137
138
Ton Dijkstra
lexical decision task including interlingual homographs, but, apart from nonwords, Dutch words were also incorporated, requiring a no-reaction. Strong
inhibition eects were now obtained for interlingual homographs relative to
English control words. The size of the inhibition eect depended on the relative
frequency dierence of the two readings of the homograph. It was largest when
the Dutch reading of the homographs had a high frequency relative to the
English reading. In Experiment 3, Dijkstra et al. (1998) used the same stimulus
materials but changed the task demands. Participants now performed a general
lexical decision task, responding yes if a word of either language was presented
(rather than saying no to Dutch words). In this experiment, frequencydependent facilitation eects were found for the interlingual homographs
(relative to English control words).
The authors argued that the null-results for interlingual homographs in the
rst experiment did not constitute conclusive evidence that bilingual word
recognition involves a language selective access process, because in that case the
dierent stimulus list composition of Experiment 2 should not have aected the
results. Instead, the results of Experiment 2 were considered as evidence
supporting the language non-selective access view, and this view was tested and
supported again under the dierent task demands of Experiment 3. This last
experiment further showed that task demands may aect the direction of the
observed eects: changing the task from language specic lexical decision in
Experiment 2 towards generalized lexical decision in Experiment 3 turned the
inhibition eects of Experiment 2 into facilitation eects in Experiment 3.
The null-eects in Experiment 1 make one wonder if the Dutch readings of
the interlingual homographs were activated at all in this English lexical decision
task. Recent reanalysis of the data suggests they were. A regression analysis
showed that (despite the over-all null results) homograph responses became
slower as the frequency of their Dutch reading increased, while they became
faster with increasingly high English frequency readings. Furthermore, De Moor
(1998) demonstrated that the L1 semantics of the interlingual homographs was
apparently activated as well. De Moor rst replicated the null-result for homographs relative to controls. Then, on the trial after the homograph appeared, she
presented the English translation of its Dutch reading. For instance, brand was
followed by re, which is the English translation of the Dutch word brand. A
small but reliable translation priming eect of 11 ms was found. In a replication
of this experiment with dierent stimulus materials, Van Heste (1999) observed
a reliable 35 ms dierence between translation and control trials. The Dutch
reading of the homograph on the previous trial had apparently been activated
even though this did not aect its RT (cf. De Bruijn, Dijkstra, Chwilla and
Schriefers 2001).
Finally, Dijkstra et al. (1999) performed an analogous experiment that was
reviewed in Section 2, also involving an English lexical decision task with
interlingual homographs and controls. In this study, signicant facilitation
eects were found for homographs having cross-linguistic overlap in orthography but not in phonology (stage), and no eects for items with overlap in both
(step). The items in this study were comparable to those in Dijkstra et al. (1998),
making it likely that the null-eects in the earlier study were due to mixing the
two types of items. (Indirect support for this reasoning comes from a Spanish
lexical decision study involving French-Spanish bilinguals by Font (2001: 115),
who found facilitatory eects for French-Spanish homographs that had little
phonological similarity across languages.)
Several other accounts have been proposed for the null-results in Experiment 1 and the inhibitory eects in Experiment 2 from Dijkstra et al. (1998).
These accounts have either referred to dierences in the relative activation of
words from the two languages in Experiments 1 and 2 (Dijkstra et al. 1998,
Grosjean 2001), or to dierences in participants decision strategies (De Groot et
al. 2000, Dijkstra et al. 2000). Let us take a closer look at the various proposals.
Dijkstra et al. (1998) assumed that the degree of activation of Dutch (the
non-target language) was higher in Experiment 2 than in Experiment 1, because
Dutch words were only included in Experiment 2. As a consequence, the
English readings of the interlingual homographs suered from more competition by the Dutch reading in Experiment 2. As an underlying mechanism, this
view assumes that lexical activation eects can last across trials and can aect
relative language activation. In sum, the dierent results in the two experiments
are assumed to be a consequence of dierent bottom-up activation processes
due to the composition of the stimulus list. This is basically an explanation in
terms of lexical context eects.
Similarly, Grosjean (2001) interpreted the results in terms of the participants language mode, referring to the relative state of activation of the
bilinguals languages and language processing mechanisms. The mode is
monolingual if only one language is relatively active and bilingual if both
languages are active (though one language may be more active than the other).
In Experiment 1, the participants only read English words and non-words
(although some words were homographs and cognates) and they were instructed to decide whether the items were English words or not. This would have
positioned them towards the monolingual end of the mode continuum, but
they did not reach this position totally as they knew they were being tested as
bilinguals. Thus, although their Dutch was partly active (which would explain
the cognate eect) it was not suciently active to create a homograph eect. In
sum, Grosjean (2001) proposed that both the participants expectancies with
respect to the English lexical decision task and the degree of language intermixing (encountering mostly English words) aected the bilinguals performance.
This explanation implies that both non-linguistic and linguistic context aspects
aected relative language activation.
De Groot et al. (2000) replicated the null-results observed in Experiment 1
by Dijkstra et al. (1998) using dierent stimulus materials and dierent DutchEnglish bilinguals. They proposed that the participants were instructed to
perform a language specic English lexical decision task, but on some trials
may instead have treated the task as a language neutral lexical decision task.
The adoption of a language specic processing mode would induce slower
responses to homographs than to matched controls due to lexical competition
between the activated target and non-target readings of the interlingual
homographs (just as in Experiment 2 by Dijkstra et al. 1998). In contrast, in a
language neutral processing mode the response to a homograph would be
based on the availability of any reading, irrespective of language, and homographs could then be responded to faster than controls (as in the generalized
lexical decision task of Experiment 3 by Dijkstra et al. 1998). In sum, a mixture
of the two processing modes adopted by the participants led to a mixture of
facilitation and inhibition eects for homographs relative to controls, yielding
an overall null-result. (Note that this account would predict larger standard
deviations for the homographs in the condition where the Dutch reading of the
homograph has a high frequency than in the conditions where it has a low
frequency, because in the former type of condition, the Dutch reading would be
available much faster than the English reading, while that would not be the case
for the last type of words.)
Dijkstra, De Bruijn, Schriefers and Ten Brinke (2000) pointed out that the
participants in the three studies that reported null-results were apparently not
told in advance that some of the presented letter strings would be words in both
Dutch and English. Participants might sometimes have adopted a language
neutral processing mode because they were in an uncertain situation. To
disentangle the eects of instruction and language intermixing, Dijkstra et al.
designed an experiment that combined features of Experiments 1 and 2 by
Dijkstra et al. (1998). Participants were explicitly instructed that they would
encounter Dutch words requiring a no response and were provided with
141
examples in the practice set. However, exclusively Dutch words were presented
only in the second part of the experiment. No signicant RT dierences were
found between the interlingual homographs and matched English control items
in the rst part of the experiment. In contrast, strong inhibitory eects for
homographs relative to control words appeared in the second part. Examination
of the transition from Part 1 to Part 2 showed that, as soon as Dutch items
started to come in, the RTs to homographs were considerably slowed down
compared to control words. These results converge completely with those of
Experiments 1 and 2 by Dijkstra et al. (1998) discussed above. They suggest that
language intermixing rather than instruction-based expectancies drives the
bilingual participants performance. Instead of interpreting the pattern of
results in Part 1 and Part 2 of the experiment as evidence for dierences in
relative language activation (depending on the local absence or presence of
Dutch items in the stimulus list), Dijkstra et al. propose that participants used
dierent decision criteria in the two parts of the experiment, depending on the
types of lexical items they encountered.
Ignoring the details of the proposed underlying mechanisms, we can draw
a number of general conclusions on the basis of these and other studies (see
Dijkstra and Van Heuven (2002), for an elaborated model of bilingual word
recognition based on this evidence). First, word candidates from both target
and non-target languages are activated in parallel in a bottom-up way (via the
signal), even though their rates of activation may dier depending on subjective
frequency. Second, stimulus list composition and task demands are important
determinants of the response patterns. Third, task demands, instruction details,
and other top-down information sources do not override the activated
bottom-up information; instead, the activated representations in the two
lexicons are used for responding in accordance with the requirements of the
task at hand. In all, the conclusion is that for isolated words presented in
stimulus lists, bilingual word recognition is based on the input signal and is
basically automatic. Non-linguistic context eects (due to the composition of
the stimulus list, the specic instruction, or the task to be performed) appear to
aect the decision criteria that are used to accept one lexical candidate or
another during the lexical selection process, but not to aect the relative degree
of activated word candidates from one or the other language.
information source) and the sentence constraint, and not of language membership and the sentence constraint. This suggests that (just like for isolated words)
lexical characteristics are more important than language characteristics in the
determination of word recognition in sentences.
Only a limited number of studies have investigated syntactic eects of
sentence context on word recognition in some detail (for a full review, see Kroll
and Dussias in press). Here we will briey describe a few recent studies that
used Event-Related brain Potentials (ERPs) to compare syntactic and semantic
aspects of sentence processing in bilinguals (Weber-Fox and Neville 1996,
Hahne 2001, Hahne and Friederici 2001). Like the authors of these studies, we
will argue that there are processing dierences between monolinguals and
bilinguals with respect to semantic aspects that appear to be especially quantitative in nature, while the dierences with respect to syntax appear to be qualitative as well.
In the study by Weber-Fox and Neville (1996), ve groups of ChineseEnglish bilinguals performed an acceptability judgment task for sentences in
their L2, English, while their EEG was recorded. These groups of participants
had learned English at dierent ages. Apart from normal control sentences, they
read semantically anomalous sentences in English (e.g. The scientist criticized
Maxs event of the theorem), sentences that contained violations of English
phrase structure rules (e.g. The scientist criticized Maxs of proof the theorem),
sentences that contained specicity constraint violations (e.g. What did the
scientist criticize Maxs proof of?), and sentences that contained subjacency
constraint violations (e.g. What was a proof of criticized by the scientist?). In
terms of their brain activity, early L2 learners (those who had learned English
before age 11) responded to the semantic anomalies in a very similar way as
monolingual language users. The other bilingual groups diered from the
monolinguals only quantitatively: an N400 eect, a marker associated with the
processing of semantic anomalies, was present in their EEGs, but it was delayed
in time relative to that in monolinguals.
In contrast, several qualitative dierences between prociency groups were
found with respect to the syntactic processing of phrase structure violations.
First, none of the bilingual groups displayed a so-called early left anterior
negativity (N125) in the EEG that was present in monolinguals. The N125 is an
early eect in the EEG that may reect automatized rst-pass parsing processes.
Second, a second left lateralized negativity (N300500) was found in all groups,
which was left lateralized (found in the left hemisphere of the brain) in monolinguals and early bilinguals, but more bilaterally distributed in late bilinguals.
Third, a P600 eect was present in the monolinguals and early bilinguals but
not in the late learners. The P600 eect is considered to be the most important
EEG marker associated with syntactic reanalysis and repair. In sum, late L2
learners consistently displayed large dierences in ERPs patterns relative to
monolinguals, suggesting that (especially late) syntactic processes are dierent
in late L2 learners.
Hahne (2001) came to similar conclusions on the basis of an auditory
sentence processing study involving procient late Russian-German bilinguals and
German monolinguals. Her participants listened to German sentences that were
either correct (e.g. Die Tr wurde geschlossen, The door was being closed),
contained a semantically incorrect item (selection restriction violation: Die Ozean
wurde geschlossen, The ocean was being closed), or a syntactically correct item
(word category violation: Das Geschft wurde am geschlossen, The shop was
being on closed). As before, ERP dierences in processing semantic incongruities between native and L2 speakers were only quantitative in nature, while
there were qualitative dierences with respect to syntactic processing between
the two participant groups. This suggests that the second language learners did
not process syntactic information in the way that native listeners did.
Hahne and Friederici (2001) examined sentence comprehension in Japanese speakers who had learned German as a second language after puberty.
These bilinguals listened to German sentences that were correct or contained
semantic and/or syntactic violations. A variety of dierences was found in the
ERPs for the Japanese-German bilinguals and German monolinguals. Semantically incorrect sentences induced an ERP pattern similar for the two groups (an
N400 eect), while correct sentences led to a dierent pattern (greater positivity) in L2 learners than in native listeners. The latter nding may reect the
greater diculties the learners had with respect to syntactic integration. For
sentences containing a phrase structure violation, L2 learners, in contrast to
native listeners, did not show signicant modulations of the syntax-related ERP
components mentioned above (the early anterior negativity and the P600).
Furthermore, sentences containing a pure semantic or combined syntactic/
semantic violation elicited eects not found in native listeners. These eects
may reect additional conceptual-semantic processing in late bilinguals.
These ERP studies indicate that future RT studies examining sentence
processing in bilinguals are likely to yield evidence for complex interactions
between lexical and syntactic knowledge in L1 and L2. Of course, there is a vast
number of research issues to be addressed. For a psycholinguist working on the
bilingual lexicon, one interesting issue to explore is to which extent the language
145
non-selective access mechanism found at the lexical level also holds at the
syntactic level.
The assumption that the syntactic rules and syntactic categories of dierent
languages are not incorporated in language-specic databases but in an
integrated store leads to a variety of predictions. For instance, for a DutchEnglish bilingual processing dierences could arise for a noun phrase like the
light of a distant star and a noun phrase like the man sat in his room, because
the interlingual homograph star is an adjective in Dutch (meaning rigid) but
a noun in English, while the homograph room is a noun in both languages
(which means cream in Dutch).
Furthermore, language non-selective access of syntactic rules might lead to
specic cross-linguistic priming eects. For instance, hearing a sentence like
the librarian handed the reader a book might prime the production of de vader
gaf het meisje een appel (The father gave the girl an apple) but not de vader gaf
een appel aan het meisje (The father gave an apple to the girl) (cf. Bock 1986).
Another interesting issue is whether there is a separate eect of the language
of the sentence context on the recognition of a target word. In other words,
could a noun phrase or larger sentence context elicit some kind of language
frame that aects the processing of later arriving words? In that case, processing
a sentence with a code-switch like I see a huis might be more dicult than
processing a regular sentence like I see a house, simply because the words in
the rst sentence context do not all belong to one and the same language.
Finally, it seems likely that we may be expecting some unexpected results in
future RT studies on bilingual syntactic processing, due to the complex interactions between lexical, syntactic, and semantic factors. One possibility is that
quantitative dierences in working memory capacity for L2 syntactic processing
may lead to qualitative processing consequences between L2 monolinguals and
bilinguals (cf. Michael, Dijkstra and Kroll 2002). For instance, in a pilot study
in our lab we found that although both Dutch and German readers tended to
resolve local ambiguities in subject- and object-relative clauses in their L1 by
using syntactic information only, Dutch-German readers in their L2 used
semantic information as well (Caelen 1998, but also see Frenck-Mestre and
Pynte 1997). A similar pattern was found in L1 readers under higher processing
load conditions.
All these exciting questions and many others are amenable to empirical
research by means of existing research techniques. Unfortunately, the collection
of empirical data addressing these questions has only just started and we have
no answers to these questions yet.
4. General conclusions
In this chapter, we have considered a number of questions about bilingual
lexical processing and provided answers based on the presently available
empirical evidence with respect to interlingual homographs and cognates. First,
we have argued that during the recognition of isolated words by bilinguals,
lexical candidates from several languages are activated in parallel. Such parallel
activation does not only hold for orthographic representations, but also for
phonological and semantic codes. Moreover, there is evidence that language
non-selective access occurs even when bilinguals are processing words in their
native language and are not aware that their second language knowledge is
important. These ndings indicate that the bilingual word identication system,
just like the monolingual system, is to a large extent automatic in nature, in
the sense that lexical candidates from both languages are activated in a fast
recognition process that in itself is largely unaected by intentional and
attentional factors.
Second, we have examined the time-course of lexical activation with respect
to L1 and L2 and found that L2 is slower to be activated than L1, depending on
relative L1/L2 prociency and therefore on (subjective) L1/L2 word frequency.
For interlingual homographs, we have found that in spite of dierences in
L1/L2 activation rates, both readings of interlingual homographs remain active
during lexical processing for a relatively long time. This nding has several
important theoretical consequences. For instance, if the language membership
of word candidates could be used quickly to suppress lexical candidates that are
irrelevant in the experimental context, eects of the non-targeted reading of the
homographs should quickly disappear. However, if language membership
information becomes available late during processing, both readings of the
homograph would remain active for quite long. The available empirical studies
support the latter position, which is in correspondence with the automatic
nature of bilingual word recognition.
Third, we have demonstrated that both non-linguistic experimental and
linguistic context factors may aect the result patterns that are observed in
experiments. It appears that non-linguistic factors such as task demands and
instruction aect the performance of bilingual participants at the level of task
and decision processes as well as participant strategies. Linguistic factors such
as lexical, syntactic, and semantic aspects of the sentence context appear to
aect the word identication process more directly. Evidence from ERP studies
indicates that syntactic processing in bilinguals may dier both quantitatively
<DEST "dyk-n*">
and qualitatively from that in monolinguals, but RT studies are badly needed in
order to specify from which underlying mechanisms the dierences originate.
To conclude, empirical studies on bilingual word recognition in the last
decade have uncovered a number of fundamental characteristics of the bilingual
word recognition system. They have answered some major questions that are
unique to the bilingual domain, such as that about language selective or nonselective access, as well as more generally important questions, such as how
language users handle lexical ambiguity and how task and stimulus context
aect word recognition. The conclusions of these studies will have to be taken
into account during the development of a more general model of bilingual
processing. However, much more empirical evidence on the interaction
between lexical, syntactic, and semantic processing is needed before we can
even attempt to build such a model.
Note
* The author thanks Folkert Kuiken and two anonymous reviewers for their comments on
a previous version of this paper. The author also thanks Judy Kroll for her continuous
support and the many discussions that shaped the ideas in this chapter.
References
Altarriba, J., Kroll, J. F., Sholl, A. and Rayner, K. 1996. The inuence of lexical and conceptual constraints on reading mixed-language sentences: Evidence from eye xations and
naming times. Memory & Cognition 24: 477492.
Altenberg, E. P. and Cairns, H. S. 1983. The eects of phonotactic constraints on lexical
processing in bilingual and monolingual studies. Journal of Verbal Learning and Verbal
Behavior 22: 174188.
Bock, J. K. 1986. Syntactic persistence in language production. Cognitive Psychology 18:
355387.
Caelen, M. 1998. Extending the study on the processing of relative clauses to bilingualism.
Unpublished Masters Thesis, University of Nijmegen.
Cristoanini, P., Kirsner, K. and Milech, D. 1986. Bilingual lexical representation: The status of
Spanish-English cognates. Quarterly Journal of Experimental Psychology 38A: 367393.
Davelaar, E., Coltheart, M., Besner, D. and Jonasson, J. T. 1978. Phonological recoding and
lexical access. Memory & Cognition 6: 391402.
De Bruijn, E., Dijkstra, A., Chwilla, D. and Schriefers, H. 2001. Language context eects on
interlingual homograph recognition: Evidence from event-related potentials and
response times in semantic priming. Bilingualism: Language and Cognition 4: 155168.
De Groot, A. M. B., Delmaar, P. and Lupker, S. J. 2000. The processing of interlexical homographs in a bilingual and a monolingual task: Support for nonselective access to
bilingual memory. Quarterly Journal of Experimental Psychology 53: 397428.
De Moor, W. 1998. Visuele woordherkenning bij tweetalige personen. [Visual word recognition
in bilinguals.] Unpublished Master Thesis, University of Ghent.
Dijkstra, A., De Bruijn, E., Schriefers, H. J. and Ten Brinke, S. 2000. More on interlingual
homograph recognition: Language intermixing versus explicitness of instruction.
Bilingualism: Language and Cognition 3: 6978.
Dijkstra, A., Grainger, J. and Van Heuven, W. J. B. 1999. Recognition of cognates and
interlingual homographs: The neglected role of phonology. Journal of Memory and
Language 41: 496518.
Dijkstra, A., Timmermans, M. and Schriefers, H. 2000. Cross-language eects on bilingual
homograph recognition. Journal of Memory and Language 42: 445464.
Dijkstra, A. and Van Heuven, W. J. B. 2002. The architecture of the bilingual word recognition system: From identication to decision. Bilingualism: Language and Cognition 5:
175197.
Dijkstra, A., Van Jaarsveld, H. and Ten Brinke, S. 1998. Interlingual homograph recognition: Eects of task demands and language intermixing. Bilingualism: Language and
Cognition 1: 5166.
Font, N. 2001. Rle de la langue dans laccs au lexique chez les bilingues: Inuence de la
proximit orthographique et smantique interlangue sur la reconnaissance visuelle de mots.
Unpublished Doctoral Thesis of the Universit Paul Valery, Montpellier, France.
Frenck-Mestre, C. and Pynte, J. 1997. Syntactic ambiguity resolution while reading in
second and native languages. Quarterly Journal of Experimental Psychology 50: 119148.
Gerard, L. D. and Scarborough, D. L. 1989. Language-specic lexical access of homographs
by bilinguals. Journal of Experimental Psychology: Learning, Memory and Cognition 15:
305313.
Gottlob, L. R., Goldinger, S. D., Stone, G. O. and Van Orden, G. C. 1999. Reading homographs: Orthographic, phonologic, and semantic dynamics. Journal of Experimental
Psychology: Human Perception and Performance 25: 561574.
Grainger, J. and Beauvillain, C. 1987. Language blocking and lexical access in bilinguals.
Quarterly Journal of Experimental Psychology 39A: 295319.
Grainger, J. and Dijkstra, A. 1996. Visual word recognition. In Computational Psycholinguistics: AI and connectionist models of human language processing, A. Dijkstra and K.
De Smedt (eds), 139165. London: Taylor and Francis.
Grosjean, F. 2001. The bilinguals language modes. In Language processing in the bilingual,
J. L. Nicol and T. D. Langendoen (eds), 125. Oxford: Blackwell.
Hahne, A. 2001. Whats dierent in second-language processing? Evidence from eventrelated brain potentials. Journal of Psycholinguistic Research 30: 251266.
Hahne, A. and Friederici, A. 2001. Processing a second language: late learners comprehension mechanisms as revealed by event-related brain potentials. Bilingualism: Language
and Cognition 4: 123141.
Kroll, J. F. and Dussias, P. E. In press. The comprehension of words and sentences in two
languages. Chapter to appear in Handbook of bilingualism, T. Bhatia and W. Ritchie
(eds). Cambridge, MA: Blackwell Publishers.
</TARGET "dyk">
Lemhfer, K. and Dijkstra, A. Submitted. Recognizing cognates and interlingual homographs: Time course and code similarity eects in generalized lexical decision.
Macnamara, J. and Kushnir, S. L. 1971. Linguistic independence of bilinguals: The input
switch. Journal of Verbal Learning and Verbal Behavior 10: 480487.
Michael, E., Dijkstra, A. and Kroll, J. F. 2002. Individual dierences in the degree of language
nonselectivity in uent bilinguals. Paper presented at the meeting of the International
Linguistic Association, Toronto, Canada.
Sebastin-Galls, N. and Kroll, J. F. In press. Phonology in bilingual language processing:
Acquisition, perception, and production. In Phonetics and phonology in language
comprehension and production: Dierences and similarities, N. Schiller and A. Meyer
(eds). Berlin: Mouton de Gruyter.
Van Hell, J. and Dijkstra, A. 2002. Foreign language knowledge can inuence native
language performance in exclusively native contexts. Psychonomic Bulletin and Review
9: 780789.
Van Heste, T. 1999. Visuele woordherkenning bij tweetaligen. [Visual word recognition in
bilinguals.] Unpublished Master Thesis, University of Leuven.
Van Heuven, W. J. B., Dijkstra, A. and Grainger, J. 1998. Orthographic neighborhood eects
in bilingual word recognition. Journal of Memory and Language, 39: 458483.
Van Heuven, W. J. B. and Dijkstra, A. April 1999. The role of phonology in the recognition of
interlingual homographs and cognates. Paper presented at the Second International
Symposium on Bilingualism, Newcastle, UK.
Von Studnitz, R. E. and Green, D. 2002. Interlingual homograph interference in GermanEnglish bilinguals: Its modulation and locus of control. Bilingualism: Language and
Cognition 5: 123.
Weber-Fox, C. M. and Neville, H. J. 1996. Maturational constraints on functional specializations for language processing: ERP and behavioral evidence in bilingual speakers.
Journal of Cognitive Neuroscience 8: 231256.
<TARGET "wil" DOCINFO AUTHOR "John N. Williams"TITLE "Inducing abstract linguistic representations"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 7
1.
Introduction
152
John N. Williams
that it could then indicate the gender for nouns that it had not received during
training, although its performance was not perfect (ranging between 73% and
75%). This indicates that there are regularities in the form (in this case spelling)
of French words which can to a certain extent predict gender category.
Yet there are always words which fall stubbornly outside such generalisations. In the case of French, Carroll (2001) argues that in any case, the kinds of
phonological cues that have been appealed to are more subtle than could
reasonably be expected to be represented in the lexicon. This is not to say that
phonological and semantic cues do not play a role in learning gender systems,
or that they do not aect how easy it is to remember the gender of specic
words. But ultimately gender classes impose an abstract categorisation on words
which is independent of their phonological and semantic properties. Learning
gender systems, then, requires the formation of abstract grammatical categories,
and producing grammatically well-formed utterances involves applying
agreement rules which make reference to those categories.
There is a growing body of evidence which suggests that even quite advanced second language learners continue to make gender errors (Hawkins
2001, Holmes and De la Batie 1999). In contrast, such errors are relatively rare
in rst language acquisition (Caselli, Leonard, Volterra and Campagnoli 1993).
There is also evidence for qualitative dierences between rst and second
language acquisition and processing of gender. A number of studies have shown
that second language learners are more sensitive to phonological agreement
patterns that correlate with gender classes than either children or adults in their
native language. For example, for the Italian il pettine (the comb, masculine
singular) a second language learner might produce *le pettine, using the article
which is more often associated with the -e ending on feminine plural nouns
(Holmes and De la Batie 1999). In contrast a child would be more likely to
produce *il pettino, choosing an article that is correct for the nouns gender and
number, and providing the noun with the characteristic -o ending for masculine singulars. This demonstrates a grasp of the nouns abstract gender as the
controlling inuence in determiner selection (Caselli et al. 1993). In reaction
time tasks on adults, Taraban and Kempe (1999) showed that non-native
speakers of Russian are more sensitive to phonological cues to gender than are
natives. Finally, a study by Guillelmon and Grosjean (2001) showed that
whereas native speakers of French and early bilinguals show certain gender
congruency eects in reaction time tasks, such eects are absent in late bilinguals. These studies suggest that second language learners do not achieve
native-like representation or processing of gender information.
In this chapter I shall explore the possibility that the reason why gender is
a persistent problem for second language learners is precisely because the
underlying abstract grammatical concepts are dicult to acquire through
associative learning. I shall address this issue through behavioural studies of
semi-articial language learning in tandem with computational (connectionist)
simulations. These simulations were used as a means of assessing the viability,
and potential limitations, of a purely associative learning account of the
behavioural data.
153
154
John N. Williams
sequences which violated phrase structure). Because no such test was performed
it is dicult to know whether abstract word classes had really been learned.
More stringent tests of word class learning become possible when noun
classes, such as gender systems, are considered. Brooks, Braine, Catalano and
Brody (1993) used an articial language in which there were two noun classes,
and each class used dierent axes to mark the location of the actor in relation
to the object denoted by the noun. Neither the form nor meaning of the nouns
provided any clue to their class. Adults were rst taught the vocabulary, and
then performed both comprehension and production tasks (e.g. acting out
phrases, or describing pictures with feedback in the form of the correct answer).
After training they were tested on knowledge of the trained items, and also on
their ability to produce the correct response for noun-ax combinations that
had not been presented during training. Whilst their performance on trained
items was at around 75%, they were at chance on the generalisation items. Not
one of the 16 subjects showed evidence of having learned the system. Similar
results have been obtained in a number of other studies (Braine 1987, Braine et
al. 1990, Frigo and McDonald 1998). Frigo and McDonald (1998: 237) argue
that models of noun class learning that depend on pure distributional analysis
(Anderson 1983, Maratsos and Chalkley 1980, Pinker 1984) are too powerful.
The question is, then, does connectionism fall into this class of overly
powerful learning mechanisms for learning noun classes? The experiments and
simulations presented below further explored the circumstances under which
arbitrary and non-arbitrary noun class systems can be learned by humans and
connectionist networks.
4. Experiment 1
Williams and Lovatt (2003) tested whether humans can learn the arbitrary noun
class system shown in Table 1. There were eight nouns divided into two arbitrary classes masculine and feminine. Words in the masculine class occurred
with the determiners ig, i, ul, and tei. Words in the feminine class occurred
with the determiners ga, ge, ula, and tegge.1 The training items were the nonitalicised phrases shown in Table 1. The italicised items were withheld for
testing generalisation. It would only be possible to know that the ball should
be translated as ig johombe by knowing that johombe belongs to the masculine
class. Neither its form, its -e ending, nor its meaning provide any clues.
The participants rst learned the nouns and determiners as isolated
vocabulary items. They then received the determiner-noun combinations for
155
Table 1.The items employed in Experiments 1 and 2. Items used for testing generalisation are in italics.
denite singular
(the)
denite plural
(the)
indenite singular
(a)
indenite plural
(some)
masculine
ball
house
ght
bird
ig johombe
ig zabide
ig wakime
ig migene
i johombi
i zabidi
i wakimi
i migeni
ul johombe
ul zabide
ul wakime
ul migene
tei johombi
tei zabidi
tei wakimi
tei migeni
feminine
shoe
kiss
cake
nose
ga shosane
ga tisseke
ga chakume
ga nawase
ge shosani
ge tisseki
ge chakumi
ge nawasi
ula shosane
ula tisseke
ula chakume
ula nawase
tegge shosani
tegge tisseki
tegge chakumi
tegge nawasi
157
158
John N. Williams
train the network in a way which resembled as closely as possible the training
task performed by the participants in Experiment 1. I decided to focus on the
recall component of the training task. The network was taught to produce the
correct determiner for each phrase in the training set shown in Table 1. The
input consisted of representations of the noun stem, the inection, the English
determiner, and the number of the noun. For example, the input for the item
tei johombi was johomb, -i, some, and plural. This is the information that
is relevant to predicting the determiner, and which was explicitly provided to
the participants in the recall component of the training task in Experiment 1.2
Following Elman (1990) one unique input node was used to represent each
element of the input (for example one unit was used to represent johomb),
yielding a total of 15 input nodes (eight stems, two inections, three English
determiners, singular, plural). The input nodes were connected to ve hidden
units, which were in turn connected to eight output units, one for each of the
eight possible determiners. For each input pattern the network was taught to
produce the correct determiner. For example given the input johomb, -i, some,
and plural it was taught to predict tei. This involved comparing the actual
output from the network with the correct output, and making appropriate
changes to the connection weights within the network according to the degree
of error. In this sense the network was provided feedback in the same way as the
participants in the experiment.
The network was initially trained until the root mean square (RMS) error
for the training items was 0.1 (this required an average of 2,479 cycles through
the training set).3 An error of this magnitude indicated that for each input
pattern the network was able to activate the correct determiner on the output
layer to a value close to the target value of 1.0, and all other output units had
values close to zero. Testing involved presenting the input patterns for the
generalisation items in Table 1 (i.e. the network was presented with patterns that it
had not received during training). For each input pattern, the activation level of
the output units was recorded, compared to the correct answer and the degree of
error calculated. The training and test procedure was repeated 20 times, and on
each run the connection weights were given random starting values.
Generalisation performance on each run was perfect in the sense that the
activation on the node for the correct determiner was far greater than that of
the others. Over 20 runs the mean RMS error was 0.118 (which is not much
greater than that for trained items). That is, the network was able to correctly
predict the determiner for input patterns that it had never encountered during
training with an accuracy which was almost as high as for the trained items.4
5. Experiment 2
This experiment employed a training task that was thought to be unlikely to
induce an intentional learning strategy. Participants rst performed the same
phonological short-term memory test and vocabulary learning exercise as in
Experiment 1. Determiner-noun combinations from the training set were then
auditorily presented in a semi-random sequence, avoiding immediate repetitions of the same noun or determiner. For each item the participants had to
perform the following tasks: (1) repeat the phrase aloud, (2) indicate whether
it refers to a living or non-living thing by pressing one of two response keys, and
(3) translate the phrase into English. For example, for the item ul johombe they
would respond by saying ul johombe, pressing the non-living key, and saying
a ball. The meanings of the words were altered so that half of the nouns in each
class referred to living things and half to non-living things. The living/nonliving decision was included because this experiment was also a control for a
subsequent version in which noun animacy predicted noun class membership
(see Experiment 3 below). Here it serves as a means of increasing task demands
so that participants would be less likely to attempt to engage explicit learning
processes. The participants were told that the purpose of the experiment was to
see how their decision and translation performance improved with practice and
so they were encouraged to make their responses as quickly and as accurately as
possible. Training extended over 15 cycles through the 24-item training set,
giving a total of 360 training trials. This took between 60 and 75 minutes
including rest breaks after every ve cycles.
The training phase was followed by the test phase. On each trial the English
translation of a test phrase was visually presented (e.g. the ball) and the
participants had to choose between a grammatical and ungrammatical translation in the target language, where the determiner for the ungrammatical item
was always of the correct number and deniteness, but the incorrect gender
(e.g. ig johombe versus ga johombe for the ball). First the eight generalisation
items were presented (see Table 1) followed by 16 trained items.
There were 18 participants who were selected on the basis of their good
knowledge of gender languages so as to increase the potential for obtaining
learning in this experiment. They all rated themselves as intermediate or better
in at least two gender languages (mean = 2.8, range = 2 to 6). Twelve of the
participants spoke a gender L1. Using the same scale for assessing knowledge of
gender languages as employed by Williams and Lovatt (2003) they scored 5.8,
which is much higher than the mean of 2.6 for the participants in Experiment 1.
Their phonological short-term memory was also somewhat superior, the mean
score being 71% as opposed to 64%.
None of the participants were aware of the noun class system either during
training or test phases. The average percentage correct on the generalisation
items was 56%, which was not signicantly dierent from the chance level of
50%, t = 1.34, p > 0.1. On the other hand, performance on the trained items was
69%, which is signicantly better than chance, t = 5.53, p < 0.001, and signicantly better than performance on generalisation items, t = 2.58, p < 0.05. Thus,
although the participants had quite good memory for trained items, there was
no evidence of learning the underlying noun class distinction. This conclusion
is emphasised by the fact that the ten participants who scored 75% or better on
the trained items (mean = 80%) had a mean generalisation score of 50%. Nor
were there any correlations between generalisation test performance and either
phonological short-term memory or language background, and participants
who spoke a gender L1 did no better than those that did not (generalisation test
scores were 56% for both groups).
Given the failure to obtain learning in this experiment one may conclude
that Simulation 1 was in fact too powerful, and that the learning that occurred
in Experiment 1 was a result of purely explicit processes which fall outside the
scope of the model. However, there is an alternative possibility. We should also
consider the relationship between the task performed by Simulation 1 and the
tasks performed by the participants in Experiments 1 and 2. Simulation 1 was
intended as a model of the recall component of the training task used in
Experiment 1. But in Experiment 2 the participants task was very dierent.
They did not have to generate any determiners at any point during training, but
only had to perform animacy decisions and produce English translations.
Simulation 1 could not be said to be a good model of this task. A second
simulation was therefore conducted that made dierent assumptions about the
learning task.
5.1 Simulation 2
Incidental learning is best regarded as a relatively passive process of recording
correlations between attended features in each experience. Cleeremans and
161
Jimnez (2002), following OReilly and Munakata (2000: 18), have referred to
this as model learning, the goal of which is to enable the cognitive system to
develop useful, informative models of the world by capturing its correlational
structure. Connectionist models of model learning do not require feedback
because the system merely attempts to represent the structure of the inputs it is
provided. This is in contrast to task learning which has the aim of mastering
specic input-output mappings (i.e. achieving specic goals) in the context of
specic tasks through error-correcting learning procedures (ibid. p. 18).
Crucially for present purposes they assume that model learning operates
continuously, regardless of the task. Simulation 1 instantiated task learning, and
was successful because the underlying noun class distinction happened to be
relevant to the task the network was required to perform. But in Experiment 2
the tasks that the participants were performing (animacy decisions and translation) exerted no pressure to learn the noun class distinction. The same would
be true of simulations of those tasks. The only way in which the noun class
distinction could be learned, therefore, would be through model learning,
which requires a dierent kind of network from that used in Simulation 1.
One way of instantiating model learning is to train a three-layer network to
associate each input to itself. That is, the network learns to reproduce the input
pattern on the output layer. These are called autoassociation networks
(Plunkett and Elman 1997). Because there are fewer hidden than input/output
units the network is forced to discover an economical means of representing the
patterns so that they can be reproduced on the output. This gives the network
the potential to extract generalisations. Autoassociation networks do not
require feedback because the input itself provides the reference point against
which the accuracy of the output can be judged. How does such a network fare
on the arbitrary noun class induction problem?
In Simulation 2 there were 31 input units representing the eight determiners, eight stems, two inections, three English determiners, eight English
nouns, and units for singular and plural. All of the relevant information in a
training item such as ul johombe, a ball was represented as a pattern over the
input layer. The 31 output units represented the same information as the input
units. The network had 20 hidden units.5 For each item in the training set the
network was trained to reproduce the input pattern on the output layer.
Training continued until output error ceased to decline (which was after about
2,500 cycles).
In Experiment 2, learning was assessed by forcing participants to choose
between two translations for a phrase, for example, between ga johombe and ig
johombe as translations of the ball. The model can be tested in the same way by
presenting both grammatical and ungrammatical determiner-noun combinations and comparing the strength of the output on the determiner units. For a
trained item, such as ul johombe, the strength of activation of the corresponding
determiner in the output, in this case ul, was, as one would expect, very high
(0.996 when averaged over eight training items on ve separate runs, where the
required activation level was 1.0). Ungrammatical items such as ula johombe
produced much weaker activation of the corresponding output determiner
node, in this case ula (0.214). Clearly the network had not simply learned to
reproduce input patterns on the output layer. Rather, its ability to do so was
aected by whether it had received those patterns during training. In human
terms this would be the equivalent of a greater feeling of familiarity for ul
johombe than ula johombe. But for generalisation items the output activation on
determiners in both grammatical and ungrammatical items, e.g. ig johombe
versus ga johombe, was very low and not signicantly dierent (0.054 and 0.055
respectively). In other words, both items appeared equally unfamiliar to the
network. Therefore, like the human participants in Experiment 2, the autoassociation network had good memory for trained items, but was unable to
distinguish between grammatical and ungrammatical generalisation items.
The contrast between Simulations 1 and 2 demonstrates that task learning
enabled a connectionist network to become sensitive to an abstract noun class
distinction whereas model learning did not. This is a rather surprising result
when one considers that there is a sense in which the networks were performing
rather similar tasks. In both cases they had to remember which determiners
occurred with which congurations of noun, deniteness, and number in the
training items. The dierence was that in Simulation 1 the networks resources
were focused on predicting the determiner from the cues that it was provided,
whereas in Simulation 2 the network was actually attempting to remember the
unique combination of determiner, noun, deniteness, and number that
occurred in each training item. This exercise in episodic memory for entire
training episodes apparently did not exert sucient pressure on the network to
discover the underlying noun class distinction.
The contrast between task learning and model learning is reminiscent of the
procedural-declarative distinction in Andersons ACT framework (Anderson
1983). Productions are sets of rules which match their IF conditions against
the current contents of working memory, and if these are satised, they THEN
produce some action, or deposit some other kind of representation in working
memory. Although stated in a symbolic formalism in ACT, a connectionist
network can be conceived as a subsymbolic model of the entire set of productions which perform the transformation between one type of input to another
type of output (Sun et al. 2001). Both procedural learning and connectionist
learning of this type depends on error correction. In contrast, Simulation 2
could be identied with the declarative memory component of the ACT
framework. The idea that these two kinds of memory system might have
dierential power to extract generalisations from the environment is clearly
relevant to attempts to construct a theory of second language acquisition in
terms of their interaction (Towell and Hawkins 1994), or to identify them with
dierent brain regions (Ullman 2001). Indeed, the idea that procedural learning
is more powerful at extracting abstract linguistic rules would be consistent with
the proposal that such a mechanism supports rst language acquisition,
whereas second language acquisition is supported by declarative learning
(Ullman 2001).
If Simulation 1 is accepted as a valid model of the learning process in
Experiment 1 then there is another interesting consequence. The learning that
was occurring in that experiment was characterised as explicit. Not only did
the participants appear to have an intention to learn, but some of them also
made comparisons between consciously recalled input items, and formed
conscious hypotheses. Simulation 1 captures the intentional component of the
learning process, since it too evaluated its outputs with respect to feedback for
the purpose of learning in order to be able to generate determiners. But it
obviously does not model the other components of what, in human terms, we
regard as explicit learning. However, this does not necessarily detract from the
relevance of the model. Shanks (1995) reviews a range of studies on human
learning where there is a good t between human behaviour and connectionist
models. Yet in many of these experiments the participants were actively
searching for rules. For example, in a medical diagnosis task (see Shanks
1995: 42) participants were presented with hypothetical patients with certain
symptoms and were instructed to diagnose what illness each patient had. Each
trial was accompanied by feedback in the form of the correct diagnosis. Performance was directly related to the degree of contingency between dierent cues
(symptoms) and outcomes (diseases) in the training data. Shanks showed that
the results could be adequately modelled by a simple connectionist network in
which symptoms were presented as inputs, diagnoses as outputs, and the correct
diagnosis was provided as feedback (Shanks 1995:120). Yet the participants in the
experiment presumably had the experience of actively trying to work out the
relationship between symptoms and diseases. Whilst it is presently unclear how the
conscious states of the learner inuence the learning mechanism, it should not
be assumed that the possibility of there being such interactions rules out a
unied associative explanation (Cleeremans and Jimnez 2002).
6. Experiment 3
With the benet of hindsight Experiment 2 was a poor test of implicit learning
because the kind of associative learning mechanism supposed to underlie
implicit and incidental learning would not be expected to learn the underlying
generalisations. We6 therefore decided to run Experiment 2 again, but this time
using a system that would be learnable even by the kind of autoassociation
network used in Simulation 2. The language was essentially the same as that
shown in Table 1 except that the meanings of the words were altered so that all
of the words in class I referred to living things and all of the words in class II
referred to inanimate objects. For simplicity, the living/non-living distinction
will be referred to here in terms of an animacy cue to noun class. It has been
shown that under intentional learning conditions humans have no problem
grasping semantically-based noun classes (Braine 1987, Carroll 1999). A version
of Simulation 2 that included animacy information conrmed that the present
system was also learnable by an autoassociation network. This is presumably
because there are direct associations between the units that encode animacy and
certain determiners. Note, therefore, that in this experiment we are no longer
concerned with whether implicit learning of abstract noun classes is possible.
Rather the issue is whether implicit learning of a noun class distinction can be
obtained under conditions where the connectionist model predicts that there
should be an eect.
The tasks and procedure were exactly the same as in Experiment 2. For each
phrase presented during training the participants had to repeat it, indicate
whether it referred to a living or nonliving thing, and translate it into English.
Note that this time the living/nonliving decision coincided with the noun class
of the word. The same learning tests were used as in Experiment 2. There were
37 participants with varied language backgrounds.
Only seven of the participants became aware of the noun class distinction
and its relation to animacy during the training phase, and their performance
was perfect, or near perfect, on the generalisation and trained items. None of
the remaining 30 participants became aware of the system during the training
phase and none of them claimed to have been consciously trying to work out
the system during the generalisation test. Even at the end of the whole testing
phase none of them realised the relevance of animacy. Nevertheless, performance on generalisation items was 61%, which was signicantly above the
chance level of 50%, t = 3.25, p < 0.01. They scored 71% correct on trained
items, which was also signicantly above chance, t = 6.09, p < 0.001. Therefore,
Experiment 3 succeeded in demonstrating at least some degree of implicit
learning of a system that was also learnable by an autoassociation network.
However, there were large individual dierences in generalisation test
performance. Just as in Experiment 1 there were correlations with phonological
short-term memory (r = 0.50, p < 0.01) and knowledge of gender languages
(r = 0.586, p < 0.001), which in this case was quantied simply in terms of the
number of gender languages in which the participants rated their prociency as
intermediate or better (mean = 1.8, range = 0 to 5). We also evaluated whether,
amongst the 30 unaware participants, speakers of gender L1s did better than
speakers of non-gender L1s. For the 13 speakers of gender L1s mean generalisation test performance was 71%, which is signicantly above chance, t = 4.08,
p < 0.01, whereas for the 17 speakers of non-gender L1s it was 54%, which is not
signicantly above chance, t = 0.96. The dierence between these two groups
was signicant, t = 2.78, p < 0.01. The two groups did not dier signicantly in
terms of the number of L2s spoken to an intermediate level or better (3.54 and
3.12 respectively, t < 0.92), the number of gender languages known as an L2 (the
means were 1.46 and 1.23 respectively), but they did dier slightly in terms of
phonological short term memory (77% versus 68%, p = 0.08). Better matched
groups resulted from removing the three participants with the lowest phonological short term memory scores from the sample (all scores were less than 50%,
and all three participants were in the non-gender L1 group). The 13 speakers of
gender L1s and remaining 14 speakers of non-gender L1s were well matched in
terms of number of gender languages spoken as an L2 (1.46 and 1.43 respectively) and in terms of phonological short term memory scores (77% and 73%). Yet
the generalisation scores were 71% and 55% (the dierence being signicant,
t = 2.31, p < 0.05). Note that for the non-gender L1 group the mean for the
trained test items was well above chance (67%, t = 3.74, p < 0.01).
7. Discussion
In one sense it could be argued that there is a good alignment between the
connectionist models and the human data in the present studies, provided
that form and meaning are unitised at encoding; learners must actually perceive
them as being relevant to each other. Merely paying attention to the relevant
elements does not appear to be sucient, at least not under the task conditions
of Experiment 3. In terms of the model learning mechanism instantiated by
Simulation 2 this means that even though animacy information was attended,
it did not enter the same memory trace as information about the determiner,
noun identity, deniteness, and number. The problem is, therefore, to explain
why participants who spoke a gender L1 deed this principle and were able to
unconsciously associate the determiners with animacy information. There is no
obvious connectionist answer to this problem. Is the classical linguistic approach any more promising?
Linguistic (Carroll 1989, Hawkins 2001) and psychological (Levelt, Roelofs
ans Meyer 1999, Vigliocco, Antonini and Garrett 1997) analyses of gender
representation and processing in the L1 assume abstract gender features that are
attached to nouns in the lexicon. How gender features are acquired is not often
considered. However, Carroll (2001) proposes an induction procedure which
is triggered by the presence of alternating determiner forms in the input (e.g.
two words for some). The rst occurrence of one of the determiners, for
example in tei johombi (some monkeys) has no eect. But when another
phrase involving a word for some is encountered, for example tegge nawasi
(some vases), the learner seeks to rationalise the contrast by marking the noun
with a [+gender] feature. In this way, one of the determiners becomes an
assigner of the gender feature whilst the other remains the default. Remembering which of the alternating pair of determiners assigns the gender feature is
likely to be problematic, however. In the (admittedly articial) case that
[+gender] also corresponds to some other active feature of the noun, such as
[+inanimate], one can imagine that this problem would be alleviated. To
account for the inuence of gender L1s in Experiment 3 it would have to be
assumed that this kind of induction mechanism can only operate in L2 if it was
used in the L1. This is perhaps not too implausible if one considers that each
time a speaker of a gender language encounters a novel noun the same process
of using the accompanying determiner to assign gender to it must operate. On
the other hand, it is another matter to assume that, when confronted with a new
language, learners are able to assign new gender features on the basis of newly
observed alternations between determiners. It is also relevant to consider that
at present there is no evidence that speakers of gender L1s have any less
problem with gender in an L2 than do speakers of non-gender L1s (Bruhn and
White 2000). Thus, although the gender L1 advantage found in Experiment 3
Only in one of three experiments of Frigo and McDonald (1998) was performance on unmarked generalisation items signicantly above chance, and this
was when word class was indicated by a characteristic initial and nal syllable
(e.g. wanersumglot, wanolovglot, wanalglot versus kaisalmrish, kaisilvrish,
kaisalbrish). Braine (1987) also obtained good generalisation to unmarked
words, but half of the nouns in one class referred to males and the other half to
females. Thus, generalisation appears to be limited to cases where the cues are
more salient than in natural languages.
Somewhat counter-intuitively, where the above studies did nd evidence of
generalisation to unmarked items was when entirely novel nouns were introduced in the nal test phase. The equivalent test in the context of the language
used here would involve telling participants that ul vark means a dog and
asking them to produce the translation of the dog (the correct answer being ig
vark). Such a test only requires knowledge of the associations between the
determiners. Therefore, it does appear that partial phonological cues can
facilitate acquisition of inter-determiner associations (or rather, their equivalent
in the languages that were used). Determiners in the same class presumably
become associated by virtue of their frequent association to the same phonological cue. Generalisation is then achieved by a process of inference from another
determiner-noun combination that is provided at test or recalled from memory.
As argued by Frigo and McDonald (1998), poor performance on generalisation
tests involving nouns that occurred in training could be because of problems
recalling an example of a determiner that occurred with that noun. But the
native speaker of a gender language is assumed to generate an appropriate
determiner directly on the basis of an abstract specication of the nouns
gender in the lexicon, not by inference. It is far from clear that the participants
in these experiments acquired knowledge of noun classes in that sense.
The results from these studies do not, therefore, oer much prospect of
incidental learning of noun classes. This is of course consistent with the claim
that gender is a persistent problem for second language learners. Assuming an
underlying model learning mechanism such as that in Simulation 2, learning
would be predicted to be limited to rote storage of determiner-noun combinations, and associations between determiners and partial phonological and
semantic cues. This would explain L2 learners sensitivity to phonological cues
in gender processing tasks (Guillelmon and Grosjean 2001, Holmes and De la
Batie 1999, Taraban and Kempe 1999). Unmarked nouns would have to be
dealt with through rote storage, putting a strain on phonological memory
(Williams and Lovatt 2003). The lack of a true underlying noun class organisation
Notes
1. This language was derived from Italian. The determiners were derived from the Italian il,
i, un, dei, la, le, una, and delle by systematically substituting consonants (lg, dt, nl).
The nouns correspond to Italian nouns which end in -e in the singular and -i in the plural
regardless of gender, e.g. cliente (masculine), stazione (feminine). Note that none of the
participants in Experiments 1 and 2 (reported below) had any knowledge of Italian, and only
two participants in Experiment 3 knew Italian at an intermediate level or better as an L2.
2. The only dierence was that in the experiment they also had to produce the inection,
whereas in the simulation the inection was provided on the input. However, in the
experiment the participants learned the correct plural inections in the preliminary
vocabulary learning phase, and not in the training phase of the main experiment. In any case
the inection provides no clue as to the correct determiner over and above the presence or
absence of the plurality of the noun.
3. A Root Mean Square error of 0.1 means that over all of the input patterns presented on a
particular cycle the average dierence between the actual output and the required output on
each node was 0.1 units of activation. The point at which the correct output node was simply
the most active occurred well before an RMS error of 0.1 was achieved.
4. The Luce ratio was also used as a measure of network performance the activation level
of the correct output node divided by the sum of the activation over all output nodes. Perfect
output would be indicated by a Luce ratio of 1.0. In this simulation the mean Luce ratio over
20 runs was 0.87.
5. The number of hidden units was set to about two thirds of the number of input/output
units so as to force the inputs through a reduced representational space, exerting pressure on
the network to extract generalisations. Other simulations were performed with either 10 or
40 hidden units but the generalisation performance was similar to that reported here.
6. This experiment was run in collaboration with Helen East.
171
References
Anderson, J. R. 1983. The architecture of cognition. Cambridge MA: Harvard University Press.
Berent, I., Marcus, G. F., Shimron, J. and Gafos, A. I. 2002. The scope of linguistic generalizations: Evidence from Hebrew word formation. Cognition 83: 113139.
Braine, M. D. S. 1987. What is learned in acquiring word classes: A step towards an acquisition theory. In Mechanisms of language acquisition, B. MacWhinney (ed.), 6587.
Hillsdale, NJ: Lawrence Erlbaum.
Braine, M. D. S., Brody, R. E., Brooks, P. D., Sudhalter, V., Ross, J. E., Catalano, L. and Fisch,
S. M. 1990. Exploring language acquisition in children with a miniature articial
language: Eects of item and pattern frequency, arbitrary subclasses, and correction.
Journal of Memory and Langage 29: 591610.
Brooks, P J., Braine, M.D S., Catalano, L. and Brody, R. 1993. Acquisition of gender-like
noun classes in an articial language: The contribution of phonological markers to
learning. Journal of Memory and Language 32: 7695.
Bruhn, J. and White, L. 2000. L2 acquisition of Spanish DPs: the status of grammatical
features. In Proceedings of the 24th annual Boston University conference on language
development. Vol. 1, S. C. Howell, S. A. Fish and T. Keith-Lucas (eds), 164175. Somerville, Mass.: Cascadilla Press.
Carroll, S. 1989. Second language acquisition and the computational paradigm. Language
Learning 39: 535594.
Carroll, S. E. 1999. Input and SLA: Adults sensitivity to dierent sorts of cues to French
gender. Language Learning 49: 3792.
Carroll, S. E. 2001. Input and evidence: The raw material of second language acquisition.
Amsterdam: John Benjamins.
Caselli, M. C., Leonard, L. B., Volterra, V. and Campagnoli, M. G. 1993. Toward mastery of
Italian morphology: A cross-sectional study. Journal of Child Language 20: 377393.
Cleeremans, A. and Jimnez, L. 2002. Implicit learning and consciousness: A graded,
dynamic perspective. In Implicit learning and consciousness, R. M. French and A.
Cleeremans (eds), 140. Hove: Psychology Press.
Corbett, G. 1991. Gender. Cambridge: Cambridge University Press.
Ellis, N. C. 1998. Emergentism, connectionism and language learning. Language Learning
48: 631664.
Elman, J. L. 1990. Finding structure in time. Cognitive Science 14: 179211.
Fodor, J. A. and Pylyshyn, Z. W. 1988. Connectionism and cognitive architecture: A critical
analysis. Cognition 28: 371.
Frigo, L. and McDonald, J. L. 1998. Properties of phonological markers that aect the
acquisition of gender-like subclasses. Journal of Memory and Language 39: 218245.
Gmez, R. L. and Gerken, L. 2000. Infant articial language learning and language acquisition. Trends in Cognitive Sciences 4: 178186.
Guillelmon, D. and Grosjean, F. 2001. The gender marking eect in spoken word recognition: The case of bilinguals. Memory and Cognition 29: 503511.
Hawkins, R. 2001. Second language syntax: A generative introduction. Oxford: Blackwell.
173
</TARGET "wil">
Sokolok, M. E. and Smith, M. E. 1992. Assignment of gender to French nouns in primary and
secondary language: A connectionist model. Second Language Research 8: 3958.
Sun, R., Merrill, E. and Peterson, T. 2001. From implicit skills to explicit knowledge: a
bottom-up model of skill learning. Cognitive Science 25: 203244.
Taraban, R. and Kempe, V. 1999. Gender processing in native and nonnative Russian
speakers. Applied Psycholinguistics 20: 119148.
Tomasello, M. 2000. The item-based nature of childrens early syntactic development.
Trends in Cognitive Sciences 4: 156163.
Towell, R., and Hawkins, R. 1994. Approaches to second language acquisition. Clevedon:
Multilingual Matters.
Ullman, M. T. 2001. The neural basis of lexicon and grammar in rst and second language:
The declarative/procedural model. Bilingualism: Language and Cognition 4: 105122.
Vigliocco, G., Antonini, T. and Garrett, M. F. 1997. Grammatical gender is on the tip of
Italian tongues. Psychological Science 84: 314317.
Williams, J. N. In preparation. Implicit learning of form-meaning connections.
Williams, J. N. and Lovatt, P. 2003. Phonological memory and rule learning. Language
Learning 53: 67121.
<LINK "sab-n*">
<TARGET "sab" DOCINFO AUTHOR "Laura Sabourin and Marco Haverkort"TITLE "Neural substrates of representation and processing of a second language"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 8
1.
Introduction
1993, Kolk 2002, among others) that aphasics exhibit syntactic and semantic
priming eects. In a syntactic priming task, unimpaired subjects are quicker in a
lexical decision task if the target is a word that syntactically ts into the sequence of
words heard or read up to the point of presentation of the target. It reects the fact
that language users have clear expectations about what syntactic category is to
come next. For aphasics, these eects also show up, but only with stimulus onset
asynchronies (SOA) that are larger than the optimal SOA for unimpaired subjects:
whereas the optimal SOA for unimpaired subjects is 300 ms (with longer SOAs the
eect gradually disappears), the optimal SOA for the aphasics is much larger.
Haarman (1993) presents data from a syntactic priming study. He compared sentences such as those in (1) and found a priming eect for an unimpaired control group of about 65 ms on the last word: if that word t the
syntactic context, the unimpaired control group made the lexical decision 65 ms
quicker than if it did not. The agrammatic patients showed the same priming
eect (a quicker response to words that t the syntactic context), but only when
the SOA was increased from 300 (normals) to 1100 ms.
(1) a.
The fact that the aphasics showed a syntactic priming eect can only be
explained under the assumption that they have the relevant knowledge (regarding phrase structure and subcategorization) at their disposal and hence have
syntactic expectations as to what word class the next word will be; otherwise, no
eect should be found. The fact that the optimal SOA is a little over three times
as large for the aphasic population as for the control group, however, indicates
that the aphasics cannot make use of the relevant knowledge quick enough online; as soon as they are given more time, the exact same eect shows up as for
the unimpaired population. However, for the unimpaired control group, the
priming eect disappeared when items were presented at longer SOAs.
A similar priming eect has been shown to exist in ller-gap dependencies,
using semantic priming. Burkhardt et al. (2001), using sentences with moved
wh-phrases and DPs, presented semantically related or unrelated words
(Examples 2 and 3 below) at the trace or 600 ms after the trace in object
position (indicated by ti).
(2) The kid loved the cheese whichi the brand new microwave melted ti
yesterday afternoon while the entire family was watching tv.
(3) The butteri in the small white dish melted ti after the boy turned on the
brand new microwave.
In this experiment as well, the priming eect can only be observed in a small
window, which for normals is immediately at the object position; if the semantic prime is presented with a delay, the priming eect is gradually lost in the
unimpaired population, an indication that it is indeed the reactivation of the
semantic content of the moved phrase at the trace position that causes the
eect. Here, again, the aphasics exhibit a priming eect, but only when the
semantically (un)related word is presented with a delay of 600 ms, indicating
that the patients can construct the trace associated with the moved wh-phrase
or DP. Thus, the representation of the relevant syntactic knowledge must be
available to them; otherwise no priming eect would be expected.
These observations all point in the direction that a processing-based
account of aphasic behaviour is on the right track: the knowledge base seems to
be available, and can be used by the patients under particular conditions.
However, the task cannot be too complicated or involve too many sub-tasks,
and the patients need to be given sucient time to do the task. At the behavioural level, young children and second language learners behave similarly to
the aphasics in a number of respects (subjects are omitted, verbs are not
inected for tense and agreement but occur in the innitival form in the
corresponding syntactic position instead, and functional categories conjunctions, determiners, pronouns, auxiliaries, copula verbs and prepositions, for
instance are omitted), which suggests that their behaviour should be
explained along similar lines, viz. in terms of processing limitations, in line with
Ockhams razor (see also Avrutin, Haverkort and Van Hout 2001 and the
dierent papers in that volume). We hypothesize that in other populations,
particularly second language learners, however, these limitations are of a
dierent nature: not so much timing restrictions, as in the aphasic population,
but the use of qualitatively dierent processing strategies (see below).
second language learners have native-like knowledge, just like the aphasics
(suggesting that access to Universal Grammar (UG) for a second language is
possible). However, they may actually process this knowledge in a non-nativelike manner. Their non-native processing, though, may not be due to timing
limitations as in aphasic populations (a quantitative dierence) but may be a
qualitative dierence.
We will now look at studies that investigate whether advanced second
language learners of Dutch, even if they exhibit the same knowledge as native
speakers in an o-line grammaticality judgment task, exhibit the same neurophysiological responses to grammatical violations. This would indicate that,
even though their knowledge is comparable to that of native speakers, their online processing diers qualitatively. Comparison of data obtained using the
traditional grammaticality judgment technique with those obtained by tapping
directly into electrophysiological activity in the brain associated with a specic
processing phenomenon allows us to study knowledge and processing separately. Grammatical gender is specically interesting in this respect, because it
involves both lexical and syntactic aspects; hence storage, computation, and
their interaction can be studied simultaneously.
neuter gender nouns. The indenite determiner is the same for both genders,
i.e. een, and the agreement is only evident on the adjective (adjective-noun
agreement). For indenite common gender nouns, the sux -e is added to the
adjective, while for indenite neuter nouns the adjective remains uninected,
as shown in the following examples:
(4) a.
Een klein
kind.
a small- child-neut
b. Een klein-e
tafel.
a small-agr table-com
Participants
In total 59 participants were tested on this task: 34 native speakers of Dutch
formed the control group, while there were 25 second language learners with
German as their native language. As we were interested in studying advanced
second language learners, participants were required to have a high level of
prociency. Therefor attain such participants had to have been using Dutch for
at least three years. A prociency score was also obtained from each second
language participant; a score of 90% or more correct was required. This
prociency score was determined by testing participants on their knowledge of
number and niteness agreement.1 Information about the participants is
summarized in Table 1.
Materials and methodology
The grammaticality judgment test contained 80 sentences of interest, each of
which contained the critical determiner-adjective-noun sequence. Half of these
items belonged to the determiner-noun agreement condition and the other half
to the adjective-noun agreement condition. For the rst condition, the critical
Exposure to Dutch
Dutch (n = 34)
N/A
Range: 90100%
Average: 98%
German (n = 25)
Range: 92100%
Average: 97%
* The one German subject who had less then 3 years of exposure to Dutch, started teaching himself
Dutch while still living in Germany, but those years were not counted, as the amount of Dutch used
before moving to the Netherlands could not be determined.
nouns in the sentences were preceded either by the correct denite determiner
or by the incorrect denite determiner (Example 5). In the second condition
there were sentences containing indenite NPs in which the critical noun was
preceded either by the correctly or incorrectly inected form of the adjective
(Example 6). The test also included 200 ller sentences with dierent types of
violations: 80 sentences that were used in the prociency measure, 80 sentences
looking at the use of the relative pronouns and 40 sentences looking at the form
of the predicative adjective. Full details can be found in Sabourin (2003).
(5) Het/*De
kleine kind
probeerde voor het eerst te lopen.
the-neut/*com small child-neut tried
for the rst to walk.
The small child tried to walk for the rst time. (DET-N agreement)
(6) Hij loopt op een gekke/*gek
manier. (A-N agreement)
he walks in a funny-com/*neut way.
The critical nouns used in this experiment were controlled for frequency. Half
of the items were of high frequency while the other half were of a middle
frequency. The middle frequency items were still of a fairly high frequency to
ensure that second language participants would know them. The frequency of
each item was determined through the CELEX database (Burnage 1990). The
log frequency of each high frequency item was between 1.96 and 2.98 (average
2.28); for each middle frequency item, it was between 1.11 and 1.49 (average
1.31). Items were also broken down into gender class: half of the nouns were
common gender nouns (de) and the other half were neuter gender ones (het).
Each subject received a grammaticality judgment questionnaire. The
participants were asked to rst go through the test making a yes/no decision as
to the grammaticality of each sentence. They were required to complete this
181
task in 30 minutes. After judging the grammaticality of each sentence, they were
asked to go back to the beginning and correct every sentence they had marked
as ungrammatical. This was done to ensure that subjects were rejecting a
sentence for the right reasons and not, for instance, due to the fact that they felt
that an incorrect preposition or incorrect word order had been used.
In scoring the grammaticality judgments only sentences that were both
correctly judged as grammatical or ungrammatical and that contained a
relevant correction were considered as correct answers. For example, in the
ungrammatical version of the sentence in (5), repeated below as (7), the
participant correctly might have said the sentence was ungrammatical, but, in
the correction of the sentence only changed the position of the prepositional
phrase voor het eerst. If this was the case, the answer was scored as incorrect.
(7) *De
kleine kind
probeerde voor het eerst te lopen.
the-com small child-neut tried
for the rst to walk
The small child tried to walk for the rst time.
Results
The results for this experiment will be analysed using a four-way Anova
(analysis of variance). Only responses to the ungrammatical items will be
analysed as scores on the grammatical items were near perfect for both groups.
The within-subjects eects were deniteness (denite and indenite), frequency (high and middle), and gender (common and neuter). The between-subjects
eect was native language (Dutch and German). Only one signicant interaction was found in this analysis; deniteness signicantly interacted with L1
(F(1,57) = 15.8, p < .001). This eect can be seen in Figure 1. The main eects of
deniteness (F(1,57) = 32.84, p < .001) and L1 (F(1,57) = 25.37, p < .001) were
also signicant.
What is most important to note here is that while there was a dierence
between the native speakers and the L2 learners, this is only clearly the case
when indenite NPs are being used. The L2 learners perform signicantly less
100
90
80
def
indef
70
60
50
Dutch
German
Figure 1.Scores (in percent) comparing the Dutch and German scores on the denite
and indenite NPs.
worse than the native speakers on the denite NP items compared to the
indenite NP items.
To summarize, the German group shows that, for the denite NPs (the
determiner-noun agreement condition), their knowledge is similar to that of
the native speakers. However, for indenite NPs (the adjective-noun agreement
condition), the German group performs very poorly. One way to interpret these
results is by noting that determiner-noun agreement is similar to simply
assigning gender to nouns and can, therefore, be done on the basis of lexical
information rather than via a syntactic process. On the other hand, adjectivenoun agreement requires the participants to take the lexical knowledge of which
gender an item is and apply this information in order to correctly inect the
adjective. In the next experiment we look at the on-line processing of the same
sentences. The question now is how the L2 group processes these dierent kinds
of data on the on-line version of the task.
4.1.2 Experiment 2: Processing
As was seen in the rst experiment, the German group showed that when they
are given an overt determiner they can judge the grammaticality quite accurately. If only this o-line measure had been used, the conclusion might be that
second language speakers acquire a native-like competence of their second
language. But, as was argued above, there are some reasons to think that the
representation of knowledge on the one hand and language processing on the
other can be dissociated. For aphasics, evidence was presented, indicating that
they still have knowledge of the language but that they cannot process the
language on-line in a quick enough manner. It is therefore also possible that
183
Participants
In total 39 participants were tested on the ERP version of the above experiment.
There were 23 native speakers of Dutch and 14 second language speakers with
German as their native language. The L2 participants have lived in the Netherlands between two and 32 years with an average of 9.8 years. None of the
participants took part in Experiment 1.
Materials and methodology
The critical stimulus sentences used in this experiment were the same sentences
used in Experiment 1. While the task used in the ERP version also contained a
grammaticality judgment, there were a few important dierences in how the
on-line task was run compared with the o-line task from Experiment 1.
During the ERP measurement, participants were seated in a dimly lit soundproof room facing a computer monitor. Sentences were presented word by
word in the middle of the screen (the word was on the screen for 250 ms,
followed by a blank screen for 250 ms before the next word appeared). Each
sentence was preceded by an asterisk (to let participants know that a new
sentence was about to start). After each sentence, a delay screen was displayed,
followed by a screen requesting subjects to give a grammaticality judgment by
pushing one of two buttons. After each sentence participants were given two
seconds in which they were allowed to blink.2 The experiment started with a
practice session to allow participants to get used to the presentation style of the
sentences and to practice not blinking during the sentence trials. The actual
experiment lasted approximately one hour. The native speakers were given
three breaks while the L2 participants were given a total of seven breaks; the
length of the pause was chosen by the participant, so the total testing time
varied, depending on the length of the pauses that were taken.
EEG recording
The EEG activity was recorded by means of tin electrodes mounted in an elastic
cap (Electro-Cap International) from 12 electrode sites, based on the international 1020 system. The 12 electrodes analyzed were: F7, Fz, F8, T3, Cz, T4, T5,
Pz, T6, O1, Oz and O2. The F represents frontal electrodes, T represents
temporal electrodes, C represents central electrodes, P represents parietal
185
electrodes and O represents occipital electrodes. Odd numbers represent electrodes on the left half of the scalp, even numbers represent electrodes on the right
half, and the z represents electrodes along the midline. All electrodes were
referenced to linked mastoids. Both horizontal and vertical electro-oculograms
(EOGs) were measured for both eyes. Electrode impedances were kept below 5 k.
EEG and EOG signals were sampled at 1000 Hz, amplied and digitally ltered
with a cut-o frequency of 30 Hz; eective sample frequency was 100 Hz.
Results
First the behavioural results (accuracy on the grammaticality judgment) will be
presented followed by separate analyses of the on-line ERP data for the native
speakers and the German group. For the L2 group only the sentences which
they correctly judged as grammatical will be analyzed.
The average accuracy scores for the Dutch and German groups are presented in Table 2.
Table 2.Accuracy scores (in %) ont the ERP version of the grammaticality judgment task
NP Denites
Dutch (n = 23)
German (n = 14)
NP Indenites
grammatical
ungrammatical
grammatical
ungrammatical
95%
88%
90%
80%
93%
93%
92%
68%
Average ERPs were computed at the above electrode sites for each participant in all conditions. The averaging was done for an interval starting at the
onset of the critical noun and continuing for 1500 ms post-onset. All averages
were aligned with a 200 ms pre-stimulus baseline (200 ms before onset of the
critical noun is set to zero for both conditions to correct for pre-existing
dierences in the EEG). For analysis purposes, averaged ERPs of each 1500 ms
epoch were divided into 50 intervals of 30 ms. This method allows one to see
the onset and duration of eects more clearly. In each of these 50 intervals, mean
amplitudes were statistically analysed with a Manova. Eects will be reported only
if three or more successive intervals reach signicance at the .05 level for native
speakers and at the 0.1 level for the German group. Three successive signicant
intervals are more likely to reect a real and reliable eect despite the use of
multiple comparisons. A .1 level will be allowed for the L2 speakers in order to
avoid false negative results of the P600 as less of their data can be analysed (only
the sentences for which they made a correct judgment) and it is expected that
their data will also be more variable. Both of these reasons will likely make it
more dicult to nd signicant dierences in the wave patterns so a less strict
level of signicance will be taken, but the reader must be aware then, that
unexpected signicant results should be looked at carefully.
Each language group will be analysed separately and then discussed
comparatively. For each 30 ms interval a three-way Manova was carried out
looking at the eects of grammaticality (two levels), front to back scalp distribution (four levels: frontal, temporal, parietal and occipital), and left to right
scalp distribution (three levels: left, midline and right). The four levels of the
front to back scalp distribution are: F7, Fz and F8 as the rst level, T3, Cz and
T4 as the second level, T5, Pz, and T6 as the third level and O1, Oz and O2 as
the fourth level. The three levels of the left to right scalp distribution are: F7, T3,
T5 and O1 as the left hemisphere electrodes; Fz, Cz, Pz and Oz as the midline
electrodes and; F8, T4, T6 and O2 as the right hemisphere electrodes.
Denite NPs
Upon a visual inspection of the ERP patterns for the denite NPs for the native
speakers a clear P600 pattern is seen; the ungrammatical sentences are more
positive than the grammatical sentences over the more posterior electrodes.
This can be seen by looking at Figure 2.
Looking at the waves statistically the presence of a P600 is supported.
Within each of the 30 intervals from 570 to 1500 ms the eect of grammaticality
is signicant at the .05 cut-o. Within this same time frame there was also a
signicant interaction with the front to back factor. From 570 to 900 ms, there
is a largely distributed P600 component which is signicant over the following
electrodes: Fz, C3, Cz, C4, T5, P3, Pz, P4, T6, O1, Oz and O2. From 900 ms to
the end the posterior positivity is maintained, while a frontal negativity starts
which is only signicant to the .05 level at electrodes F3 and F8.
Upon visually inspecting the ERP patterns for the German participants (see
Figure 3) we can also see a P600 component.
Statistically we see that indeed a P600 eect is present. However, it is much
more restricted and starts later than the one found for native speakers. Between
840 and 990 ms there is a signicant positivity for the ungrammatical sentences at
electrodes C3, Cz, C4, T5, P3, Pz, P4 and T6. This can be seen in Figure 4 where
the dierence waves at electrode Pz are compared for the Dutch and German
groups. Dierence waves represent the wave found for the ungrammatical
sentence minus the wave found for the grammatical sentence; thus a positivity
in the dierence wave reects a positivity in the ungrammatical sentences.
Figure 2.ERP wave patterns for the grammatical (darker line) versus ungrammatical
sentences for the native speakers in the NP denite condition. The y-axis represents a
voltage of 5 microvolts with positive plotted up.
In Figure 4 we can see that the P600 eect starts later for the German group
and that its maximal amplitude occurs later as well. Another dierence between
the ERP waves for the German and Dutch groups can be seen. The German group
does not show the late frontal negativity that is seen in the native speaker ERP.
Indenite NPs
A visual inspection of the native speaker data (see Figure 5) shows that a P600
eect is also present for these sentences. Statistically, the P600 eect is signicant from 600 to 710 ms at electrodes C3, Cz and C4 and between 600 and 1500
ms at electrodes T5, P3, Pz, P4, T6, O1, Oz and O2. Electrodes Fz and F8 show
a late frontal negativity. This can be seen in Figure 5.
Visual inspection of the German data reveals no obvious eects (see
Figure 6). It is important to note that only sentences that were correctly judged
Figure 3.ERP wave patterns for the grammatical (darker line) versus ungrammatical
sentences for the German group in the NP denite condition. The y-axis represents a
voltage of 5 microvolts with positive plotted up.
Figure 4.The dierence waves for the ungrammatical minus the grammatical condition for
the NP denite sentences. The P600 component as seen at electrode Pz for the native
speakers (the darker line) and the German speakers. The y-axis represents a voltage of
8 microvolts with positive plotted up.
In the case of the denite NP where gender agreement between the overt
determiner and noun can be seen as equivalent to gender assignment to nouns,
which is similar in Dutch and German, the Germans are able to perform well
o-line and their on-line processing looks very similar to that of native speakers, although the P600 component found for the ungrammatical sentences
occurs later and is more restricted. However, in the case of Dutch indenite
NPs, where agreement can only occur at a more purely syntactic level, since
more than just knowing whether an item is common or neuter is required, the
German speakers have quite a bit of diculty in the o-line judgment, and in
looking at processing of only the items they correctly judged in the on-line task
they do not show a P600 component.
Figure 5.ERP wave patterns for the grammatical (darker line) versus ungrammatical
sentences for the native speakers in the NP indenite condition. The y-axis represents
a voltage of 5 microvolts with positive plotted up.
5. General conclusions
The main goal of this paper was to show that in the study of language behaviour
in general and in the eld of second language acquisition in particular the
representation of grammatical knowledge and the processing in which this
knowledge is employed need to be carefully distinguished. We have shown that
for the phenomenon of grammatical gender and for one specic group of
second language learners this is indeed an important distinction, because this
group shows a dierence in processing at least for the indenite NPs. While
quantitative dierences in language processing can be seen in aphasic populations, when compared with unimpaired language users, we have shown that
there is actually a qualitative dierence between native speakers and second
language learners in the processing of language. The results presented in this
191
Figure 6.ERP wave patterns for the grammatical (darker line) versus ungrammatical
sentences for the German group in the NP indenite condition. The y-axis represents a
voltage of 5 microvolts with positive plotted up.
paper suggest that the German participants may be making use of a translation
strategy to learn Dutch gender assignment knowledge and that they may be
using their L1 processing strategies to process their L2 for cases where the
grammars are very similar. Further support for this is seen in the ERP processing patterns for sentences involving subject-verb agreement and niteness, both
phenomena for which the Germans show native-like knowledge in the o-line
task. For niteness structures, for which the German group can translate their
L1 strategies, they show native-like processing but for subject-verb agreement,
which exists in German but is dierent in this language, the processing is
dierent from that observed in native speakers (see Sabourin 2003). Another
important thing to note is that the very clear late frontal negativity which was
seen in the ERPs of the native speakers was not found for the German group
even for the NP denite condition for which a P600 eect can be seen.3 These
Figure 7.The dierence waves for the ungrammatical minus the grammatical condition for
the NP indenite sentences. The P600 component as seen at electrode Pz for the native
speakers (the darker line) and the German speakers. The y-axis represents a voltage of
8 microvolts with positive plotted up.
ndings suggest that linguistic processing (as reected by the P600) can only
occur in the L2 when the processing strategy from the L1 can be used relatively
directly in L2 processing. Processing of grammatical structures that are not
similar in the L1 and L2 may be learned and handled by more general cognitive
strategies. Ullman (2001) discussed the declarative/procedural memory
distinction in terms of L1 processing, lexical knowledge being declarative and
syntactic knowledge being procedural in nature. For L2 speakers, Ullman claims
that both lexical and syntactic knowledge usually rely on declarative memory,
although Ullman suggests that factors such as age of exposure and practice may
inuence the ability to use procedural memory in L2 learners. Using this terminology the results presented here suggest that in L2 processing of grammatical
knowledge, only in the case where the L1 and the L2 are similar can procedural
memory be used by advanced adult L2 learners although probably with a
quantitative dierence compared to native speakers, viz. a temporal delay.
<DEST "sab-n*">
Notes
* We would like to thank Laurie Stowe, John Hoeks, Liz Temple and two anonymous
reviewers for their comments on an earlier draft of this paper. The research of the rst author
was funded by the School of Behavioral and Cognitive Neurosciences (BCN) of the University of Groningen; the research of the second author was funded by a grant from the Royal
Netherlands Academy of Sciences (KNAW).
1. For more details on the prociency test see Sabourin (2001, 2003).
2. Participants were asked to try their best to not blink during presentation of the sentences
as eye movements and blinks greatly distort the EEG signal.
3. This suggests that whatever the L2 speakers are doing, it is not exactly the same as the
native speakers.
References
Avrutin, S., Haverkort, M. and Van Hout, A. 2001. Introduction: Language acquisition and
language breakdown. Brain and Language 77: 269273.
Burkhardt, P., Piango, M. and Wong, K. 2001. The role of the anterior left hemisphere in
real-time sentence comprehension: Evidence from split intransitivity. Ms. Yale University.
Burnage, G. 1990. A guide for users. Nijmegen: CELEX Centre for Lexical Information.
Caplan, D. and Hildebrandt, N. 1988. Disorders of syntactic comprehension. Cambridge: MIT
Press.
Chomsky, N. 1995. The minimalist program. Cambridge: MIT Press.
Grodzinsky, Y. 1990. Theoretical perspectives on language decits. Cambridge: MIT Press.
Grodzinsky, Y. and Finkel, L. 1998. The neurology of empty categories: Aphasics failure to
detect ungrammaticality. Journal of Cognitive Neuroscience 10 (2): 281292.
Gunter, T. C., Friederici, A. D. and Schriefers, H. 2000. Syntactic gender and semantic
expectancy: ERPs reveal early autonomy and late interaction. Journal of Cognitive
Neuroscience 12 (4): 556568.
Haarmann, H. 1993. Agrammatic aphasia as a timing decit. Doctoral dissertation, University
of Nijmegen.
Hagoort, P. and Brown, C. 1999. Gender electried: ERP evidence on the syntactic nature
of gender processing. Journal of Psycholinguistic Research: Special Issue on Processing of
Grammatical Gender 28 (6): 715728.
</TARGET "sab">
Hagoort, P., Brown, C. and Groothusen, J. 1993. The syntactic positive shift (SPS) as an
ERP-measure of syntactic processing. Language and Cognitive Processes 8: 439483.
Kolk, H. 1995. A time-based approach to agrammatic production. Brain and Language 50:
282303.
Kolk, H. 1998. Disorders of syntax in aphasia: Linguistic-descriptive and processing
approaches. In Handbook of neurolinguistics, B. Stemmer and H. Whitaker (eds),
250260. San Diego: Academic Press.
Kolk, H. 2002. Language production in agrammatic aphasics: an experimental study. Paper
presented at the University of Nijmegen Linguistics Colloquium.
Kutas, M. 1993. In the company of other words: Electrophysiological evidence for singleword and sentence context eects. Language and Cognitive Processes 8 (4): 533632.
Linebarger, M., Schwarz, M. and Saran, E. 1983. Sensitivity to grammatical structure in socalled agrammatic aphasics. Cognition 13: 361392.
Osterhout, L. and P. J. Holcomb. 1992. Event-related brain potentials elicited by syntactic
anomaly. Journal of Memory and Language 31: 785806.
Rugg, M. D. and Coles, M. G. H. 1995. The ERP and cognitive psychology: Conceptual
issues. In Electrophysiology of the mind: Event-related brain potentials and cognition,
M. D. Rugg and M. G. H. Coles (eds), 2739. Oxford: Oxford University Press.
Sabourin, L. 2001. L1 eects on the processing of grammatical gender in L2. In Eurosla
Yearbook, Volume 1, S. Foster-Cohen and A. Nizegorodcew (eds), 159169. Amsterdam:
John Benjamins.
Sabourin, L. 2003. Grammatical gender agreement in L2 processing. Doctoral dissertation,
University of Groningen.
Ullman, M. T. 2001. The neural basis of lexicon and grammar in rst and second language:
The declarative/procedural model. Bilingualism: Language and Cognition 4 (1):
105122.
<LINK "gre-n*">
<TARGET "gre" DOCINFO AUTHOR "David W. Green"TITLE "Neural basis of lexicon and grammar in L2 acquisition"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 9
1.
Introduction
In acquiring a second language (L2) individuals must grasp its grammar and its
vocabulary but dierences in the context of acquisition from a rst acquired
language (L1) may mean that dierent learning mechanisms are involved. Such
dierences, in turn, carry implications for the neural representation of L1 and
L2. Understanding the neural basis of the representation of L1 and L2 can
therefore contribute to a deeper understanding of the interface of syntax and
lexicon in L2 acquisition.
The basic orienting question is this: is a persons lexical and grammatical
knowledge represented dierently if it is learned as an L2 as opposed to an L1?
In this chapter, I contrast this proposal, termed the dierential representation
hypothesis, with an alternative, termed the convergence hypothesis. This
hypothesis states that as prociency in L2 increases, non-native speakers
represent, and process, the language in the same way as native speakers of that
language. In addressing these hypotheses, I consider both their computational
basis, and relevant neuropsychological and neuroimaging data.
The chapter is structured as follows: I rst consider the dierential representation hypothesis. On one version an L2 is represented in a persons righthemisphere rather than in their left-hemisphere. I reject this possibility. Next I
consider a version in which in L2, in contrast to L1, grammatical and lexical
information is represented in a common memory system. This version of the
hypothesis relies on a distinction between two memory systems that has been
justied primarily by studies on amnesic patients. However, computational
modelling shows that a single memory system is sucient to generate the data
of these amnesic patients. Such a result prompts consideration of the computational basis for the alternative, convergence hypothesis. By themselves computational arguments are not decisive and so we consider neuropsychological and
neuroimaging data in an eort to adjudicate between the two hypotheses
empirically. A nal section considers the kind of studies needed to further our
understanding of the issue.
attainment in L2 post puberty. They also found evidence of native-like attainment in late learners of L2. They argue, in line with Flege, Yeni-Komshian and
Liu (1999), that practice is an important factor in determining the eventual level
of attainment. The nature of the L1 and L2 pairing may also be relevant.
Turning now to the key element of Ullmans (2001b) proposal, namely the
idea that emphasis is shifted to the declarative memory system in L2 learners
and that there is little or no involvement of a procedural memory system in
grammatical processing. This proposal presumes that there is good evidence for
the existence of these two types of memory systems. Data from amnesic patients
appear to provide compelling support. Amnesics have poor declarative memory
but show normal performance on various tasks involving nondeclarative
memory (Gabrieli 1998). A study by Knowlton and Squire (1993) is exemplary.
They used dot patterns created by systematically distorting a prototype pattern.
Amnesic patients were able to classify these patterns normally (a nondeclarative
task) but were severely impaired in their ability to recognize whether or not a
particular pattern had been presented previously (a declarative task). Knowlton
and Squire interpreted these data as evidence that performance in the two tasks
was mediated by two dierent memory system one of which (the declarative
system) was impaired and the other of which was not. However, computational
work by Nosofsky and Zaki (1998) challenged this interpretation by showing
that dierences in performance on these two tasks can be obtained within a
single memory system. A slight reduction in the value of a sensitivity parameter
in their computational model reduced classication performance marginally
but exerted a marked eect on recognition performance.
More pertinent to the present concern is work on the learning of articial
grammars. Knowlton and Squire (1994, 1996) contrasted normals and amnesics
in their ability to learn an articial grammar. In such studies individuals rst
memorise a set of strings generated by a (nite-state) grammar and are then
informed about a set of rules generating the strings. In the classication task,
they have to classify a new set of strings into those that are grammatical and
those that are not. In the recognition task, they have to indicate whether or not
a string of symbols was presented. In the studies by Knowlton and Squire
classication performance in amnesic patients was normal but recognition
performance was impaired. Knowlton and Squire interpreted this as evidence
that the two tasks are mediated by dierent memory systems only one of
which, the declarative memory system, is impaired in amnesics.
Other studies support a dissociation between recognition and repetition
priming in amnesic patients (e.g. Hamann and Squire 1997: Experiment 1).
by Ullman can emerge from single networks (see also Ullman 2001b, Note 1:10).
At a minimum, such results encourage the search for an alternative formulation. In fact, computational considerations lead us to expect a rather dierent
outcome for L2 acquisition. The next section considers a computational
justication for the convergence hypothesis.
5. The convergence hypothesis and its computational basis
According to the convergence hypothesis, any qualitative dierences between
native speakers of a language and L2 speakers of that language disappear as
prociency increases. Such a hypothesis is broadly in line with the idea that
prociency in language involves identifying, and using, the various cues to
meaning see, for example, the competition model (MacWhinney 1997). The
convergence hypothesis is specically concerned with the neural, and not
simply the cognitive, representation of L2. Given the diversity of languages, in
order to consider the computational grounds for the convergence hypothesis,
we need to characterise languages at a certain level of abstractness. There are
four linguistic means for communicating experience (see, for example, Tomasello 1995): individual symbols (lexical items); markers on symbols (grammatical morphology), ordering patterns of symbols (word order) and prosodic
variations of speech (e.g. stress, intonation, timing). Languages dier in the
weight they attach to these dierent linguistic means. In some languages, word
order is basically free and information on who did what to whom is conveyed
by word endings or by prosody in tone languages. By contrast, in English, such
information is conveyed by word order and this is relatively rigid. These
dierent linguistic means or signals require dierent devices for their processing. The rst step of the computational argument for the convergence hypothesis is that the neural representation of the various linguistic devices is similar
across languages. The next paragraph spells out the basis for this argument.
Such devices may be represented by specic networks with a distinct neural
anatomical representation or they may be mediated by a specialised network
with a distributed neural representation. Specialised networks can emerge from
unique interactions amongst a set of regions each fullling a number of
dierent functions (e.g. Mesulam 1990). Consider the development of a system
using these devices to communicate meaning. First, dierent neural regions
may compete to process input. Those regions, whether innately specied, or
possessing some small processing advantage, will come to mediate processing of
a given linguistic means. Neural regions active at the same time will connect
together (Hebb 1949, Robertson and Murre 1999) giving rise to a specialised
network. Second, once a network has come to process signals of a particular
type it will resist processing other types of signals unless input to it is curtailed
in which case it may process signals of another type given that plastic reorganization is possible. One line of support for this view is evidence of crowdingout: when language is displaced to the right-hemisphere as a result of neurological damage to the left-hemisphere, a persons visuo-spatial skills (typically
mediated by the right-hemisphere) are impaired (Teuber 1974, Strauss, Satz
and Wada 1990). Third, given commonalities across brains in the initial
sensitivities of dierent regions call this the commonality assumption2
there will be commonalities in the neural representation of the dierent devices
for speakers of dierent languages.
The second step of the computational argument for the convergence
hypothesis is that the acquisition of an L2 arises in the context of an already
specied, or partially-specied system, with a specic neural network mediating
each device.3 It follows that L2 will receive convergent representation with L1.
Further, given the commonality assumption (see above) the representation of
L2 will converge with the representation of that language learned as an L1.
The convergence hypothesis does not entail that a speaker of L2 will
necessarily achieve native-like levels of performance (for example in achieving
certain phonetic norms, Flege 1995) nor does it exclude the possibility that tasks
such as mental arithmetic are carried out exclusively in L1. Clearly, also, the
contexts of acquisition (e.g. a formal school setting versus an immersion
setting) aect the initial registration of linguistic information. However, in
contrast to the dierential representation hypothesis, the convergence hypothesis is committed to the prediction that as prociency in L2 increases, the same
linguistic means involve the same neural networks as native speakers. The
hypothesis would be refuted if there is no change in representation with
prociency and if a normal, procient L2 speaker activated neural networks
disjoint from those of a native speaker, especially when encoding and decoding
syntactic information. The fact that explicit, declarative representations of
grammatical information, play only an initial role in on-line processing,
according to the convergence hypothesis, does not mean that they are unimportant. Explicit (metalinguistic) representation may well benet the recovery of L2
over L1 following brain-damage (e.g. Lebrun 2002, Paradis 1994, 1997). But
such a possibility, it seems to me, cannot be used to claim a continuing role for
such representations in on-line processing, once the relevant procedures are in
place. However, this possibility is open to test.
P600 (found 600 ms after an event) is sensitive to syntactic violations. ERP data
cannot provide direct evidence of the neural sources of such eects but haemodynamic studies are informative. Studies on grammatical processing and
encoding in native speakers (Hagoort, Brown and Osterhout 2000) suggest a
common syntactic component subserved by the left frontal area (a dorsal part
of Brocas area and adjacent parts of the middle frontal gyrus) and studies on
the semantic representation of words identify regions in the temporo-parietal
region the left extrasylvian temporal cortex and the left anterior inferior
frontal cortex (Price 2000). Neuropsychological data (Donkers, Redfern and
Knight 2000) and also neuroimaging data (e.g. Price, Moore, Humphreys and
Wise 1997) suggest a specic area in the anterior temporal region as a site
critical for the interpretation of sentences (i.e. syntactic-semantic integration).
6.2 The representation of L2
What empirical evidence is there that L2 is represented dierently from L1 as
proposed by the dierential representation hypothesis? According to Ullman
(2001b), L2 learned late will be sensitive to damage to neocortical temporal/
temporal-parietal regions for those linguistic forms that depend on grammatical
processing in L1. A case reported by Ku, Lachmann and Nagler (1996) seems to
support his position. A 16 year old native Chinese speaker who had been living
in the United States for six years and who had received intensive training in
English over this period suered a circumscribed lesion to the left temporal lobe
(as a result of herpes simplex encephalitis). For three weeks following the lesion
he lost the ability to comprehend and to speak English. In contrast, naming in
Mandarin was normal. However, in speaking Mandarin his syntax was simplied and so this case is not decisive support for the claim that grammatical
information is represented dierently in L2.
The notion that L1 grammatical processing is mediated by a frontal-basal
ganglia circuit predicts that damage to the basal ganglia will lead to a selective
loss of L1. Fabbro and Paradis (1995) report the case of E. M. with such a lesion
and, true to prediction, her spontaneous speech in her L1 (Venetan) was poor
whereas her speech was better in her L2 (Italian) that she rarely used prior to
the lesion. Ullman (2001b) considered the nature of her errors. There was a
similar proportion of word nding diculties in both languages but a tendency
for poorer grammatical performance in L1 (e.g. the omission of grammatical
function words in obligatory contexts). However, these eects are small and the
overwhelming dierence is her spontaneous use of L2 in preference to L1.
Green and Price (2001) argue that language control (e.g. the ability to select
between one language and another) is also mediated by frontal-basal ganglia
circuits. An impairment in this system will also give rise to problems in modulating the output from the lexical-semantic system.
Leaving the neuropsychological data on one side, how convincing is the
ERP and neuroimaging data for a distinct representation of L2? Individuals
acquiring L2 can vary in terms of when they acquired L2, how they acquired L2
and how procient they are in using it. Typically, prociency is confounded
with the age of acquisition. In terms of prociency it is natural to expect that
less procient users of L2 will show quantitative dierences on a range of
measures (e.g. naming time, ERP eects and activation patterns). The critical
issue is whether or not there are qualitative dierences indicating that dierent
neural mechanisms are involved. If there are, it is important to determine
whether these necessarily imply dierent representations.
ERP data point to both quantitative and qualitative dierences in processing between L1 and L2. Kutas and Kluender (1991), for instance, found that the
N400 component in response to a semantic anomaly was delayed and of lower
amplitude in a bilinguals less uent language. Likewise, Webber-Fox and
Neville (1996) found N400 present in all groups of Chinese-English second
language learners though it was more delayed in those learning L2 after
reaching the age of 1113 years. More critically, in contrast, to monolinguals,
there was a distinct pattern of response to phrase structure violations in
bilinguals. Only individuals acquiring L2 before the age of four showed no
dierence from native learners of L2. Such data are compatible with the notion
that there is a critical period for language learning and are consistent with the
notion that dierent brain mechanisms mediate syntactic processing in late
learners of L2. But they are not decisive as later exposure to English was
associated with worse performance in identifying syntactically anomalous
sentences. Such individuals may have been circumventing syntactic processing.
Hahne and Friederici (2001) examined the eects of phrase structure
violations and semantic anomaly in Japanese late learners of German. These
individuals also showed substantial error rates in a grammaticality judgement
task and so cannot be considered procient. Hahne and Friederici (2001)
conrmed a delayed N400 eect in response to semantic anomaly but also found
a right anterior central negativity. Unlike native German speakers there was no
early anterior negativity in response to a syntactic violation. They propose that
late learners identify lexical content independently of morphological form (e.g.
the past participle form of the verb) and construct a representation directly
7. Ways forward
The dierential representation hypothesis and the convergence hypothesis concur
that the initial representation of L2 may dier from that of native speakers of that
language. The fundamental research requirement then is to conduct withinparticipant longitudinal studies to chart changes in behaviour on various tasks,
e.g. picture naming (Kroll, Michael, Tokowicz and Dufour 2002) and to
examine changes in ERP and neuroimaging proles as prociency in L2
changes. Further, such studies need to involve both syntactic and lexical tasks.
To my knowledge there are currently relatively few such studies.
As discussed above, in contrast to L1, grammatical knowledge of L2 may be
represented explicitly and declaratively. According to the convergence hypothesis, ERP responses to syntactic anomalies should change with prociency.
Osterhout and McLaughlin (2000) studied responses to semantic and syntactic
anomalies in native speakers of French and in novice learners. Semantic
anomalies yielded N400 and syntactic anomalies yielded a P600 in native
French speakers. French learners after four weeks of instruction showed an
N400 in response to semantic anomalies. In contrast, there was either an N400
or no eect in response to syntactic anomalies. After just four months, however,
syntactic anomalies yielded a P600 but no N400. These data suggest that if there
are qualitative dierences between native speakers and L2 learners, these can be
rather short-lived. Any dierences in responding to syntactic anomalies are
presumably negatively correlated with the increasing grasp of syntax. Consistent
with the convergence hypothesis, Weber-Fox and Neville (submitted, cited in
Ullman 2001b) examined responses to open-class and closed-class words.
Native speakers of English showed a left anterior negativity to closed class words
(N280) and an N400 for open-class words. L2 speakers of English (with Chinese
as their rst language) showed the same open class N400 as native English
speakers. Interestingly, the response to closed class words related to an independent test of their grammatical ability. The higher the score on the test, the
earlier the anterior negativity for closed class words.
211
<DEST "gre-n*">
explicitly encode the direction of motion i.e. make that property of the scene
salient rather than the manner of motion. Prociency should be associated with
changes in the explicit representation of the properties of scenes. In terms of
activation patterns, there should be changes in the activation and connection of
cortical regions mediating those properties so that the correct words can be
selected and expressed in a suitable syntactic frame.5
8. Conclusion
This chapter has contrasted two main hypotheses about the representation and
processing of lexicon and grammar in L2. The subtle form of the dierential
representation hypothesis proposes that declarative representations play a much
more important role in the representation of grammar in L2 than in L1. This
chapter has considered the computational and empirical basis for the dierential representation hypothesis and argued, on both computational and neuroimaging grounds, for an alternative, convergence hypothesis. According to this
hypothesis, as prociency in L2 increases, the networks mediating L2 converge
with those mediating language use in native speakers of that language. Current
evidence marginally favours the convergence hypothesis. However, we lack
appropriate longitudinal studies of L2 acquisition. Crucially, we have little or no
information about the functional integration of dierent neural regions during
second language use. Scope indeed for discovery!
Notes
* I thank the editors of this volume for the opportunity to contribute to our understanding
of the interface between syntax and the lexicon in L2 acquisition and for constructive
comments on a previous draft of this chapter.
1. It would be useful to have converging evidence for the existence of distinct neural systems
beyond that oered by amnesia. Fortunately, we do not need to rely exclusively on neuropsychological data. Neocortical activity is reduced for stimuli that have been processed before
(Ungerleider, 1995) and this datum has been interpreted as evidence that these structures
mediate priming. But amnesics can also be impaired on priming in certain conditions
(Ostergaard, 1999).
2. The commonality assumption is compatible with anatomical variability. Brains dier in
the location of quite major features (e.g. Rickard 2000). The commonality assumption refers
to the sensitivity of neural regions, not to their precise anatomical location. Neuroimaging
evidence for convergence must take such variability into account.
3. At the lexical level, the impact of a prior representation is captured by cognitive models
such as the revised hierarchical model (Kroll and Stewart 1994) and the distributed features
model (e.g. De Groot 1993, Kroll and de Groot 1997). It might also be argued that prior
representation of L1 induces a radically distinct representation of L2. Jiang and Forster
(2001) propose, on the basis of experimental evidence, that lexical items in L2 (they tested
native Chinese speakers with English as the L2) are represented in a non-lexical memory.
This memory allows the meaning of translation equivalents to be retrieved indirectly via the
L1 lexical item. However, it is unclear what it means for a non-lexical system to represent the
syntactic properties of lexical items. Their ndings, as they acknowledge, need to be
replicated with procient L2 speakers and with a dierent language pairs.
4. ERPs provide high resolution temporal evidence of the existence of dierent processes
during language processing. These are derived by averaging signals from an eletroencephalogram over a series of trials that time-locked to the presentation of a particular type of
stimulus. An ERP itself comprises various components (i.e. positive and negative voltage
peaks). Where these are aected by some experimental manipulation they are termed ERP
eects. ERP data are compatible with an innite number of neural generators (the inverse
problem) and so they need to be complemented by data from haemodynamic methods.
Haemodynamic methods (Positron Emission Tomography, PET; or functional Magnetic
Resonance Imaging, fMRI) rely on a close coupling between changes in the activation of a
population of neurons and change in blood supply. A haemodynamic eect arises only when
there is a change in the overall metabolic demand in a neuronal population. PET and fMRI
track dierent signals. PET measures the decay of a short-lived isotope which accumulates
in a neural region in proportion to the amount of blood owing through that region. The
most typical fMRI method indexes metabolic demand and hence relative neural activity by
assessing the ratio of deoxy- to oxyhaemoglobin in the blood (see Rugg (2000) for a critical
appraisal of these methods). It is worth noting that haemodynamic methods, along with
other electrophysiological methods, allow us to identify regions that are sucient for task
performance but they do not allow us to identify regions that are necessary for task performance. Other data are needed to identify which regions are necessary. For instance, if task
performance is impaired in a patient with a lesion at given site then this region, or the
network of which it is part, is necessary for task performance. Likewise, virtual lesions
induced by drugs or by transcranial magnetic stimulation, may help identify regions
necessary for task performance.
5. Lower levels of prociency in L2 might also be associated with more reliance on conceptual/pragmatic information (see Hahne and Friederici above). In native speakers, conceptual
factors aect grammatical encoding (Vigliocco and Hartsuiker 2001) and it is reasonable to
expect that such eects might be more marked in novice learners. A sentence completion
task oers one behavioural measure. Vigliocco and Franck (2001) showed that there were
more errors in generating a predicate when the sex of the referent was incongruent with the
gender of the noun. Given that novice learners of French or Italian, for instance, know the
syntactic gender of the noun, they might also show greater eects of incongruity.
213
References
Abutalebi, J., Cappa, S. F. and Perani, D. 2001. The bilingual brain as revealed by functional
imaging. Bilingualism: Language and Cognition 4: 179190.
Albert, M. L. and Obler, L. K. 1978. The bilingual brain: Neuropsychological and neurolinguistic aspects of bilingualism. New York: Academic Press.
Bialystok, E. 1992. Selective attention in cognitive processing: The bilingual edge. In
Cognitive processes in bilinguals, R. J. Harris (ed.), 501513. Amsterdam: Elsevier Science
Publishers B. V.
Birdsong, D. and Molis, M. 2001. On the evidence for maturational constraints in secondlanguage acquisition. Journal of Memory and Language 44: 235249.
Black, M and Chiat, S. 2003. Noun-verb dissociations: a multi-faceted phenomenon.
Journal of Neurolinguistics 16: 231250.
Breedin, S. D. and Saran, E. M. 1999. Sentence processing in the face of semantic loss:
A case study. Journal of Experimental Psychology: General 128: 547562.
Bchel, C., Frith, C. and Friston, K. 2000. Functional integration: Methods for assessing
interactions amongst neuronal systems. In The neurocognition of language, C. M.Brown
and P. Hagoort (eds), pp. 337355. Oxford: Oxford University Press.
Chee, M. W. L., Tan, E. W. L. and Thiel, T. 1999. Mandarin and English single word
processing studied with functional Magnetic Resonance Imaging. Journal of Neuroscience 19: 30503056.
Chomsky, N. 1995. The minimalist program. Cambridge, MA: MIT Press.
Cleeremans, A. 1993. Mechanisms of implicit learning: Connectionist models of sequence
processing. Cambridge, MA: MIT Press.
Dehaene, S. D., Dupoux, E., Mehler, J., Cohen, L., Paulesu, E., Perani, D., van de Moortele,
P. F., Lehricy, S. and Le Bihan, D. 1997. Anatomical variability in the cortical representation of rst and second languages. Neuroreport 8: 38093815.
De Groot, A. M. B. 1993. Word-type eects in bilingual processing tasks: Support for a
mixed representational system. In The bilingual lexicon, R. Schreuder and B. Weltens
(eds), 2751. Amsterdam: John Benjamins.
De Groot, A.M.B., Delmaar, P. and Lupker, S.J. 2000. The processing of interlexical homographs in translation recognition and lexical decision: Support for non-selective access to
bilingual memory. Quarterly Journal of Experimental Psychology 53A: 397428.
Dijkstra, A., van Jaarsveld, H. and ten Brinke, S. 1998. Interlingual homograph recognition:
Eects of task demands and language intermixing. Bilingualism: Language and
Cognition 1: 5166.
Dronkers. N. F., Redfern, B. B. and Knight, R. T. 2000. The neural architecture of language
disorders. In The new cognitive neurosciences, M. S. Gazzaniga (ed.), 949960. Cambridge, Mass: MIT.
Edelman, G. M. 1989. The remembered present: A biological theory of consciousness. New York:
Basic Books.
Ellis, N. C. 1995. The psychology of foreign language vocabulary acquisition: Implications
for CALL. Computer Assisted Language Learning 8: 103128.
Fabbro, F. 1999. The neurolinguistics of bilingualism: An introduction. Hove, Sussex: Psychology Press.
Fabbro, F. and Paradis, M. 1995. Dierential impairments in four multilingual patients with
subcortical lesions. In Aspects of bilingual aphasia, M. Paradis (ed.), 139176. Oxford,
UK: Pergamon.
Falk, A. R., Durwen, H. F., Mller, C., Knig, M., Mller, E. and Heuser, L. 1999. Determination of eloquent cortical areas in Russian bilinguals performing a word generation
task. Neuroimage 6: S1002.
Flege, J. E. 1995. Second-language speech learning: Theory, ndings and problems. In
Speech perception and linguistic experience: Issues in cross-language research, W. Strange
(ed.), 233277. Timonium, MD: York Press.
Flege, J. E., Yeni-Komshian, G. H. and Liu, S. 1999. Age constraints on second-language
acquisition. Journal of Memory and Language 41: 78104.
Frenck-Mestre, C. and Pynte, J. 1997. Syntactic ambiguity resolution while reading in second
and native languages. Quarterly Journal of Experimental Psychology 50A: 119148.
Gabrieli, J. D. 1998. Cognitive neuroscience of human memory. Annual Review of
Psychology 49: 87115.
Gollan, T. H. and Kroll, J. F. 2001 Lexical access in bilinguals. In A handbook of cognitive
neuropsychology: What decits reveal about the human mind, B. Rapp (ed.), 321345.
New York: Psychology Press.
Green, D. W. 1986. Control, activation and resource: A framework and a model for the
control of speech in bilinguals. Brain and Language 27: 210223.
Green, D. W. 1998. Mental control of the bilingual lexico-semantic system. Bilingualism:
Language and Cognition 1: 6781.
Green, D. W. and Price, C. 2001. Functional imaging in the study of recovery patterns in
bilingual aphasics. Bilingualism: Language and Cognition 4: 191201.
Gupta, P. and Dell, G. S. 1999. The emergence of language from serial order and procedural
memory. In The emergence of language, B. MacWhinney (ed.), 447481. Mahwah:
Lawrence Erlbaum Associates.
Hagoort, P., Brown, C. M. and Osterhout, L. 2000. The neurocognition of syntactic
processing. In The neurocognition of language, C. M. Brown and P. Hagoort (eds),
273316. Oxford: Oxford University Press.
Hahne, A. and Friederici, A. D. 2001. Processing a second language: Late learners comprehension mechanisms as revealed by event-related brain potentials. Bilingualism:
Language and Cognition 4: 123141.
Hamann, S. B. and Squire, L. R. 1997. Intact priming for novel perceptual representations
in amnesia. Journal of Cognitive Neuroscience 9: 699713.
Hebb, D.O. 1949. The organization of behaviour: A neuropsychological theory. New York: Wiley.
Hermans, D. 2000. Word production in a foreign language. Doctoral dissertation, Katholieke
Universiteit Nijmegen.
Jiang, N. and Korster, K. I. 2001. Cross-language priming asymmetries in lexical decision
and episodic recognition. Journal of Memory and Language 44: 3251.
Johnson, J. S. and Newport, E. L. 1989. Critical period eects in second language learning:
The inuence of maturational state on the acquisition of English as a second language.
Cognitive Psychology 21: 6099.
215
Jus, A. 1998. Main verb versus reduced relative clause ambiguity resolution in L2 sentence
processing. Language Learning 48: 107147.
Kim, K. H. S., Relkin, N. R., Lee, K. M. and Hirsch, J. 1997. Distinct cortical areas associated
with native and second languages. Nature 388: 171174.
Kinder, A. and Shanks, S. 2001. Amnesia and the declarative/non-declarative distinction:
A recurrent network model of classication, recognition and repetition priming.
Journal of Cognitive Neuroscience 13: 648669.
Klein, D., Milner, B., Zatorre, R., Meyer, E. and Evans, A. 1995. The neural substrates
underlying word generation: A bilingual functional-imaging study. Proceedings of the
National Academy of Sciences USA 92: 28992903.
Knowlton, B. J. and Squire, L. R. 1993. The learning of categories: Parallel brain systems for
item memory and category knowledge. Science 262: 17471749.
Knowlton, B.J. and Squire, L.R. 1994. The information acquired during articial grammar
learning. Journal of Experimental Psychology: Learning, Memory, and Cognition 20: 7991.
Knowlton, B. J. and Squire, L. R. 1996. Articial grammar learning depends on implicit
acquisition of both abstract and exemplar-specic information. Journal of Experimental
Psychology: Learning, Memory, and Cognition 22: 169181.
Kroll, J. F. and Dussias, P. E. (in press). The comprehension of words and sentences in two
languages. In Handbook of bilingualism, T. Bhatia and W. Ritchie (eds), Cambridge,
MA: Blackwell Publishers.
Kroll, J. F. and de Groot, A. M. B. 1997. Lexical and conceptual memory in the bilingual:
Mapping form to meaning in two languages. In Tutorials in bilingualism: Psycholinguistic perspectives, A. M. B. de Groot and J. F. Kroll (eds), 169199. Mahwah, NJ:
Lawrence Erlbaum Associates.
Kroll, J. F. and Stewart, E. 1994. Category interference in translation and picture naming:
Evidence for asymmetric connections between bilingual memory representations.
Journal of Memory and Language 33: 149174.
Kroll, J. F., Michael, E., Tokowicz, N. and Dufour, R. (2002). The development of lexical
uency in a second language. Second Language Research 18: 137171.
Ku, A., Lachmann, E. A. and Nagler, W. 1996. Selective language aphasia from herpes
simplex encephalitis. Pediatric Neurology 15: 169171.
Kutas, M. and Kluender, R. 1991. What is who violating? A reconsideration of linguistic
violations in light of event-related brain potentials. In Cognitive Electrophysiology, H.-J.
Heinze, T. F. Mnte and G. R. Mangun (eds),183210. Boston: Birkhuser.
Lebrun, Y. 2002. Implicit competence and explicit knowledge. In Advances in neurolinguistics of bilingualism, F. Fabbro (ed.), 299313. Udine: Forum.
Lenneberg, E. H. 1976. Biological foundations of language. New York: Wiley.
Levelt, W. J. M., Roelofs, A. and Meyer, A. S. 1999. A theory of lexical access in speech
production. Behavioral and Brain Sciences 22: 175
Loring, D. W., Meador, K. J., Lee, G. P., et al. 1990. Cerebral language lateralization:
evidence from intracorotid amobarbital testing. Neuropsychologica 28: 831838.
McClelland, J.L., McNaughton, B.L. and OReilly, R.C. 1995. Why there are complementary
learning systems in the hippocampus and neocortex: Insights from the successes and
failures of connectionist models of learning and memory. Psychological Review 102:
419437.
</TARGET "gre">
Robertson, I. H.and Murre, J. M. 1999. Rehabilitation of brain damage: Brain plasticity and
principles of guided recovery. Psychological Bulletin 125: 544575.
Rugg, M. R. 2000. Functional neuroimaging in cognitive science. In The neurocognition of
language, C.Brown and P. Hagoort (eds), 1536. Oxford: Oxford University Press.
Scoresby-Jackson, R. E. 1867. Case of aphasia with right hemiplegia. Edinburgh Medical
Journal 12: 696706.
Segalowitz, N. S. and Segalowitz, S. J. 1993. Skilled performance, practice, and the dierentiation of speed-up from automatization eects: Evidence from second language word
recognition. Applied Psycholinguistics 14: 369385.
Slobin, D. I. 1996. From thought and language to thinking for speaking. In Rethinking
linguistic relativity, J. J. Gumperz and S. C. Levinson (eds), 177202. Cambridge:
Cambridge University Press.
Springer, J. A., Binder, J. R., Hammeke, T. A. et al. 1999. Language dominance in neurologically normal and epilepsy subjects: A functional MRI study. Brain 122: 20332046.
Squire, L. R., Knowlton, B. and Musen, G. 1993. The structure and organization of
memory. Annual Review of Psychology 44: 453495.
Squire, L. R. 1994. Declarative and nondeclarative memory: Multiple brain systems
supporting learning and memory. In Memory systems, D. L. Schacter and E. Tulving
(eds), 203231. Cambridge, MA: MIT Press.
Strauss, E., Satz, P. and Wada, J. 1990. An examination of the crowding hypothesis in epileptic
patients who have undergone the carotid amytal test. Neuropsychologia 28, 12211227.
Teuber, H. L. 1974. Why two brains? In The Neurosciences. Third study program, F. G.
Worden (ed.), 7174. Cambridge: MIT Press.
Tomasello, M. 1995. Language is not an instinct. Cognitive Development 10: 131156.
Ullman, M. T. 2001a. The declarative/procedural model of lexicon and grammar. Journal
of Psycholinguistic Research 30: 3769.
Ullman, M. T. 2001b. The neural basis of lexicon and grammar in rst and second language:
the declarative/procedural model. Bilingualism: Language and Cognition 4: 105122.
Ungerleider, L. G. 1995. Functional brain imaging studies of cortical mechanisms for
memory. Science 270: 769775.
Vaid, J. and Hull 2002. Re-envisioning the bilingual brain using functional neuroimaging:
Methodological and interpretive issues. In Advances in neurolinguistics of bilingualism,
F. Fabbro (ed.), 315355. Udine: Forum.
Vigliocco, G. and Franck, J. 2001. When sex aects syntax: Context eects in sentence
production. Journal of Memory and Language 45: 368390.
Vigliocco, G. and Hartsuiker, R. J. 2001. The interplay of meaning, sound, and syntax in
sentence production. Psychological Bulletin (under review).
Weber-Fox, C. and Neville, H. J. 1996. Maturational constraints on functional specializations for language processing: ERP and behavioral evidence in bilingual speakers.
Journal of Cognitive Neuroscience 8: 231256.
Weber-Fox, C. and Neville, H. J. (submitted). Sensitive periods dierentiate processing
subsystems for open and closed class words: An ERP study in bilinguals. cited in
Ullman, 2001b (above).
<TARGET "hou" DOCINFO AUTHOR "Roeland van Hout, Aafke Hulk and Folkert Kuiken"TITLE "The interface"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Chapter 10
The interface
Concluding remarks
Roeland van Hout, Aafke Hulk and Folkert Kuiken
University of Nijmegen (Hout) / Utrecht University (Hulk) /
Universiteit van Amsterdam (Kuiken)
1.
In its bare form, a grammar can be dened as a set of elements or symbols (the
lexicon) and a set of rules (the syntax) that together produce output strings, the
utterance of the language belonging to the grammar. The format of the lexicon
is evident: it has no structure, and it does not need to have one. Chomsky
(1965: 84) characterizes the lexicon as simply an unordered list of all lexical
formatives. Although the assumed absence of order may be accepted as a
heuristic device, the lexicon of human languages and of human speakers is not
an unordered list, far from that, as is apparent from many recent lexical studies
as well as from the chapters in this book. The form, role and structure of lexical
items, words, lemmas, or whatever the lexical formatives are called, and the
relationships or connections between them constitute the pivotal domain of
research in the chapters of this book. The nearest neighbour to which lexical
items and their structural properties connect is syntax, the machinery by which
utterances can be computed, on the basis of lexical input. The questions then
are: How are lexicon and syntax linked precisely? What kind of information do
they exchange? What is the interface between lexicon and syntax? Which
interface levels need to be distinguished? There are several possible answers to
this question depending on ones theoretical perspective.
Within the generative, modular paradigm, Chomsky (1995: 131) distinguishes two interface levels: the level of phonetic form (PF) is the interface with
sensorimotor systems, the level of logical form (LF) is the interface with systems
of conceptual structure and language use. The two performance systems involved
are the articulatory-perceptual system and the conceptual-intentional system,
semantics
This diagram illustrates the central position of syntax, and, at the same time, it
raises the question of the relationship between lexicon and syntax. In the
chapter by Norbert Corver, this relationship was identied as the third interface
level. Corver cited Chomsky (1991, 46):
that there are three fundamental levels of representation: D-structure, PF
and LF. Each constitutes an interface of the syntax (broadly constructed) with
other systems: D-structure is the projection of the lexicon, via the mechanisms
of X-bar theory; PF is associated with articulation and perception, and LF with
semantic interpretation.
syntax
semantics
lexical items will govern how the phrase structure is implemented, notably the
argument structure and the feature composition of the lexical items is essential
to the implementation of the syntax in the language production process.
This book was designed to examine the relative contribution of these two
dimensions in a clear fashion, through illustrations of exemplary research
carried out within each paradigm and to examine how they can be made to
inter-relate in a way which would enable us to explain better the overall process
of SLA. An examination of the interface between syntax and the lexicon is both
timely and important. Both groups of researchers are now coming to an
understanding that their particular view of the world may not suce to account
for the overall process and that each will have to understand more about what
the other knows. From the point of view of the researcher interested in syntax
(generally coming from a background in linguistics), the shift of linguistic
theory away from principles and parameters and into minimalism has resulted
in a crucial increase of the signicance of the lexicon. Within minimalism so
much of the information essential for the working of the system has been
assigned to the lexicon that it has become crucial for syntacticians to reect
more on how the lexicon works. From the point of view of the researcher
interested in the lexicon (generally coming from a background in psychology),
it is important to integrate the outcome of lexical learning within the overall
acquisition process. Unless one adopts a purely connectionist position, it is
clear that the use of the lexical items studied can only take place within the
syntactic system. An understanding of the acquisition of the syntax is therefore
essential to the understanding of the whole second language processing and
acquisition. The introductory chapter by Richard Towell explicates the complementarity of linguistics and psychology in doing language research. Taking the
other chapters in this book in consideration, he balances the linguistic and
psychological dimensions of basic questions in second language acquisition
research. Towell argues that the lexical part of the lexicon-syntax interface is the
driving force of language acquisition and that we need to investigate the
psychological mechanisms behind this development.
L2 acquisition, procedural vs. declarative knowledge, structure vs. process, competence vs. performance, etc. One main conclusion is that, over and over again, we
need to plea for a cocktail of treatments, for many sorts of data, and for many sorts
of expertises to give an appropriate answer to the questions belonging to such
contrastive pairs. In this cocktail, the following distinctions can be made.
3.1 Variety in methodology
It will be clear that no single particular methodology can produce all the
answers we need. We need the linguistic analysis of spontaneous and elicitated
speech data (see the chapters by Hawkins and Liszka, Van de Craats, Corver),
but also on-line and o-line grammatical judgment tasks (see the chapters by
Dueld, and Sabourin and Haverkort). We need to carry out psycholinguistic
experiments with reaction times (see the chapters by Dueld and Dijkstra), but
also newer methodologies should be applied like eye-tracking, and neuroimaging techniques (see the chapters by Sabourin and Haverkort, and Green).
Another promising methodology is the use of computer simulations (see the
chapter by Williams).
3.2 Variety in learners and languages
The discovery and testing of general or universal principles and parameters in
language and language acquisition require the full range of learners: from real
beginners (see Corver and Van de Craats) to (near)native speakers (see, e.g.
Hawkins and Liszka), from unguided acquisition to classroom learning. At the
same time we need to consider the full language typological range, both as a
source and a target language in L2 acquisition (the chapters in this book cover
a range larger than in many other books on L2 acquisition). As for the computer simulations, dierent learning algorithms should be probed, including
algorithms where previous knowledge (L1) can be implemented to explore its
eect on acquiring another language system (L2). Without studying real
beginners, it seems not feasible to get an answer to the question of the role of
cognitive vs. linguistic principles in speaking a new language (cf. Klein and
Perdue 1997), and how, perhaps, cognitive principles are matched by linguistic
structures. Cognitive principles can be inuential in an indirect way via the
syntax-semantic interface, but maybe they directly trigger specic syntactic
mechanisms. That would contradict approaches based solely on lexical formal
features as the main sources of information for syntactic structures.
</TARGET "hou">
References
Bowerman, M. and Levinson. S. (eds). (2001). Language acquisition and conceptual development. Cambridge: Cambridge University Press.
Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge. Mass: MIT Press.
Chomsky, N. (1991). Some notes on economy of derivation and interpretation. In Principles and
parameters in comparative grammar, R. Freidin (ed.), 417454. Cambridge MA: MIT Press.
Chomsky, N. (1995). The minimalist program. Cambridge MA: MIT Press.
Jackendo, R. (2002). Foundations of language. Brain, meaning, grammar, evolution. Oxford:
Oxford University Press.
Klein, W. and Perdue, C. (1997). The basic variety. Or: Couldnt natural languages be much
simpler? Second Language Research 13, 301/347.
Muysken, P. (2000). Bilingual speech. A typology of code-mixing. Cambridge: Cambridge
University Press.
Plunkett, K. and Juola, P. (1999). A connectionist model of English past tense and plural
morphology. Cognitive Science 23, 436490.
Schreuder, R. and Baayen, H. (1997). How complex simplex words can be. Journal of
Memory and Language 37, 118139.
<TARGET "ni" DOCINFO AUTHOR ""TITLE "Name index"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Name index
A
Abutalebi 209, 214
Allen 54-56, 67, 98, 99, 124
Altarriba 143, 148
Avrutin 124, 178, 194
B
Barlow 106, 124
Bever 97, 98, 124
Birdsong 97, 122, 124, 201, 214
Bley-Vroman 46, 67, 108, 120, 125
Brown 184, 194, 195, 206, 207, 214, 215,
218
C
Cappa 20, 209, 214, 217
Carlson 114, 127
Chomsky 1-3, 6, 19, 20, 24, 35, 39, 43, 45,
50, 51, 62, 66, 67, 69, 72, 78, 79, 93,
94, 95, 115, 123, 125, 176, 194, 199,
214, 219, 220, 226
Chung 99, 125
Clahsen 94, 95, 109, 120, 125
Coppieters 97, 125
Corver 3, 5, 7, 10, 45, 48, 49, 67, 68, 70,
77, 79, 81, 85, 94, 95, 220, 223, 225
Culicover 99, 123, 125
D
De Groot 20, 140, 141, 149, 206, 213, 214,
216, 217
De Moor 139, 149
Dijkstra 3, 5, 12, 18, 129, 132-136,
138-142, 146, 148-150, 206, 214,
221, 223, 225
Dueld 3, 5, 7, 97, 107, 108, 110, 114,
116, 118, 120, 121, 124, 125, 223
Dussias 144, 149, 211, 216
E
Ellis 4, 5, 19, 97, 125, 151, 172, 200, 214,
217
F
Fabbro 198, 207, 215, 216, 218
Fodor 1, 7, 14, 19, 20, 99, 125-127, 154,
172
Font 134, 140, 149
Foster 195
Freedman 108-110, 126
Friederici 144, 145, 149, 184, 194, 208,
209, 213, 215, 217
G
Gerard 132, 134, 149
Grainger 132, 133, 149, 150
Green 3, 5, 10, 15, 16, 18, 134, 150, 197,
206, 208, 209, 215, 217, 223, 225
Greenbaum 97, 126
Grosjean 140, 141, 149, 152, 170, 172
H
Hagoort 184, 194, 195, 206, 207, 214, 215,
218
Hahne 144, 145, 149, 208, 209, 213, 215
Hamann 202, 203, 215
Hardt 115, 126
Hebb 205, 215
Hedgcock 97, 126
Hong 109, 125
K
Kemmer 106, 124
Kinder 203, 216
Kirsner 135, 148
Kluender 99, 100, 113, 126, 208, 216
Kornlt 60, 64, 68, 85, 86, 95
</TARGET "ni">
<TARGET "si" DOCINFO AUTHOR ""TITLE "Subject index"SUBJECT "Language Acquisition & Language Disorders, Volume 30"KEYWORDS ""SIZE HEIGHT "220"WIDTH "150"VOFFSET "4">
Subject index
A
abstract representations 15, 154, 159
abstraction 4, 153, 154
acceptability judgment 117, 120, 144
acquisition of vocabulary 199, 200
age of acquisition 201, 208-210, 217
agreement 22, 25, 44, 48, 61-66, 77, 78,
83, 84, 87, 93, 152, 157, 178-181,
183, 184, 190, 192, 195
alternative feature realization 61
amnesic patients 197, 198, 201-203
animacy 14, 154, 160-162, 165-168
aphasia 175, 176, 194, 195, 198, 215-218
aphasics 12, 15, 16, 176-179, 183, 184,
194, 195, 198, 215
articial grammar 153, 173, 202, 203, 216
articial language 153, 155, 172
aspect hypothesis 38, 39, 41
association 22, 131, 132, 134, 150, 170
asymmetric spell out 61
auxiliary selection 112, 113, 127
B
bilingual 20, 130-132, 134-136, 138-140,
142-150, 198, 206, 208, 214-218,
221, 225, 226
bilingual syntactic processing 146
D
declarative 4, 10, 15, 16, 18, 20, 163, 164,
174, 193, 195, 199-203, 205, 212,
216, 218, 223
declarative memory 15, 193, 199-203
derivational theory of complexity 2, 107
discourse hypothesis 39, 41
distributed morphology 23, 35, 43
distributional analysis 18, 154, 155, 157
dual competence 104, 107, 111, 116, 117,
119
Dutch go/no-go 137
C
categorical 56, 97, 99-101, 104, 106, 107,
111-114, 116-119, 121-123
causative verb 108
clitic placement 107, 108, 110, 119, 125
closed-class words 210
cognate 135, 141
common gender 179-181
E
EEG 144, 145, 175, 184-186, 194
eective connectivity 211
electroencephalography 184
English go/no-go 132, 137
English lexical decision task 131, 138-141
ERP data 186, 206-208, 213
ERPs 12, 15, 16, 130, 144, 145, 184, 186,
192, 194, 213
Event-Related brain Potentials 144, 149,
195, 215, 216
explicit learning 160, 164, 199
F
feature 9, 10, 17, 22-25, 33-37, 40-42, 48,
51-54, 56, 59-64, 66, 69, 70, 72, 73,
76, 78-81, 85-88, 91-93, 113, 117,
143, 168, 222
feature bundle 73, 76, 81, 85, 86, 88, 91,
93
formal feature 51, 73, 78, 80, 85
frequency 4, 12, 13, 18, 19, 22, 28, 29, 37,
42, 43, 111, 112, 117, 124, 129, 135,
136-139, 141-143, 147, 159, 172,
181, 182, 186
frequency of use 12, 18
frontal-basal ganglia circuit 207
functional category 9, 34
functional head 69, 78, 80, 83
I
idealisation of the data 7
implicit learning 18, 153, 159, 165-167,
169, 172, 174, 200, 214
incidental learning 154, 161, 165,
169-171, 224
initial sensitivities 205
innate knowledge 5, 225
interlanguage 4, 7, 8, 10, 17, 18, 43, 46,
48-50, 53, 65, 66, 70, 73, 94, 109
interlingual homograph 132, 134, 136,
137, 146, 148-150, 214
interpretable feature 79
invariant principles 3
invisible category principle (ICP) 59, 60
G
generalised blocking principle (GBP) 34, 36
generalized lexical decision 131, 136, 139,
141, 150
generativist 4-7, 10, 98, 101, 122, 220
go/no-go 131, 132, 137
gradience 97, 99-101, 111-113, 115, 117,
120, 125
grammar 1, 4, 8, 9, 15, 17, 19-21, 23, 26,
34, 35, 40, 43, 44, 46, 48, 51-54, 65,
67, 68, 70, 81, 88, 94, 95, 98, 100,
103, 123, 151, 153, 173, 174, 175,
176, 179, 184, 195, 197, 199, 200,
202, 203, 212, 216, 218, 219, 220,
221, 226
grammatical gender 151, 173-175, 179,
184, 191, 194, 195
grammaticalise 34
L
language decision 131, 137
language faculty 2, 6, 10, 21, 22, 24, 26,
40, 41, 45, 47, 65, 66, 98, 123
language intermixing 138, 141, 142, 149,
214
language mode 140
language non-selective access 132-134,
139, 146, 147
language processing 19, 140, 149, 150,
176, 178, 183, 184, 191, 198, 213,
218, 221, 222, 224
language processor 175
language-selective access 132
language-specic lexical decision 131
lexical decision 131, 133-136, 138-141,
150, 177, 184, 214, 215
lexical gradience 111, 112, 115
Subject index
231
</TARGET "si">
T
task learning 13, 162, 163, 167
task-dependent variation 176
terminal node 23, 35
test of knowledge of morphology 26
time-course of lexical activation 135, 147
triggering 1, 7, 8, 17, 91, 221
V
VP-ellipsis 100, 113-115, 120, 125
W
word association 131, 132, 134
word naming 131, 132
working memory 146, 163, 176