THE DIGITAL NINETEENTH

CENTURY
Series Editor: Andrew Stauffer

NINETEENTH-
CENTURY
ILLUSTRATION
AND THE DIGITAL
Studies in Word
and Image
Julia Thomas
The Digital Nineteenth Century

Series Editor
Andrew Stauffer
University of Virginia
Charlottesville
Virginia, USA
The Palgrave Pivot series publishes short-form monographs on topics at
the intersection of nineteenth-century studies and the digital humani-
ties. Partnering with the NINES Center (Networked Infrastructure
for Nineteenth-Century Electronic Scholarship) at the University of
Virginia, this series will be retrospective and prospective, involving not
only explications of digital projects and theoretical considerations of
methods, results, rhetorics, and audiences, but also projections that chart
a course for future work. The series will also include free-standing titles
for scholars throughout the world not tied to a specific digital project,
but rather synoptic studies of a particular method, approach, or thematic
in digital nineteenth-century studies. The series aims to provide a grow-
ing archival record of the digital nineteenth century across the years.

More information about this series at

http://www.springer.com/series/15607
 Julia Thomas

Nineteenth-Century
Illustration
and the Digital
Studies in Word and Image
Julia Thomas
Cardiff University
Cardiff, Wales, UK

The Digital Nineteenth Century

ISBN 978-3-319-58147-7 ISBN 978-3-319-58148-4  (eBook)
DOI 10.1007/978-3-319-58148-4

Library of Congress Control Number: 2017940600

© The Editor(s) (if applicable) and The Author(s) 2017

This work is subject to copyright. All rights are solely and exclusively licensed by the
Publisher, whether the whole or part of the material is concerned, specifically the rights
of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and
retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and
information in this book are believed to be true and accurate at the date of publication.
Neither the publisher nor the authors or the editors give a warranty, express or implied,
with respect to the material contained herein or for any errors or omissions that may have
been made. The publisher remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

Cover illustration: Pattern adapted from an Indian cotton print produced in the 19th century

Printed on acid-free paper

This Palgrave Macmillan imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
For Tim Killick, Ian Harvey and Nicky Lloyd, who helped
to make DMVI and The Illustration Archive
 Acknowledgements

This book is the product of more than a decade’s research on digital

illustration projects. I have received help and advice during this time
not only from the people with whom I have worked but also from those
whom I have met along the way in conferences and workshops. First, I
would like to thank the Arts and Humanities Research Council for sup-
porting and funding the Database of Mid-Victorian Illustration and The
Illustration Archive. Thanks also to Cardiff University for a Research
Leave Fellowship that allowed me to write this book and to the editors at
Palgrave for bringing it to publication. There are many people who have
been generous in giving feedback on these projects and have helped to
formulate the research ideas on which the book is based. In particular, I
would like to thank the team at Advanced Research Computing@Cardiff
(ARCCA), Giorgio Bacci, Neil Badmington, Giles Bergel, Charlotte
Boman, John Bradley, the team at British Library Labs, Quentin Blake,
Dean Burnett and Nathan Heslop, Luisa Calè, Abbie Enock and Felicity
Bazell at Capture, Benjamin Colbert, Simon Cooke, Damian Walford
Davies, the Digital Public Library of America, Martha Fogg, Scott
Gibbens, Paul Goldman, Mikey Goodman, Simon Grennan, Ortwin de
Graef, Alison Harvey, Ian Haywood, Ann Heilmann, Kate Holterhoff,
Claire Horrocks, Alan Hughes, Lorna Hughes, Martin Kayman, Mario
Klingemann, Dawn Knight, Lorraine Janzen Kooistra, Simon Mahoney,
Brian Maidment, Becky Munford, Sara Pepper, Mike Pidd, Etienne
Posthumus, Andrew Prescott, Susan Matthews, Mary Shannon, Bethan

vii
viii  Acknowledgements

Stevens, Jahn Thon, Rhys Tranter, Frederik Truyen, Chris Veness at

Movable Type, Martin Willis, and Alison Wray.
It has been a pleasure working on these projects with David Skilton,
Omer Rana, and Paul Rosin. I have learnt so much from them. Special
thanks to Anthony Mandal for commenting on the manuscript, founding
the CEIR library, letting me tap into his digital networks, and for being
my companion on numerous excursions. The success of DMVI and The
Illustration Archive owes so much to Tim Killick, Ian Harvey, and Nicky
Lloyd, the research associates who have shaped these projects. They may
have moved on to other things, but I will always be thankful for their
insight, knowledge, commitment and humour. This book is dedicated to
them. The birth of these digital resources has coincided with the birth of
my two boys, Jude and Gabriel. A final thanks to them and to Stuart for
keeping the domestic show on the road while I’ve been on the digital one.
 Contents

1 Frontispiece 1

2 (In)visibility 15

3 Searchability 33

4 Crowdsourcing 65

5 Tailpiece 95

Bibliography 
105

Index 
115

ix
 List of Figures

Fig. 2.1 Titlepage for Sir Gilbert Edward Campbell, Bart.,

Mysteries of the Unseen; or, Supernatural Stories of English Life
(London: Ward, Lock and Co., 1889). Available on
The Illustration Archive: http://illustrationarchive.
cardiff.ac.uk 17
Fig. 3.1 John Everett Millais, illustration for Harriet Martineau,
The Anglers of the Dove, Once a Week 7 (16 August 1862)
p. 197. Engraved by Joseph Swain. Available on the
Database of Mid-Victorian Illustration: www.dmvi.org.uk 34
Fig. 3.2 Arthur Boyd Houghton, ‘True or False?’, illustration
for Adelaide Anne Proctor, ‘True or False?’, Good Words
3: 12 (December 1862) p. 721. Engraved by the Dalziel
Brothers. Available on the Database of Mid-Victorian
Illustration: www.dmvi.org.uk 45
Fig. 3.3 Detail from Myles Birket Foster, illustration for Robert
Bloomfield, ‘The Farmer’s Boy’, Favourite English Poems
(London: Sampson Low, 1862) p. 174. Engraved
by Edmund Evans. Available on the Database
of Mid-Victorian Illustration: www.dmvi.org.uk 49
Fig. 3.4 Arthur Boyd Houghton, illustration for J. Hamilton Fyfe,
‘About Toys’, Good Words 3: 12 (December 1862) p. 753.
Engraved by the Dalziel Brothers. Available
on the Database of Mid-Victorian Illustration:
www.dmvi.org.uk 58

xi
xii  List of Figures

Fig. 4.1 Charles Bennett, illustration for ‘Paterfamilias Reading

the “Times”’, London Society 2 (September 1862)
facing p. 233. Engraved by the Dalziel Brothers.
Available on the Database of Mid-Victorian Illustration:
www.dmvi.org.uk 67
Fig. 4.2 Screenshot from the ‘Tag Illustrations’ page,
The Illustration Archive: http://illustrationarchive.
cardiff.ac.uk 82
Fig. 5.1 ‘The End’, tailpiece from Edward Lamplough,
Yorkshire Battles (Hull: William Andrews and Co., 1891).
Available on The Illustration Archive: http://
illustrationarchive.cardiff.ac.uk 102
 CHAPTER 1

Frontispiece

Abstract  The introduction establishes the significance of this book as the

first extended analysis of the intersection between nineteenth-century illus-
tration and the digital. Drawing on a range of nineteenth-century source
materials and twenty-first-century digital archives, Thomas establishes a
conceptual model for understanding how its migration to a digital envi-
ronment challenges our perception of illustration and the relation between
word and image. In a discussion that takes place within the framework of
digital humanities and illustration studies, the introduction argues that it is
as much in the development of the digital resource as in the end product
that new ways of viewing historic illustrations are generated.

Keywords  Nineteenth century · Illustration studies · Digital

humanities · Book history

This book, perhaps, is an illustration. Although I cannot claim that what

follows ‘makes brilliant and illustrious’ or offers ‘spiritual enlightenment’
(two of the original meanings of the word ‘illustration’), the book does
adopt another sense of this plural signifier: to ‘illuminate’ or ‘light up’.
Its objective is to illuminate nineteenth-century illustrations, to engage
critically with these unseen images through the digital mechanisms that
render them visible again.

© The Author(s) 2017 1

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4_1
2  J. THOMAS

The digital might appear an alien environment for historic illustrations

that are seemingly tied to their material contexts—the books, periodicals,
magazines and newspapers in which they were originally published—but
it has already been proved remarkably accommodating in the form of
online illustration archives like the William Blake Archive (http://www.
blakearchive.org), the Rossetti Archive (http://www.rossettiarchive.org)
and The Yellow Nineties Online (http://www.1890s.ca). In some cases,
the only practical way that illustrations can be accessed is through digital
resources. Eduardo Urbina, the editor of the Cervantes Project (http://
cervantes.tamu.edu), which is a digital archive that includes tens of thou-
sands of illustrations of Don Quixote, points out that many illustrated
editions of this text are rare and held in different repositories around the
world, while the quantity of illustrations has meant that there is no com-
plete catalogue. A digital environment, Urbina suggests, allows ‘for the
kind of access and knowledge until now unimaginable’.1 Indeed, the dig-
ital format promises to emphasise the significance of illustration in a way
that, paradoxically, print has not. Jerome McGann points to the printed
editions of the poems that Letitia Elizabeth Landon wrote to accompany
images, which often leave out the visual contexts within which Landon
was working, including the specific images that she ‘illustrated’ with her
words and the space of the illustrated gift book where her picture–poems
were published.2
The digital, however, is also a challenging environment for illustra-
tions, as McGann recognises in his reflections on the creation of the
Rossetti Archive. Illustrations are the pictorial obstacles that computa-
tional tools come up against. While tools for analysing and data mining
nineteenth-century texts are becoming ever more sophisticated,3 those
for analysing nineteenth-century images are falling behind. The prob-
lem with images is that they cannot be treated in the same computational
way as texts: they cannot be marked up, retrieved or ‘mined’ like words.
As McGann puts it, ‘while digital texts lie open to automated search and
analysis, digital images do not. Consequently, our critical commentaries
[in the Rossetti Archive] never adequately reflected the reality we knew
was there’.4
This book is all about the mysterious ‘reality’ that is illustration. It
draws on a range of nineteenth-century source materials and twenty-
first-century digital archives in order to offer the first extended articula-
tion of the intersection between the digital and historic illustrations and
to develop new theoretical paradigms for understanding this intersection.
 1 FRONTISPIECE  3

The fundamental question that the book addresses is: how does its
migration to a digital environment challenge and change our under-
standing of nineteenth-century illustration, both in terms of its cultural
importance and the way that it signifies? A key component of this signifi-
cation is the dialogue between word and image that illustration puts into
play. W.J.T. Mitchell has claimed that ‘all media are mixed media’, there
is no purely visual, or, for that matter, verbal genre.5 But illustration is
more of a mixed medium than most. To turn to a related theoretical
discourse, illustration is the ‘intermedial’ genre that Irina O. Rajewsky
defines as ‘media combination’, where at least two distinct media com-
bine: ‘These two media or medial forms of articulation are each present
in their own materiality and contribute to the constitution and significa-
tion of the entire product in their own specific way’.6 It is the ‘media
combination’ of word and picture that distinguishes illustration, I would
argue, from Rajewsky’s other example of intermedial relations: ‘medial
transposition’, where a media product is transformed into another
medium (a book, for example, is turned into a film).7 I want to make this
distinction clear from the outset because illustration has often been (mis)
understood in precisely these terms, with the ‘original’ text regarded
as the creative inspiration for and source of the image into which it is
transposed. Illustration here is little more than a pale reflection of the
originary words. However, this is to misconstrue how nineteenth-cen-
tury illustration works. The text is not always the primary, or even prior,
partner in the illustrative relation, as Landon’s picture–poems written
to accompany images indicate. Moreover, an illustration is not simply a
‘transposition’ or ‘transformation’ of the words, but stands in a nuanced
relation of complementarity and conflict, sameness and difference.
Notions of what an illustration is and how it makes its meanings have
been central to my research for some years, but these notions were thrown
into relief during the development of the Database of Mid-Victorian
Illustration (DMVI) (http://www.dmvi.org.uk) and The Illustration
Archive (http://illustrationarchive.cardiff.ac.uk). Both of these resources,
funded by the Arts and Humanities Research Council in the UK, make
nineteenth-century illustrations available online, but their content and
methodologies are very different. Work on DMVI began in 2004 at a time
when major digitisation projects were underway (in December of that
year, Google announced its ‘Library Project’).8 DMVI was funded as part
of a ‘Resource Enhancement’ programme, the resource that it enhanced
being a little-known collection of illustrations from periodicals of the
4  J. THOMAS

1860s and 1870s housed in the School of Art Gallery and Museum in
Aberystwyth University. In fact, so little known was this collection that
it had languished in a disused lift shaft for over 30 years.9 The project
involved travelling from Cardiff to Aberystwyth (not always easy, as any-
one who has attempted the journey by train will know) and setting up a
flatbed scanner in the reading room of the gallery. We could almost have
been Google. The project was not only concerned with the digitisation of
these illustrations, though. A major part of the remit was to explore ways
of making the content of the illustrations searchable so that the images
could be retrieved in terms of what they depicted. After considering sev-
eral image classification systems, we decided to create a bespoke keyword-
ing structure and to mark up each illustration according to this schema. It
was this aspect of the project that turned out to be the most challenging,
even more than the commute from Cardiff to Aberystwyth.
One advantage of our frequent train journeys was the time that it gave
us to think through and discuss ideas. A founding principle of DMVI
was agreed upon while the team was laden with scanning equipment and
waiting for a train to Oxford to see the Forrest Reid collection of illus-
trations in the Ashmolean Museum: the decision to include illustrations
published in 1862, a year that was chosen because it was particularly rich
in terms of illustrated books and periodicals. There were pragmatic rea-
sons for this decision: this was a 3-year project and our ambitions had
to be realised within this fixed timeframe. But there were also theoreti-
cal considerations: we wanted the database to provide a snapshot of the
range of literary illustrations published in a period that was, arguably,
one of the most significant in the history of British book illustration, and
this would not have been possible if we had focused on, say, the work of
a specific artist, illustrator or author. The choice of 1862 (which encom-
passed illustrated novels that began or ended their serialisation in this
year) meant that we could include more familiar illustrated works, like
Frederic Leighton’s illustrations for George Eliot’s Romola and John
Everett Millais’ illustrations for Anthony Trollope’s Orley Farm, along-
side the less familiar images that accompanied poems and short stories in
the popular Victorian periodicals, London Society, Once a Week and Good
Words. As it stands today, the database contains 868 illustrations from lit-
erary works, all of which are described bibliographically and iconographi-
cally (in terms of what the images depict), enabling the user to search
across multiple fields.
 1 FRONTISPIECE  5

The Illustration Archive had a different genesis. This was the output of
a Big Data project that aimed to make searchable over a million illustra-
tions from books in the British Library’s collection. The decade between
the development of DMVI and The Illustration Archive was one that
marked a defining shift from what Bethany Nowviskie describes as the
‘boutique digitization’ undertaken by humanities scholars to the mass
digitisation projects carried out by large corporations.10 The images in
The Illustration Archive had already been digitised by Microsoft; so when
we embarked on this project in 2014 there was no setting up camp with
a scanner in the reading room of the British Library. As we soon discov-
ered, though, big data presents its own transportation issues. For a start,
there is too much of it to transfer electronically. The data was copied on
to five external hard drives and these drives travelled by courier from the
British Library down the M4 to Cardiff. When they arrived, we had very
little idea of the illustrations they contained. The books in which the
images appeared covered ‘history’, ‘geography’, ‘literature’ and ‘philoso-
phy’; but these categories were established in the 1850s when they were
much broader. We were dealing, then, not only with considerably more
illustrations than in DMVI, but also different types and taxonomies of
illustrations (the focus of DMVI was on illustrations that accompanied
‘literary’ texts), although the main objective of the two projects was the
same: to create a searchable archive of illustrations that would be freely
accessible online. Because of the number of illustrations in the British
Library dataset, it was impossible to keyword all of them in-house. So,
the project developed the infrastructure for crowdsourced tagging while
exploring the potential for computer recognition of image content. The
resource that we created (a web-accessible front-end to a database of
illustrations, bibliographic metadata and descriptive tags) is currently the
largest searchable online resource dedicated to illustrations.
While these projects provide the foundation for the argument in this
book, the discussion moves beyond their specific mechanisms and out-
comes to develop a way of conceptualising the relation between the digi-
tal and illustration studies. Since I first made the case for the recognition
of illustration studies as a discipline ‘with its own methodologies, vocab-
ularies and critical strategies’, the agenda for this discipline has been set
out by a number of critics.11 Paul Goldman suggests that illustration
studies should involve a re-evaluation of the significance of illustration
across academic institutions and the development of courses devoted to
the subject.12 The images that might form the basis of illustration studies
6  J. THOMAS

are vast in type and number. Illustration crosses different periods and
genres and the potential modes of analyses that such studies embrace are
equally varied, from a close analysis of the illustrations that accompany
a particular work to an examination of illustration as a cultural practice
and its place in the history of the book. However, the critical emphasis
of illustration studies is a specific one, which, as David Skilton argues,
can be distinguished ‘from the more limited objectives of bibliographi-
cal description and aesthetic appreciation’.13 Put simply, the objective of
illustration studies is to analyse how illustrations signify, how they make
their meanings and how these meanings are embedded in the historical
moment of their production and reception.
This objective is not as straightforward as it may initially seem because
the study of illustration needs to begin by advocating the relevance of
illustration and countering the all-too-common misconceptions to be
found in scholarly work, the sorts of misconceptions that view illustra-
tion as a straightforward example of Rajewsky’s ‘medial transposition’, or
worse. Some of the typical fallacies about illustration are as follows: that
illustrations are purely decorative or ornamental; that illustrations are
supplements or ‘added extras’ and can, therefore, be extracted from the
text without loss; that illustrations have a subordinate relation to the text;
that the text exists prior to the illustration; that the text is the primary
means of signification; that the illustration replicates or reflects textual
meanings. References to ‘mere illustration’ are scattered throughout crit-
ical articles and books, so much so that they are simply taken for granted.
It is unsurprising, then, that these errors have found their way into the
field of digital humanities. Martyn Jessop, for example, uses this familiar
notion of illustration as a means of establishing his idea of ‘visualization’:

The term ‘Illustration’ implies an image which serves only to support writ-
ten language; thus the main carrier of information is the associated text not
the image. Many consider that an image can only be a true visualization
when it is the primary carrier of information, not a supplement to a piece
of text. I shall settle on a distinction in which an illustration is intended
merely to support a rhetorical device (usually textual) whereas a visualiza-
tion is intended either to be the primary rhetorical device or serve as an
alternative but parallel (rather than subordinate) rhetorical device.14

I use this quote not to undermine Jessop’s otherwise very interesting

ideas about visualization, but to expose the flawed concept of ‘illus-
tration’ that is presented as a given. Here is another example from the
 1 FRONTISPIECE  7

book, Digital Humanities: illustration is ‘the employment of a graphical

feature, photograph, map, or other representational device to elucidate,
explain, or show something in a text … the text still assumes priority, and
the illustration is meant to summarize an argument, provide a reference
point, or corroborate the text’.15
Such comments are part of a discourse about illustration that has its
roots in the nineteenth century. While this period is often seen as one
in which authors embraced illustration (Dickens and Thackeray are usu-
ally given as examples), many writers and critics expressed alarm at the
rise of illustration and took up their pens against the pencil, arguing that
illustration was inherently inferior to literary texts. As Gustave Flaubert
declared, ‘Jamais, moi vivant, on ne m’illustrera, parce que la plus belle
description littéraire est dévorée par le plus piètre dessin’.16 Henry
James shared this aversion to illustration, commenting that words and
pictures should not be served ‘on the same platter’ any more than fish
and flesh.17 These views of illustration are far from objective: they betray
an anxiety about the impact that the presence of pictures might have on
words. Ironically, what this antagonism suggests is that illustration has
a far more complex identity and interaction with the text than these
remarks imply. Illustration does not (indeed, it cannot) ‘summarize’ the
text because it is a distinct medium with its own conventions, traditions
and signifying practices. Illustration adds to or detracts from the words;
it emphasises or undermines them, but it does not merely re-produce or
reflect them. There is always a semantic gap between the illustration and
the text because they are different modes of representation, a difference
that is often exposed at the very moment that the illustration seems most
faithfully to depict what the text inscribes. Ironically, Jessop’s definition
of ‘visualization’ provides a neat description of the nineteenth-century
illustration: as ‘an alternative but parallel (rather than subordinate)’
genre.
The initial call for illustration studies took place against the backdrop
of the launch of DMVI and the belief that ‘new technologies can func-
tion alongside scholarly work to illuminate the complexities of illustra-
tion’.18 The ‘new’ technologies may no longer be as new as they once
were, but the conviction remains that the digital offers extensive pos-
sibilities for studying illustration. It is no coincidence that the growing
academic interest in nineteenth-century illustration, evidenced in recent
publications, conference papers and networks, has emerged alongside,
and out of, digital illustration resources.19 Sandro Jung comments on the
8  J. THOMAS

advances in illustration studies that have taken place in tandem with digi-
tal initiatives such as DMVI. These resources, Jung asserts, ‘have con-
tributed significantly to shedding light on, and to the understanding of
the importance of, corpora and illustration techniques that did not fea-
ture in traditional accounts of literature and book history’.20 Likewise,
Christina Ionescu points to the increased availability of image databases
and digitised editions as a key component in the growth of studies in the
eighteenth-century book illustrations.21
As Ionescu recognises, the digital resource allows a greater number of
historic illustrations to be seen than ever before. But it does more than
this. The digital archive is not simply a transparent vehicle through which
illustrations are made accessible, but a critical and analytic space in which
they are continually re-viewed. This re-viewing, moreover, takes place in
the creation of the digital resource as well as in the final product. It is
important to make this point because digital humanities has sometimes
been regarded, even among exponents, as a largely ‘instrumentalist’ field
in which the traditional interpretative questions of the ‘humanities’ (that
is, of literary scholarship) are forsaken in favour of the project-based
creation of digital tools and archives. A recent article in the Los Angeles
Review of Books makes this point: ‘What Digital Humanities is not about,
despite its explicit claims, is the use of digital or quantitative methodolo-
gies to answer research questions in the humanities. It is, instead, about
the promotion of project-based learning and lab-based research over
reading and writing’.22 It is true that a humanities research agenda has
not always been evident in digital projects that ‘apply’ innovative com-
putational tools, sometimes arbitrarily, to research areas. It is also true
that applications of computational tools, even to big data, can be narrow
in their conceptual approaches and research implications. But this is not
always the case. An alternative view sees the humanities and the digital
as mutually reinforcing principles, with digital humanities initiatives at
their strongest when the computational infrastructure and the humani-
ties research agenda are interdependent.23 Indeed, this book argues that
there is not a clear line, never mind a binary opposition, between the
development of the computational infrastructure and the research ques-
tions: these research questions are articulated as much in the creation of
the digital resource as in what the computational ‘tools’ make possible.
Questions of what an ‘illustration’ is, how it signifies, and even what
it means to say that a picture ‘illustrates’ a text, are fundamental, then,
to the development of digital illustration archives. The discussion that
 1 FRONTISPIECE  9

follows attempts to replicate something of this dynamic by focusing on

the logistical considerations involved in constructing digital illustration
resources alongside (and, indeed, as inseparable from) the theoretical
implications that underpin their methodologies. As both developers and
users, the digital form challenges our conception of historic illustrations,
moving beyond conventional models to emphasise the role of illustra-
tion as a signifying practice, how and what it means in the context of
the dialogue between pictures, words and culture. And the digital cre-
ates new dialogues, too. In this space, illustrations interact not only with
their ‘original’ texts but also with other digitally-born texts: with tags,
metadata and the language of the interface. An illustrative relationship
between text and image is generated in the very environment that ena-
bles an exploration of this relationship.
This intimate relationship between the digital and illustration studies
has evolved not only because the digital offers new possibilities for the
analysis of illustration, but also because the digital space is itself an illus-
trated one. Jay David Bolter has commented on the way that computers
are easily able to combine word and image because both are represented
in the same binary code. The result is that in ‘electronic writing’ images
and texts ‘belong to the same space’ and are able to cross traditional
domains: pictures can become textual symbols, and texts can become
pictures.24 In some senses, however, the distinctly illustrative character
of the digital has become so naturalised that it is frequently overlooked.
There is a tendency to regard more ‘recent’ modes of representation
like photography and cinema as providing the origins of new media;25
but illustration also provides a model, if not the model. There have been
hints at the parallels between these seemingly disparate fields. Mary
E. Hocks and Michelle R. Kendrick open the edited collection, Eloquent
Images (still one of the only books to deal with the relation between
word and image in the digital) by arguing that new media should not
be seen as a ‘battleground’ but as an ‘interplay that already exists and
has always existed between visual and verbal texts’.26 One of the most
prominent locations for this ‘interplay’ is illustration and, while Hocks
and Kendrick do not mention this genre, their contention that new
media can be thought of as a ‘hybrid’ situates illustration as an, albeit
absent, touchstone in their discussion. I would go further and suggest
that nineteenth-century illustration and contemporary digital cultures
are, in fact, part of the same historical trajectory, their constant repro-
duction and remediation stimulated by new possibilities for juxtaposing
10  J. THOMAS

word and image, and this juxtaposition takes place in spite of Friedrich
Kittler’s assertion that ‘computers … are not designed for image-process-
ing at all’.27 Computers might not be designed for image processing, but
perhaps it is this very fact that has spurred on attempts to bring word
and image together in productive, and explicitly ‘illustrative’, ways. Take,
for example, the creation of Mosaic, the web browser that for the first
time enabled the integrated display of word and image and led to the
popularity of the World Wide Web, a software development that bears
more than a passing resemblance to wood engraving, which for the first
time allowed pictures and letterpress to be printed simultaneously on the
same page and led to the burgeoning of illustration.28
The emergence of interface studies and its focus on the configura-
tion of the GUI (graphical user interface) and WIMP interface (win-
dows, icons, menus and pointers) may offer opportunities for exploring
the illustrative dialectic that is ever present on our screens.29 Johanna
Drucker has denaturalised the workings of the interface by stressing
its graphic, and therefore visual, nature: ‘We tend to think of graphi-
cal interface as the screen display, a portal into the online world with
menu bars, buttons, and icons to manipulate. As a result, we ignore its
graphicality, its constructedness, the very features that support its opera-
tions and make it work’.30 But the interface is not ‘purely’ graphical, if,
indeed, there is ever such a thing. These graphical features have a textual
component. Interfaces are a space where word and image are in constant
dialogue. This can be seen in our everyday encounters with the digital
where we come across pictorial capital letters, like Apple’s ‘A’ for its ‘App
Store’, which is made up of a pencil, paintbrush and ruler, and Google’s
ever-changing logo that makes pictures out of, around and between the
letters. Website menus consist of thumbnail images with text under-
neath, while websites themselves are combinations of words and images.
In social media, there are illustrated tweets, tagged images and, in the
case of platforms like Tumblr, pictures with captions. The blogging con-
vention of heading a post with an illustration comes straight out of the
Victorian illustrated press. Even with other developments—audio and
moving components, the tactile experience of touch screens—the appear-
ance of the digital, what we see, is a combination of text and image,
an Illustrated User Interface. This is not just a matter of appearances.
It is in this combination of word and image that the interface generates
meanings, meanings that are bound up in cultural values and ideologies.
What does it signify that an icon for an app store is composed of the
 1 FRONTISPIECE  11

tools of writing, painting and geometry, or that a blog post about lib-
erty is preceded with a picture of an American flag? An understanding of
the significance and workings of historical illustrations might help us to
understand and interpret this illustrated digital environment.
In 1999, Christian Vandendorpe made the astute prediction that hyper-
media would lead to a dismantling of the hierarchy that privileges text
over image and give rise to more interactions between the two forms:

In a civilization marked by the increasing prominence of the visual, we can

expect a change in this hierarchical relationship of text to image. It is far
from certain that coming generations dealing with mixed environments
will read the text first, as we so often tend to do. On the contrary, feedback
loops will proliferate between text and image, and there will be more influ-
ence between them …31

The world of hypermedia, a world that we now inhabit, is a distinctly

illustrated one, but the characteristics that Vandendorpe describes also
bear witness to past ‘civilizations’ and, in particular, to the rise of illustra-
tion in the nineteenth century. This was a culture where the hierarchical
relationship between word and image was constantly overturned, where
the text was not necessarily read first, and where the ‘influence’ between
text and image was etched on almost every printed page.
It is this, often disorienting, illustrated world that a digital environ-
ment makes manifest and that the following chapters set out to explore.
The next chapter identifies the invisibility of nineteenth-century illustra-
tions in contemporary print culture and theoretical discourse and analy-
ses the implications of making these images visible in a digital format.
The process of digital transformation, I suggest, leads to a direct con-
frontation with ideas of the visual specificity of illustration. Chapter 3
looks at the methods available for making the content of illustrations
searchable online, including ‘computer vision’, which offers the pos-
sibility of automated image retrieval, the use of textual metadata (bib-
liographic information, captions and the words that accompany the
illustration) and keywording. It also argues that these methods are not
detached or abstracted from the material that they promise to make
searchable, but are deeply implicated in the relation between word and
image that characterises illustration. Chapter 4 looks at the problemat-
ics of crowdsourced image tagging in terms of the role of the tagger,
the infrastructure of the digital illustration resource and the definition of
12  J. THOMAS

what constitutes an ‘illustration’. Crowdsourced tagging is defined as a

space where meanings are contested and where illustration refuses to be
fixed, even when this fixing is a requirement of the system. Finally, the
‘tailpiece’ turns to the conceptual paradigms opened by modes of visi-
bility and searchability in digital illustration resources, suggesting that a
digital environment results in a radically new way of viewing illustration
that is defined here as ‘affillustration’, a term that captures the interplay
or ‘kinship’ between illustrations, their allusions and references to other
illustrations, and their place in a distinct iconographical tradition.
Overall, this book argues for an intersection between the digital and
illustration studies by drawing together these apparently distinct areas.
Each chapter suggests how the concerns of the digital map onto the
concerns of illustration studies and vice versa: the process of digital
remediation and image processing is bound up in the visual identity of
illustration (Chap. 2); the mechanisms for making image content search-
able are intertwined with the politics of word and image (Chap. 3); and
crowdsourced tagging is an activity that generates questions about what
an ‘illustration’ is (Chap. 4). What I am offering, then, is a conceptual
and critical model that brings these two fields to bear on each other and
that engages with and understands nineteenth-century illustration in the
intricacies of its interaction with the digital. Like the online archives and
resources that it analyses, this book aims to take the dark art of illustra-
tion out of the shadows and place it in the, albeit flickering, luminosity
of the computer screen.

Notes
1. Eduardo Urbina, ‘Visual Knowledge: Textual Iconography of the
Quixote’, Don Quixote Illustrated: Textual Images and Visual Readings/
Iconografía del Quijote, ed. Eduardo Urbina and Jesús G. Maestro,
Biblioteca Cervantes 2 (Pontevedra: Mirabel Editorial, 2005), pp. 15–38.
Available online at http://cervantes.tamu.edu/V2/CPI/iconography/
publ.html.
2. Jerome McGann, Radiant Textuality: Literature after the World Wide Web
(New York and Basingstoke: Palgrave, 2001), pp. 63–67.
3. See, for example, Andrew Stauffer’s introduction and the accompany-
ing forum, ‘Evidence and Interpretation in the Digital Age: Searching
Engines, Reading Machines’, Victorian Studies, 54:1 (2011): 63–68.
4. McGann, Radiant Textuality, p. 173.
 1 FRONTISPIECE  13

5. W.J.T. Mitchell, ‘There Are No Visual Media’, Journal of Visual Culture,

4:2 (August 2005): 257–266, p. 260.
6. Irina O. Rajewsky, ‘Intermediality, Intertextuality, and Remediation: A
Literary Perspective on Intermediality’, Intermédialités: histoire etthéorie
des arts, des lettres et des techniques/Intermediality: History and Theory of
the Arts, Literature and Technologies, 6 (2005): 43–64, p. 52.
7. Rajewsky, ‘Intermediality, Intertextuality, and Remediation’, p. 51.
8. DMVI went live in 2007 and the database was updated with new features
in 2011.
9. For more information about the Aberystwyth collection, see Robert
Meyrick, ‘“Spoils of the lumber-room”: Early collectors of wood-
engraved illustrations from 1860s periodicals’, Reading Victorian
Illustration, 1855–1875: Spoils of the Lumber Room, ed. Paul Goldman and
Simon Cooke (Aldershot: Ashgate, 2012), pp. 179–199, pp. 194–196.
10. Bethany Nowviskie, ‘Resistance in the Materials’, Between Humanities
and the Digital, ed. Patrik Svensson and David Theo Goldberg
(Cambridge, MA and London: MIT Press, 2015), pp. 383–389, p. 386.
11. Julia Thomas, ‘Reflections on Illustration: The Database of Mid-Victorian
Wood-Engraved Illustration’, Journal of Illustration Studies, 1 (December
2007). Available online at http://jois.uia.no/articles.php?article=37. See
also the ‘Editorial’ at http://jois.uia.no/articles.php?article=42.
12. Paul Goldman, ‘Defining Illustration Studies’, Reading Victorian
Illustration, ed. Goldman and Cooke, pp. 13–32.
13. David Skilton, ‘Illustration Studies and the Infinite Archive’, Journal of
Illustration Studies, 2 (December 2013). Available online at http://jois.
uia.no/articles.php?article=50.
14. Martyn Jessop, ‘Digital visualization as a scholarly activity’, Literary and
Linguistic Computing, 23:3 (2008): 281–293, p. 283.
15. Anne Burdick, Johanna Drucker, Peter Lunenfeld, Todd Presner, Jeffrey
Schnapp, Digital_Humanities (Cambridge, MA: MIT Press, 2012),
p. 43.
16. Gustave Flaubert to Ernest Duplan, 12 June 1862, Flaubert Correspondance,
vol. III (janvier 1859–décembre 1868), ed. Jean Bruneau (Paris: Gallimard,
1991), p. 221.
17. Henry James, preface, The Golden Bowl, ed. Virginia Llewellyn Smith
(1909; Oxford and New York: Oxford University Press, 1963), p. xlvi.
18. Thomas, ‘Reflections on Illustration’.
19. I am thinking, in particular, of the creation of the Romantic Illustration
Network, Illustr4tio, and the increasing number of panels devoted to
illustration in international conferences including those hosted by the
British Association for Romantic Studies and the British Association for
14  J. THOMAS

Victorian Studies. There has also been a noticeable rise in the number
of special journal issues devoted to illustration, such as the ‘Illustration
and Gender’ issue of Nineteenth-Century Gender Studies (Summer 2015)
and the ‘Picturing the Eighteenth-Century Novel’ issue of the Journal for
Eighteenth-Century Studies (December 2016).
20. Sandro Jung (ed.), British Literature and Print Culture (Cambridge: D.S.
Brewer, 2013), pp. 1–2.
21. Christina Ionescu (ed.), Book Illustration in the Long Eighteenth Century:
Reconfiguring the Visual Periphery of the Text (Cambridge: Cambridge
Scholars, 2011), pp. 1–2.
22.  Daniel Allington, Sarah Brouillette and David Golumbia, ‘Neoliberal
Tools (and Archives): A Political History of Digital Humanities’, Los
Angeles Review of Books, 1 May 2016.
23. For a discussion of how recent work in digital humanities is turning away
from a preoccupation with process and instrumentalism to an engage-
ment with the theoretical, see Anthony Mandal, ‘Digital Humanities’,
The Year’s Work in Critical and Cultural Theory, ed. Neil Badmington
and David Tucker, vol. 24 (Oxford: Oxford University Press, 2016), pp.
317–342.
24. Jay David Bolter, Writing Space: The Computer, Hypertext, and the History
of Writing (New Jersey and London: Lawrence Erlbaum Associates,
1991), p. 72.
25. See, for example, Lev Manovich, who traces the intersections between cin-
ema and new media in The Language of New Media (Cambridge, MA:
MIT Press, 2001).
26.  Mary E. Hocks and Michelle R. Kendrick, ‘Introduction: Eloquent
Images’, Eloquent Images: Word and Image in the Age of New Media, ed.
Mary E. Hocks and Michelle R. Kendrick (Cambridge, MA: MIT Press,
2003), pp. 1–16, p. 1.
27. Friedrich Kittler, Optical Media: Berlin Lectures 1999, trans. Anthony
Enns (Cambridge: Polity, 2010), p. 226.
28. I am grateful to Michael Goodman for drawing this to my attention.
29. For an intervention in studies of the interface, see Alexander R. Galloway’s
political analysis in Alexander R. Galloway, The Interface Effect (Cambridge:
Polity, 2012).
30. Johanna Drucker, Graphesis: Visual Forms of Knowledge Production
(Cambridge, MA: metaLABprojects and Harvard University Press,
2014), p. 138.
31. Christian Vandendorpe, From Papyrus to Hypertext: Toward the Universal
Digital Library, trans. Phyllis Aronoff and Howard Scott (1999; Urbana:
University of Illinois Press, 2009), p. 96.
 CHAPTER 2

(In)visibility

Abstract  This chapter takes as its starting point the centrality of illustra-

tion in nineteenth-century culture and its subsequent disappearance from
printed editions and critical discourse today. Digital illustration archives
have in some ways restored this visibility, but their methodologies can
be problematic. Thomas identifies these problems, arguing that the
process of digital remediation leads to a direct confrontation with ideas
about the visual specificity of illustration. Referring to numerous digital
resources, including the William Blake Archive and The Yellow Nineties
Online, this chapter looks at how illustrations are put on display. This
display can mark a radical break with the print format, but it also involves
a negotiation of what constitutes the difference of illustration and the
visual components that define it.

Keywords  Nineteenth century · Illustration · Digital archives

Digitisation · Visuality · Wood engraving

Something mysterious has happened to nineteenth-century illustrations

(Fig. 2.1). The images that once swelled the pages of books and maga-
zines have vanished. Engravings, etchings, lithographs, photographs …
tens of thousands of images have all but disappeared from view. The only
fragments of this lost empire of illustrations are the few that have clung
on to visibility—Tenniel’s Alice, Cruikshank’s Fagin—the ghostly traces
of a world that was once alive with illustrative pictures.

© The Author(s) 2017 15

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4_2
16  J. THOMAS
 2 (IN)VISIBILITY  17

t Fig. 2.1  Titlepage for Sir Gilbert Edward Campbell, Bart., Mysteries of the
Unseen; or, Supernatural Stories of English Life (London: Ward, Lock and Co.,
1889). Available on The Illustration Archive: http://illustrationarchive.cardiff.
ac.uk

Illustrations dominated the cultural landscape of the nineteenth cen-

tury. Indeed, the landscape itself was beautifully illustrated. Travel writ-
ing, popular novels and children’s fiction all came with pictures, as did
Shakespeare’s works, the Bible and scientific treatise. This proliferation
of illustration, facilitated by new methods of reproduction like wood
engraving that made illustrations easier and cheaper to print, meant that
the genre crossed social divides.1 By the middle of the century, illus-
tration was no longer a feature solely of the gentleman’s library; it had
moved into more humble abodes. Disarmingly democratic and startlingly
visible, illustration pictured everything and drew everyone (in). The
Victorians could not escape illustration, even if they wanted to.
To say that the Victorian world was a visual culture, a precursor to
our own, is not entirely accurate. This was an illustrated culture. In these
years, it was this specific mode of representation that was the dominant
visual form, a fact that was recognised by contemporary commentators:
‘The pictorial printing-press is now your only wear!’ wrote one critic in
1844. ‘Everything is communicated by delineation. We are not told, but
shown how the world is wagging’.2 As this remark suggests, not only was
this period characterised by the thousands of illustrative pictures in cir-
culation, but these images were also a vehicle for self-definition, a visual
mechanism by which the Victorians shaped and made sense of their envi-
ronment. They turned themselves into illustrations, using the few square
inches of the woodblock to depict their lives, their fashions, their feats of
engineering, their houses and gardens, their wars, their politics and their
values.
An engagement with Victorian culture needs to take account of this
vast array of illustrations. The problem, however, is that they are often
nowhere to be seen. Modern editions of Victorian novels are rarely
published with the images with which they appeared, an exclusion that
leads to some striking textual anomalies. Anthony Trollope’s invoca-
tion to the reader of Orley Farm to ‘go back’ and look at one of John
Everett Millais’ illustrations makes little sense when the images are not
present in the text.3 Likewise, William Makepeace Thackeray’s frequent
18  J. THOMAS

references to his own illustrations are redundant in those editions of

the novels where the images are excised. An illustrated footnote added
by Thackeray to Vanity Fair, which comments on the gap between the
historical setting of the text and the up-to-date fashions depicted in his
illustrations is unintelligible if the pictures are removed.4 The solution
to this problem, and one that has been resorted to since the earliest unil-
lustrated editions of Vanity Fair, is to cut the references to the illustra-
tions from the text, but the traces of the illustrations cannot be so easily
erased; they seep into the very fabric of a novel that, in its original subti-
tle, Pen and Pencil Sketches of English Society, drew attention to the coex-
istence of word and image.5
Such examples not only point to the glaring absence of the illustra-
tions in editions of the novels published today, but also suggest that,
far from a marginal accessory, illustration was integral to the Victorian
novel, to the extent that it influenced the writing of texts as well as the
reading of them. It is an awareness of the significance of illustration in
shaping the words of these Victorian ‘classics’ that is lost along with their
pictures. And if this invisibility is the fate of the illustrations by high-
profile artists accompanying canonical works, there is little hope for the
less well known. The lowly status of illustration, or, at least, of popular
illustration, means that it is not high on the list of conservation priorities.
While rare and valuable illustrated books (hand-coloured travel or medi-
cal books, for instance) are usually well preserved in special collections
and archives, mass-produced material is decaying. Paul Goldman has
argued that few efforts have been made to exhibit or even properly care
for collections of illustration in museums and libraries, with the result
that ‘survival in good condition of such items is patchy at the least’.6 The
torn title page of Mysteries of the Unseen, the illustration that heads this
chapter, is emblematic of the current state of much Victorian illustration,
which is destined forever to be unseen.
It is not just illustrations that are invisible, but also a scholarly engage-
ment with them. Those researchers who have worked on historic illus-
trations have done so against the grain. The genre has been largely
ignored in those disciplinary fields that should know illustration better:
art history (where illustration is usually mentioned only if the illustra-
tor also happens to be a painter); literary studies (where the exclusion
of illustration from the analysis of texts gives us only half of the story);
book history (there is no mention of illustration in Robert Darnton’s
communications circuit7); digital humanities (which, despite fostering
 2 (IN)VISIBILITY  19

pioneering illustration projects, remains doggedly text-based in its critical

focus); and media studies (even Friedrich Kittler’s Optical Media is full
of holes when it comes to illustration8).
Illustration is theory’s blind spot: it is everywhere and nowhere. Take,
for instance, the writings of Ferdinand de Saussure. These were gathered
together by Saussure’s students after his death and provided the starting
point for the tenets of (post)structuralism. They also happen to include
illustrations, although these have rarely been examined as such.9 In fact,
the English translations of Course in General Linguistics, along with the
majority of critical works on Saussure, refer to the illustrations as ‘dia-
grams’, even though Saussure often uses the French word ‘figure’.10
The replacement of ‘figure’ and its connotations of showing, appear-
ing, representing, with ‘diagram’, situates these images securely within
a scientific discourse and outside the illustrative play of word and image.
Paradoxically, despite its visual presence in Saussure’s text, illustration
remains stubbornly invisible. In what is perhaps Saussure’s most repro-
duced illustration, an ellipse with a picture of a tree in its top half and
‘arbor’, the Latin word for tree, in the bottom, the picture of the tree is
not defined as a picture at all but as a marker of the ‘signified’ or ‘con-
cept’ (in the parallel ellipse, this picture is actually replaced by the word
‘tree’).11 To the extent that there is an ‘image’ in Saussure’s formula-
tion, it lies, ironically, with the ‘signifier’, the word ‘arbor’, which rep-
resents the ‘sound-image’. For all its obscuring of illustration, however,
Saussure’s ellipse stands as a meta-illustration, a comment on the interac-
tion between word and image in illustration, the dividing line between
the picture of the tree and the word ‘arbor’ indicating the gap between
them, ‘the opposition that separates them from each other’.12 Although
the word and picture come to seem inseparable, the connection between
the two is not natural, innate or self-evident. Rather, like Saussure’s
notion of the signifier and signified, the relation between word and
image in illustration is ‘arbitrary’ and has to be learned.13 It is the teach-
ing of the ‘proper’ interaction between word and image that is a driving
factor in children’s picture books today and, indeed, is a characteristic
feature of nineteenth-century illustrated books for children, where the
words directly refer to what is going on in the picture (‘Here is a pretty
wagon with horses of great size and strength’, ‘See, here is a ferry-man
in his boat’14).
Whilst the study of illustration has some significance for the issues of
meaning production that are central to critical theory, it remains one
20  J. THOMAS

of the most enigmatic forms of representation, and the one that seems
most resistant to analysis. Part of the reason for this murkiness is that
illustration has conventionally fallen between generic categories, seen
neither as text nor as image. Writing in 1928, Forrest Reid, one of the
first collectors of mid-Victorian illustration, recognised the problematic
status of illustration: it ‘has come to be regarded’, he states, ‘as a dubious
mixture of art and something that is not art’.15 In some ways, it is also
the very ubiquity of illustration in the nineteenth century, with which
Reid, as a collector, had to contend, that has made its analysis so pro-
hibitive. Gleeson White, another early collector, described his task as a
‘magnificent’ one in which the piles of manuscripts at his side ‘prove the
impracticability of the enterprise’.16 Illustrations are simultaneously invis-
ible and omnipresent: forgotten today, yet everywhere in nineteenth-
century culture, to the extent that there are too many to examine with
any degree of thoroughness. Gérard Genette acknowledges this in his
account of the paratext, where he deals only briefly with illustration
because the field is too large. Illustration, in his words, is an ‘immense
continent’.17 One of the main difficulties with illustration, however, is
that it is not a continent at all: it does not occupy a distinct, continuous
space. On the contrary, illustration is dangerously incontinent: an unruly
and hybrid form that refuses to be fixed.
Critics have tried to find their way around this ‘immense continent’
using the foundational map of the archive. Reid and White were com-
pelled to create their own archives by cutting the illustrations from mid-
nineteenth-century publications and using these archives as the bedrock
for their research.18 White designs an artist-centred catalogue, moving
diligently through the images that appeared in illustrated weekly papers,
some illustrated books and the key periodicals of the 1860s, including
Once a Week, the Cornhill Magazine, Good Words and London Society,
his guiding principle being to detail the work of ‘every artist of the first
rank’.19 Reid takes a similar approach, creating a ‘survey’ of the illustra-
tions of 58 British artists working in the 1860s, which is arranged by
artist and school, including ‘The Pre-Raphaelite Group’ and ‘The Idyllic
School’.20
As the examples of these early collectors suggest, the content and
organising principles of the physical archive go hand in hand with the
scholarship on illustration. The same can be said of illustration studies,
which, out of necessity, have been governed by the images that are made
available and easily identifiable in the physical archive. It is unsurprising,
 2 (IN)VISIBILITY  21

then, that discussions of nineteenth-century illustrations have focused

on canonical authors (Trollope, Dickens, Hardy, Eliot, Wilde), artists
(Millais, Tenniel), periods (the Romantic period, the 1860s, the fin de
siècle), types of publication (periodicals, magazines, gift books), indi-
vidual works (Alice’s Adventures in Wonderland, Moxon’s edition of
Tennyson’s Poems) and places of publication (Britain, France). While such
accounts of illustration undoubtedly cover a lot of ground, they depend
on textual rather than visual information: bibliographic metadata, the
publishing details listed in library catalogues.
If what is analysed and understood about illustration depends on the
principles and shape of the archive, what shifts when we move from a
physical archive to a digital one? How does a digital environment organ-
ise and define illustration? In some respects, many digital illustration
repositories are not so different from physical ones in their focus on sin-
gle authors (the William Blake Archive, the Rossetti Archive, Illustrating
Scott (http://illustratingscott.lib.ed.ac.uk), Visual Haggard (http://
www.visualhaggard.org)), specific publications (The Yellow Nineties
Online) and dates of publication (DMVI). Deriving from what illustra-
tions are available for digitisation and what illustrations are ‘known’,
digital archives often seem to replicate bibliographic imperatives, with
the result that illustrations from less familiar texts continue to remain
unseen. As Tara McPherson has recently warned, ‘We must not assume
that digitization will adequately capture the richness and diversity of the
cultural record…. We should participate in and guide decisions about
what will get digitized, ensuring that digitization does not simply rein-
stall the absences and imbalances of our physical archives within digital
realms’.21 Equally, however, there is a case to be made not just for ‘guid-
ing’ future digitisation but also for retrieving those illustrations that have
already been digitised, but are, to all intents and purposes, lost. This
was the objective of The Illustration Archive, where the largely ‘random’
dataset digitised by Microsoft eschews the idea of an illustrative canon.
While acknowledging the potential pitfalls of digital archives, it is
also important to recognise the fact that a digital environment makes
a greater number of illustrations accessible than ever before. Digital
archives currently display only a fraction of the illustrations that were
in circulation at the time, but they give a spectacular insight into the
nineteenth-century illustrated world. Taken collectively, such resources
reveal the scale of this ‘immense continent’ in a way that is impossible
in a physical archive. These resources do not simply make illustrations
22  J. THOMAS

accessible, though: they make them visible, emphasising their difference

as visual objects. It is this emphasis on the visual specificity of illustration
that unites digital illustration archives, despite their varying methodolo-
gies and content. It is an emphasis, moreover, that sets these resources
apart from other, more ‘general’, digital image repositories and hosting
sites like Flickr’s The Commons (https://www.flickr.com/commons).
The main advantage of uploading historic images onto Flickr, and the
reason that many museums and libraries have followed this route (includ-
ing the British Library, which uploaded the dataset that we used in The
Illustration Archive) is that the material is instantly available for the
public to see and make use of. But this apparent accessibility is not all
it seems. Melissa Terras laments that it is ‘nigh on impossible to navi-
gate or search Flickr in any meaningful way’.22 For Terras, this is largely
the fault of Flickr’s interface, but the problem also stems from the fact
that finding relevant images relies on the metadata attached to them, and
this metadata is highly variable. As Patricia Harpring notes, ‘Retrieval of
appropriate images depends on intelligent indexing, which one might
call the “language” of retrieval’.23 In Flickr, the inconsistency and insta-
bility of the ‘language of retrieval’, which comes about because the
images are tagged differently by different contributors, means that rel-
evant images will not always be retrieved. This is especially the case with
illustration because the word ‘illustration’ is rarely used in a folksonomic
markup that tends to focus on what the image depicts rather than what
it is. Thus, while it might be relatively easy to find thousands of pictures
of cats in Flickr, it is not so easy to find thousands of illustrations of cats.
Indeed, in some respects, Flickr’s cats are all the same. The site advo-
cates sameness, equality, congruity and the commonality of images, with
the result that a photo of my neighbour’s cat is on the same level of
‘image-ness’ as the painted cat in David Hockney’s Mr and Mrs Clark
and Percy, or the illustrated cat that dangles over a goldfish bowl in
Hood’s Own: or, Laughter from Year to Year (1855). J. Hillis Miller has
warned that ‘digital reproduction … may be in danger of putting eve-
rything on the same plane of instant availability’.24 Such technological
advancements, he asserts, can compromise notions of cultural specificity
that lie at the heart of cultural studies, or they can lead to an over-gen-
eralised model of cultural difference. It could be argued that Flickr poses
these ‘dangers’ by threatening to lose sight of historic illustrations at
the very moment that they are brought into view. Images here have the
same value and identity as other images: they are generic digital ‘images’,
 2 (IN)VISIBILITY  23

or, rather, ‘photographs’. As the homepage states: ‘The key goal of The
Commons is to share hidden treasures from the world’s public photog-
raphy archives. Please help make the photographs you enjoy more dis-
coverable by adding tags and leaving comments. Your contributions and
knowledge make these photos even richer.’25 On a day that I happen to
be browsing, the ‘photographs’ that are displayed on the homepage are
the book illustrations from the British Library’s dataset. Some are steel
engravings, some are etchings, others are wood engravings, and they are
all, of course, digital images; but none are ‘photographs’.
I am not simply being pedantic. The dilution of generic difference on
Flickr might not affect the average oil painting, but it does have reper-
cussions for historical illustrations, which are already marginalised. While
a general hosting platform like Flickr sidesteps the specificity of illustra-
tion, specialist digital illustration resources have directly to confront and
negotiate the issue of what constitutes the visual difference of an illus-
tration. In a sense, there is no more to illustration than meets the eye.
Illustrations are inescapably visual, a visuality that stands out in the nine-
teenth-century book where the pictures are often printed on thicker, yel-
low paper, with tissue interleaves separating them from the text, and in
landscape format, which means that the book has to be turned around
to view them. Indeed, Leah Price’s contention that ‘The Victorians
cathected the text in proportion as they disowned the book’ goes some
way towards explaining the deep-rooted suspicion of illustration that
emerged alongside its proliferation because the presence of illustrations
inevitably draws attention to the book as a book.26 There are other visual
constituents that define illustrations in their printed forms: the features
of the method by which they are reproduced (the scratched lines of an
etching, the cross hatching of an engraving, the porous texture of a lith-
ograph); and the size of these images (their conformity to the dimen-
sions of a page, a plate, a woodblock, although blocks were often bolted
together to make a larger image).
But, while there might be a specific ‘look’ to an illustration that dis-
tinguishes it from, say, a painting, its visual identity is never fixed or
static. On the contrary, an illustration has multiple visual incarnations:
it is the artist’s design (which is lost when the image is drawn or pinned
onto the woodblock), the engraved woodblock or etched/engraved
plate (which could go through several modifications and corrections),
the proof engraving on India paper, and the stereotype or electrotype
made from the block. Even in its printed form, the visual specificity of
24  J. THOMAS

illustration slips precariously around: the image changes from copy to

copy as the impression wears down; illustrations that are originally pub-
lished in magazines, periodicals or monthly parts take on a new visual
dimension when they are bound in books; and illustrations that started
off as engravings can turn into photographs or, conversely, photographs
can be recast as engravings.27
And then there is the illustration’s relation to the text. The differ-
ence of an illustration is located not only in the visuality of the image
itself, but also in the visuality of its conjunction with the words. The play
of word and image in illustration is pictorial as well as semantic, with
vignettes encroaching on textual borders, pictorial letters exposing the
graphic nature of words and wood-engraved images wrapping themselves
parasitically around the text.28 Even when an illustration is cut off from
the text, or ‘clipped’ in the style of Forrest Reid and Gleeson White,
there is a sense in which it retains the spectral trace of its missing limb.
An amputated illustration seems to announce its incompleteness, the fact
that it is (or was) part of a narrative trajectory, and this is the case even
in those illustrations that have only the most tenuous and enigmatic rela-
tion with the actual words (I am thinking of Alvin Langdon Coburn’s
illustrations for Henry James’ novels, the illustrations that appear in The
Yellow Book, decorative ‘ornaments’). Illustrations are pictorial fragments,
in some ways viable as independent works of art, yet always lacking.
The visual specificity of illustration comes to the fore in the digital
archive because the digital does something different with illustrations: it
puts them on display. In this respect, the visuality of illustration gener-
ated in the digital archive is starkly at odds with the visuality embodied in
an illustration’s analogue existence. Simon Cooke sees the illustrated gift
book of the 1860s as a ‘clash of outer and inner’, a space of disjunction
and mismatch between the fine gilt and coloured bindings of the exterior
and the black and white illustrations inside.29 However, there is a sense
in which this aesthetic disjunction can never actually be viewed without
a concomitant temporal disjunction: the black and white illustrations are
hidden when one looks at the binding, and the binding is hidden when
one looks at the illustrations. The problem with illustrations is that they
are not prints: they appear inside books, newspapers and magazines, and
this material location renders them invisible. After all, a book is closed
more often than it is open. Illustrations are concealed between the cov-
ers, the visual riches inside only hinted at by descriptive titles or ornate
 2 (IN)VISIBILITY  25

bindings. As Paul Goldman succinctly puts it, ‘In libraries, illustrations

are virtually ignored, encased as they are in books and periodicals’.30
Books might be closed, but the digital archive is always open. Digital
illustration resources make these ‘encased’ images visible in new ways.
This is not simply a matter of digitally ‘translating’ or ‘remediating’ his-
toric illustrations. The digital display of illustrations involves a considera-
tion of how they should be displayed and of what aspects of their visual
identity should be emphasised or effaced. The digitisation of illustra-
tion (and I am using the term ‘digitisation’ in its broadest sense here
to signify the scanning, image processing and mechanics of display that
transform an analogue illustration into a digital one) inevitably involves a
negotiation of what constitutes the difference of illustration, of the visual
components that define it. Such a negotiation takes place at the level of
the individual image (the extent to which it should be ‘cleaned up’, for
example) as well as at the level of the broader editorial principles that
shape the digital repository. Scanning illustrations, for instance, involves
choices about the resolution of the image and the level of detail that
can or should be seen, with most resources retaining their TIFF image
files for archival purposes and displaying the online images as JPEGs. Of
course, the scanning process is not always in the domain of the developer
of the resource, as the example of The Illustration Archive indicates; but
even where the digital images have come ready-made, there are still deci-
sions about how to display them. Digitisation and its associated tools set
out the parameters of an illustration’s visuality; they determine its visual
presence. The digital archive, in effect, establishes what is visually signifi-
cant and distinct about illustration.
In the case of The Yellow Nineties Online, which contains editions of
late nineteenth-century aesthetic periodicals, the specificity of illustra-
tion lies in a material integrity that is deferred to and emulated in the
‘minimal’ image processing. As the editors comment, ‘Visual images
have been minimally edited using Photoshop to adjust colour and reso-
lution in order to enhance accuracy of representation’.31 This ‘accuracy
of representation’ characterises the facsimiles of the periodicals avail-
able on the site, which include digitised images of the tissue interleaves
and a FlipBook tool that gives users the impression that they are turn-
ing the pages. Melissa Terras reminds us that the ‘digitized representa-
tion of an original analogue object is not a replacement for the object’,
but the methodologies of The Yellow Nineties Online signal an active
engagement with the difference of an illustrated text—a difference that
26  J. THOMAS

is located in the physicality of the entire periodical.32 In the words of

the editors, ‘Preserving the physical features of each periodical in virtual
form, together with paratexts of production and reception (such as cover
designs, advertising materials, and reviews), enables users to analyze the
significance of each periodical’s materiality as well as its content’.33 It is
an objective that takes on some importance because these particular peri-
odicals tend to be located in rare books and private collections and are
made out of pulp paper that is fast disintegrating. The digital editions are
the closest some users will get to the actual periodicals.
There are other competing definitions of the visual specificity of illus-
tration, though. With a nod towards archives like The Yellow Nineties
Online, the Victorian Illustrated Shakespeare Archive (www.shakespea-
reillustration.org) also includes minimally edited or ‘raw’ digitised
images of the illustrations that accompanied four Victorian editions of
Shakespeare’s works, but these images are juxtaposed with heavily pho-
toshopped and, in some cases, cropped versions of the same images. The
result is unsettling, partly because of the nostalgia that this juxtaposition
elicits for the ‘original’ illustration with its foxing and show-through let-
terpress, and partly because this juxtaposition raises questions about the
identity of an illustration, of what constitutes its visual core or essence,
and at what point an illustration becomes different from itself.
In the William Blake Archive, the visual difference of illustration is
the defining principle of a project that aims to ‘restore historical balance’
in a scholarly and print culture in which Blake’s illustrations have been
neglected.34 As the editors note, the tradition of editing Blake has been
‘overwhelmingly literary’, with the illustrations ‘largely invisible’ because
they are too expensive to be reproduced in print form.35 The digital
archive rectifies this situation by focusing on the visual integration of text
and image that informs Blake’s work: ‘we emphasize the physical object –
the plate, page, or canvas – over the logical textual unit – the poem or other
work abstracted from its physical medium’.36 These principles shape the
structure of this archive, from the XML framework to the positioning of
the user in relation to the texts and pictures:

The part-to-whole path reinforced by print – which typically starts with a

reading of Blake’s ‘poems’ (often, in fact, transcriptions extracted from illu-
minated pages) and may or may not move along to a later, secondary look
at ‘illustrations’ (which often turn out to be a predetermined editorial selec-
tion of the pictures that seem most relevant to the words) – is reversed.37
 2 (IN)VISIBILITY  27

The idea of illustration that informs the architecture of this digi-

tal archive has a direct bearing on the status of the genre. By taking
into account the ‘whole’ rather than the ‘part’, the illustrations in the
William Blake Archive are not secondary, but an integral and equal part
of this whole, and the justification for this is found in the visual specific-
ity of Blake’s illuminated pages. Blake’s digital illustrations have a holis-
tic relationship to the text: they are joined together with the words and
should not be torn asunder. Ironically, it is the books and printed edi-
tions of Blake’s works that have marginalised a visuality that is ‘restored’
by the digital, a factor emphasised in the software tools that this archive
pioneered, including the Java applet, ImageSizer, which allows for a
calibration so that Blake’s works are displayed at their true physical size
on the user’s monitor. N. Katherine Hayles has argued that this simula-
tion of visual accuracy in the William Blake Archive is deeply problem-
atic because it downplays the differences between printed and electronic
editions of Blake (changing how the work means, Hayles argues, also
changes what it means).38 Of course, Hayles has a point, but the agenda
of the William Blake Archive needs to be seen as a crucial interven-
tion that prioritises the illustrative dimension of Blake’s works and does
so by using the digital as a vehicle for emphasising the visuality of the
illustrations, a visuality that has frequently been marginalised in printed
editions. The difference of the digital in the William Blake Archive is
foregrounded in the very fact that it brings these images to light.
Tools like ImageSizer and FlipBook indicate how digital resources
can point towards notions of the analogue visuality that defines illustra-
tion, while also hinting at the possibilities of a digitally generated visu-
ality. During the development of the Rossetti Archive, Jerome McGann
distorted and ‘deformed’ Rossetti’s images in a way that revealed new
relationships between the colouring, pattern and shapes of the images.
‘The deformed images’, McGann comments, ‘suggest that computer-
ized art editing programs can be used to raise our perceptual grasp of
aesthetic objects’.39 More recently, Lev Manovich has developed meth-
ods for comparing visual patterns in big image datasets using automatic
image processing to identify the features in pages from Manga comics
(the ‘features’ here include contrast, texture, lines, curvature, shape,
brightness and colour) and then organising these images in visualisations.
Such an approach enables viewing of an individual image in the context
of bigger, overarching, patterns; a million images morph into a single,
and radically other, image. As Manovich writes, ‘This would enrich our
28  J. THOMAS

understanding of any single artefact because we would see it in relation

to precisely delineated larger patterns’.40 The seeing of illustration in
relation to ‘larger patterns’ was also central to The Illustration Archive,
although not quite in the sense that Manovich describes. A negotiation
with the visual specificity of illustration in this Archive did not lie in the
scanning of the images (over which we had no control), but in the dis-
play of the images, with the multiple display options foregrounding the
ways in which illustrations signify in their relation to other illustrations
(I will say more about this in the ‘Tailpiece’).
At times, however, the visuality of illustration that is exposed by the
digital is not so much about ‘larger’ patterns, but about those minute
details that are often overlooked. Michael Goodman, the developer of
the Victorian Illustrated Shakespeare Archive, has experimented with
Photoshop to turn digitised wood-engraved illustrations into negative
images. This is a relatively straightforward task but the result is aston-
ishing because the negative image shows what the picture on the wood-
block might have looked like before it was printed.41 Wood engravers
themselves produced what could be called ‘negative’ images, engraving
the illustration in reverse so that it appeared the right way around on the
page, and cutting out the white space, leaving the black lines to be inked
standing in relief. Goodman’s ‘negative’ illustration brings to the fore an
aspect of the visuality of illustration that is not usually ‘seen’: the process
of wood engraving, which has more or less disappeared from view today
in a focus on what the illustrations show rather than how they show it.42
The invisibility of wood engraving is not simply a product of twenty-
first-century modes of viewing, however. It is a comment on the historic
role of the engraver, which in Victorian Britain was largely that of a ‘fac-
simile’ draughtsperson, who was responsible for faithfully copying (as
opposed to interpreting) the artist’s design, a role that William Vaughan
has called ‘an excruciating feat of self-effacement’.43 The Database of
Mid-Victorian Illustration partially reverses this process of self-efface-
ment by locating the visual specificity of the illustrations in their identity
as wood engravings, a factor that is emphasised in the use of the mag-
nification tool, Zoomify (www.zoomify.com), which ‘tiles’ the images,
allowing the user to zoom in on its constituent parts. The magnifica-
tion of a wood-engraved illustration, especially when displayed on a large
monitor, never fails to raise a gasp from viewers, who can see for the first
time the complexities of the lines, those details that cannot be viewed
with the naked eye.
 2 (IN)VISIBILITY  29

Zoomify, along with Goodman’s negative illustration, might be the

tools of what J. David Bolter and Richard Grusin have called ‘remedia-
tion’, tools that enable old forms to be assimilated in new media,44 but
these tools work here in reverse, harking back to a (digital rendition of a)
visuality that existed prior to the printed illustration: the image as seen
and worked upon by the engravers, who often used magnifying glasses
as they cut the blocks.45 What I have suggested here is that the processes
and methodologies for making illustration digitally visible are part of an
active engagement with questions of what constitutes the visual specific-
ity of illustration. As James Mussell comments, ‘Every digitization pro-
ject is also an editorial project and all editorial projects must define in
some way what it is they edit’.46 The definitions of illustration articulated
in digital illustration resources might be plural and competing, but they
mark attempts to tease out the visual nuances of illustrations, to elucidate
where their difference lies, and to solve the mysteries of the unseen.

Notes
1. Richard Maxwell argues that there are three main factors that explain the
burgeoning of illustration in these years: experimentation and innovation
in book layout and typography, which led to mass-market publishing; the
rise of the notion of British literature as an institution; and a subliminal
resistance to the presence of images in books that actually spurred on
invention. Richard Maxwell (ed.), introduction, The Victorian Illustrated
Book (Virginia: University of Virginia Press, 2002), pp. xxi–xxx.
2. [Catherine Gore], ‘The New Art of Printing. By a Designing Devil’,
Blackwood’s Edinburgh Magazine, 55:339 (January 1844): 45–49, p. 47.
3. The illustrations are not present in the edition from which I am quoting.
Anthony Trollope, Orley Farm, ed. David Skilton (1861–1862; Oxford:
Oxford University Press, 1985), 2 vols., vol. ii, p. 230.
4. William Makepeace Thackeray, Vanity Fair. A Novel Without a Hero
(London: Bradbury and Evans, 1848), p. 56.
5. For a discussion of the exclusion of the illustrations from the Penguin
Classics edition of Vanity Fair (2001), see Peter L. Shillingsburg,
‘Practical Editions of Literary Texts’, Variants 4: The Book as Artefact:
Text and Border, ed. Anne Mette Hansen, Roger Lüdeke, Wolfgang
Streit, Cristina Urchuaguía and Peter Shillingsburg (Amsterdam and New
York: Rodopi, 2005), pp. 29–55, pp. 36–40.
6. Paul Goldman, ‘Defining Illustration Studies’, Reading Victorian
Illustration, 1855–1875: Spoils of the Lumber Room, ed. Paul Goldman
and Simon Cooke (Aldershot: Ashgate, 2012), pp. 13–32, pp. 17–18.
30  J. THOMAS

7. Robert Darnton, ‘What is the history of books?’, Daedalus, 111:3 (1982):

65–83, p. 68.
8. Kittler is one of the few media theorists to mention copperplate engrav-
ing, albeit in passing, but he excludes illustration from major nineteenth-
century artistic practices. His comment that ‘After 1839, there were …
two options for images: either to paint or to photograph them’ com-
pletely elides the domination of wood engraving in the middle of the
century. Friedrich Kittler, Optical Media: Berlin Lectures 1999, trans.
Anthony Enns (Cambridge: Polity, 2010), p. 136.
9. Although these aspects of Saussure’s work have not been studied as ‘illus-
trations’, they have, occasionally, been analysed as visual or iconic inter-
ventions. Sung-Do Kim, for example, discusses the ‘graphical’ nature
of Saussure’s writing in ‘La Raison Graphique de Saussure’, Cahiers
Ferdinand de Saussure, 61 (2008): 23–42.
10. The preference for ‘diagram’ occurs several times in the translated text.
See, for example, Saussure’s comment, ‘que nous figurerons comme
suit’, which turns into the rather awkward formulation ‘which I shall dia-
gram as follows’. Ferdinand de Saussure, Course in General Linguistics,
ed. Charles Bally and Albert Sechehaye, trans. Wade Baskin (New York:
McGraw-Hill, 1959), p. 12.
11. Saussure, Course in General Linguistics, p. 67.
12. Saussure, Course in General Linguistics, p. 67.
13. Saussure’s illustration can also be seen as part of a history of graphical
forms of knowledge, in which, as Johanna Drucker points out, images
of trees assume a prominent position. See Johanna Drucker, Graphesis:
Visual Forms of Knowledge Production (Cambridge, MA: metaLABpro-
jects and Harvard University Press, 2014), pp. 95–105.
14. These examples, which are typical of the time, are taken from Nurse
Rockbaby’s Easy Reading, and Pretty Pictures (London: Dean and Son,
1853), pp. 4, 10.
15. Forrest Reid, Illustrators of the Eighteen Sixties (1928; Toronto: Dover,
1975), p. 2.
16. Gleeson White, English Illustration, ‘The Sixties’: 1855–1870 (1897; Bath:
Kingsmead Reprints, 1970), p. x.
17. Gérard Genette, Paratexts: Thresholds of Interpretation, trans. Jane E. Lewin
(1987; Cambridge: Cambridge University Press, 1997), p. 406.
18. See White, English Illustration and Reid, Illustrators of the Eighteen Sixties.
19. White, English Illustration, p. x.
20. Reid, Illustrators of the Eighteen Sixties, p. 2.
21. Tara McPherson, ‘Post-Archive: The Humanities, the Archive, and the
Database’, Between Humanities and the Digital, ed. Patrik Svensson and
David Theo Goldberg (Cambridge, MA and London: MIT Press, 2015),
pp. 483–502, p. 490.
 2 (IN)VISIBILITY  31

22. See Melissa Terras, ‘Reuse of Digitised Content (1)’, blogpost, 6 October

2014, http://melissaterras.blogspot.co.uk/2014/10/reuse-of-digitised-
content-1-so-you.html.
23. Patricia Harpring, ‘The Language of Images: Enhancing Access to
Images by Applying Metadata Schemas and Structured Vocabularies’,
Introduction to Art Image Access: Issues, Tools, Standards, Strategies, ed.
Murtha Baca (Los Angeles: Getty Research Institute, 2002), pp. 20–39,
p. 20.
24. J. Hillis Miller, Illustration (London: Reaktion, 1992), p. 46.
25. https://www.flickr.com/commons.
26. Leah Price, How to Do Things with Books in Victorian Britain (Princeton
and Oxford: Princeton University Press, 2012), p. 4.
27. Before photomechanical techniques for printing photographs were
advanced, it was common for artists to trace photographs, which would
then be engraved for publication. See Gerry Beegan, The Mass Image:
A Social History of Photomechanical Reproduction in Victorian London
(Basingstoke: Palgrave Macmillan, 2008), p. 62. Even when these tech-
niques became more sophisticated, photographs were often turned into
wood engravings for the mass market. For details of Julia Margaret
Cameron’s photographs undergoing this fate, see Lorraine Janzen
Kooistra, Poetry, Pictures, and Popular Publishing: The Illustrated Gift
Book and Victorian Visual Culture 1855–1875 (Athens, OH: Ohio
University Press, 2011), p. 234.
28. For an account of how the Illuminated Magazine structured its pages
around the shapes that could be made with wood engravings, see Brian
Maidment, ‘The Illuminated Magazine and the Triumph of Wood
Engraving’, The Lure of Illustration in the Nineteenth Century: Picture
and Press, ed. Laurel Brake and Marysa Demoor (Basingstoke: Palgrave
Macmillan, 2009), pp. 17–39, p. 27, passim.
29. Simon Cooke, ‘A Bitter After-Taste: The Illustrated Gift Book of the
1860s’, Reading Victorian Illustration, ed. Goldman and Cooke,
pp. 53–78, pp. 57–58.
30. Goldman, ‘Defining Illustration Studies’, pp. 17–18.
31. ‘Editorial Principles’, The Yellow Nineties Online, ed. Dennis Denisoff and
Lorraine Janzen Kooistra, http://www.1890s.ca/Editorial_Principles.aspx.
32. Melissa Terras, ‘Cultural Heritage Information: Artefacts and Digitization
Technologies’, Cultural Heritage Information: Access and Management,
ed. Ian Ruthven and G. G. Chowdhury (London: Facet, 2015), pp. 63–88.
33. ‘Editorial Principles’, The Yellow Nineties Online.
34. ‘Editorial Principles: Methodology and Standards in the Blake Archive’,
William Blake Archive, ed. Morris Eaves, Robert N. Essick and Joseph
Viscomi, http://www.blakearchive.org/blake/public/about/principles/
index.html.
32  J. THOMAS

35. ‘Editorial Principles: Methodology and Standards in the Blake Archive’.

36. ‘Editorial Principles: Methodology and Standards in the Blake Archive’.
37. ‘Editorial Principles: Methodology and Standards in the Blake Archive’.
38. N. Katherine Hayles, My Mother Was a Computer: Digital Subjects and
Literary Texts (Chicago and London: University of Chicago Press, 2005),
pp. 90–91.
39. Jerome McGann, Radiant Textuality: Literature after the World Wide Web
(New York and Basingstoke: Palgrave, 2001), p. 85.
40. Lev Manovich, ‘How to Compare One Million Images?’,Understanding
Digital Humanities, ed. David M. Berry (Basingstoke: Palgrave
Macmillan, 2012), pp. 249–278, p. 252.
41. In a similar vein, a future project might take advantage of the sophistica-
tion of 3D printing and use the illustration to create a physical manifesta-
tion of the block from which the image was printed.
42. This exclusion of wood engraving is being rectified in a digital pro-
ject undertaken by Bethan Stevens in which the albums of the Dalziel
Brothers, one of the most prominent engraving companies of the nine-
teenth century, are being digitised, made searchable, and mounted on the
British Museum’s Collection Online site, http://www.britishmuseum.
org/research/collection_online/search.aspx.
43. William Vaughan, ‘Facsimile Versus White Line: An Anglo-German
Disparity’, Reading Victorian Illustration, ed. Goldman and Cooke
pp. 33–52, p. 34. For another account of the shift from an ‘interpretive’
engraving method, where engravers had more artistic license, to facsimile
engraving, see Beegan, The Mass Image, pp. 56–61.
44. J. David Bolter and Richard Grusin, Remediation: Understanding New
Media (Cambridge, MA: MIT Press, 1999).
45. So prolific was the practice of using a magnifying glass, whether the
engraver needed it or not, that the celebrated wood engraver, John
Jackson, made a point of warning against it on the basis that it was injuri-
ous to the sight. John Jackson, A Treatise on Wood Engraving, Historical
and Practical (London: Charles Knight, 1839), pp. 649–651.
46. James Mussell, The Nineteenth-Century Press in the Digital Age
(Basingstoke: Palgrave Macmillan, 2012), p. 4.
 CHAPTER 3

Searchability

Abstract  This chapter focuses on methods for making the content of

illustrations searchable online, including computer vision, which offers
the possibility of automated image retrieval, the use of textual metadata
(bibliographic information, captions, and the words that accompany
the illustration), and keywording. Thomas contends that these methods
are not detached from the material that they promise to make search-
able, but are deeply implicated in the relation between word and image
that characterises illustration. The chapter analyses the ways in which
the search mechanisms in digital illustration resources engage with this
problematic relation between the textal and the visual, from issues sur-
rounding the instability of the vocabulary used to describe the presence
of illustrations in books, to questions of whether an illustration ‘reflects’
the text it accompanies.

Keywords  Nineteenth century · Illustration · Digital · Computer

vision · Keywording · Bibliographic metadata

There is always a possibility, even in a digital environment, that illustra-

tions will remain unseen. The accessibility of the images depends upon
the mechanisms and systems that enable them to be searched by the user,
and this searchability, in turn, impacts on how illustration is understood
as well as how it can be studied (Fig. 3.1).

© The Author(s) 2017 33

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4_3
34  J. THOMAS

Fig. 3.1  John Everett Millais, illustration for Harriet Martineau, The Anglers of
the Dove, Once a Week 7 (16 August 1862) p. 197. Engraved by Joseph Swain.
Available on the Database of Mid-Victorian Illustration: www.dmvi.org.uk

The potential of the digital to allow illustrations to be interrogated and

retrieved in terms of their iconographic content, what they depict, not only
 3 SEARCHABILITY  35

situates these images within their historical context, but also has implica-
tions for the relation between word and image that illustration puts into
play. This relation is one that defines illustration, generates its meanings,
and determines how it is viewed, yet it has received scant attention from
critics. Lorraine Janzen Kooistra, one of the few scholars to examine the
interaction between word and image in the context of illustration of the
1890s, remarks that ‘Although most studies of illustrated books pay lip
service to the significant interactions of image and text, few critics have
gone further to investigate just how these interactions work, or to theo-
rize visual/verbal relationships’.1 Kooistra proposes a notion of ‘bitextual-
ity’ that accounts for the dialogue between word and image through an
engagement with the sexualised discourse in which illustration is described.
As Kooistra’s definition implies, the relation between word and image
in illustration is bound up in power relations. I have also discussed this
idea elsewhere, arguing that the interaction between text and picture in
illustration is frequently articulated in hegemonic terms, with the picture
positioned as subordinate to an authoritative text.2 A digital environment
is not free from these hierarchies. On the contrary, they are embedded in
the act of digitisation itself. When Microsoft set out to scan the books in
the British Library that form the basis of The Illustration Archive, it was
the words that it wanted to capture. The illustrations just happened to be
there. The identification and isolation of the illustrations in the digitised
and OCRed (Optical Character Recognition) pages were enabled not so
much because of their visual difference but because of their negative rela-
tion to the text: they were recognised as images because they were not
words. The only metadata recorded for the illustrations were the dimen-
sions of the page scan in pixels and the page number on which the image
appears as counted from the front cover (not conforming to the actual
pagination of the books).
Within this signifying system of OCRs and scans, it is the text that is
the primary signifier: it is the sought after product, the positive term by
which the illustration is defined. In this sense, a digital environment seems
to support conventional ideas of illustration that position the words as
superior to the images. As Kooistra puts it, ‘Since the image has no indi-
vidual significance, but only gains meaning in relation to the written word,
illustration can be ignored or omitted without loss’.3 But the ‘loss’ of the
illustration is never complete. The text may be privileged in a digital con-
text, but it is a text that is always already invaded by the visual. After all,
the very act of digitising the pages turns the words into images. Moreover,
36  J. THOMAS

the status of ‘illustration’ as ‘non-text’ reveals the multiple visual identity

of books: the ‘illustrations’ here include fingerprints, musical scores, signa-
tures, smudges, dead flies and thousands of British Museum stamps. Not
all ‘illustrations’ are illustrations, and not all ‘non-texts’ are the same.
This chapter looks at the shifting definitions of illustration generated
in the quest to make the content of these images searchable and identifies
the methods and tools facilitating these search mechanisms as direct inter-
ventions in the dynamic of word and image that characterises illustration.
This dynamic is engrained even in those computational tools that offer
the possibility of automated Content-Based Image Retrieval (CBIR).4
Such software, coming under the broad umbrella of ‘computer vision’,
emphasises the specificity of the visual, but the politics of word and image
are apparent in its very architecture. This is suggested in the so-called
‘Bag of Words’ approach, which was originally used in natural language
processing where a text (a sentence, for example) is broken down into a
multi-set of words, information that can then be used to classify types of
document according to the occurrence of these words. When applied to
images, the ‘Bag of Words’ model treats visual ‘features’ as words in order
to generate a vector of occurrence counts of these features. A ‘feature’
here is used in the sense of a piece of visual information (for example,
edges, points, curves, boundaries and blobs), which may be suggestive of
a particular pictorial style or the presence of a specific object.
The ‘Bag of Words’ approach implies that the seemingly neutral
domain of computational models is bound up in wider cultural issues.
Here, it is part of a critical discussion about how images can be conceived
of and described in words, and whether or not they should be ‘described’
in the first place. The ‘words’ are distinctly visual ones: shapes, patches,
colours, clusters and fragments, but the very term ‘Bag of Words ’ stands
as an example of what W. J. T. Mitchell calls ‘linguistic imperialism’,
the implicit assertion, which carries through to the realm of illustration,
that language is superior to and can stand in for the image.5 The ‘Bag of
Words’ model, or, more specifically, the ‘Bag of Visual Words’ model, is
always at odds with itself because, while it seems to indicate a blurring
of the boundaries between word and image (that images can be treated
as words, and that ‘visual’ and ‘words’ are interchangeable and synony-
mous terms), the very necessity for this software—to identify image con-
tent—suggests that images cannot be searched for computationally in
the same way as words. Images are difficult and different. And this dif-
ference is as much an issue for a digitised illustration as for an ‘original’
 3 SEARCHABILITY  37

wood-engraved one. As Matthew Kirschenbaum has argued, just as his-

toric images are distinct from words and reproduced by fundamentally
distinct means (engraving, etching, lithography, aquatint, etc.), so digital
images are radically different from electronic text.6
But it is the very difference of the digital image that has the potential
to make the content of illustrations searchable. Computer vision depends
on the fact that what is being ‘looked at’ is digital: it searches for and
makes connections between visual features at the level of the pixel.7
As Wolfgang Ernst comments, ‘This process of similarity-based image
retrieval is possible only due to the fact that the digitized image is an
image that is stored as a set of quantifiable elements’.8 An example of
how effective computer vision can be is the ImageMatch tool created for
the Broadside Ballads Online project (http://ballads.bodleian.ox.ac.uk).9
Using a search algorithm, ImageMatch retrieves other identically consti-
tuted visual features or ‘words’ from the images in the database, allowing
for the recognition of illustrations that derive from the same woodblock.
Other image similarity tools have had success in recent years, with com-
mercial applications like MatchEngine (https://services.tineye.com/
MatchEngine) used to look for similarities in art collections. John Resig
used this tool on the Frick Photoarchive, where it was able to detect cop-
ies of art works and duplicates, identify pictures before and after conser-
vation, and expose errors in cataloguing.10 Resig also employs an image
similarity engine in Ukiyo-e.org (http://ukiyo-e.org), a database contain-
ing over 213,000 Japanese woodblock prints from 24 institutions, in
which users are able to upload an image of a print and retrieve prints
with similar features from the database.
The most promising computer vision initiative in terms of nineteenth-
century illustrations are the tools designed by the Visual Geometry
Group at the University of Oxford’s Department of Engineering Science,
which is also responsible for ImageMatch. The group has worked on the
same dataset of illustrations from the British Library that is used in The
Illustration Archive to develop ‘Object Category Search’ and ‘Instance
Search’ tools.11 Object Category Search identifies features in the illus-
trations by downloading images containing those features from Google
Images. The Google images form a training set from which the classifier
is ‘learned’ and applied to the images in the British Library dataset. The
online demonstration searches for ‘arch’, ‘bird’, ‘church’, ‘flower’ and
‘train’, with impressive results, although the use of Google Images does
have its limitations. In the ‘bird’ example, the illustrations retrieved from
38  J. THOMAS

the dataset are all visually similar: the birds are positioned in side pro-
file, dominate the picture space, and are the central focus of the image.
They conform exactly to the Google ‘idea’ of a bird. While it is likely
that this tool would retrieve an illustration from an ornithological work,
it is unlikely that it would find a picture of a dead pheasant flung over
someone’s shoulder or doves flying into a dovecote, both of which fea-
ture in one of Millais’ illustrations to Trollope’s Framley Parsonage.12
Likewise, although the Instance Search tool is effective at searching for
more or less identical images, what is much more difficult computation-
ally is finding ‘sort of similar but not the same’ image content, a problem
that is exacerbated in the dataset of illustrations used in The Illustration
Archive because of the scale and variety of the images.13 Even recent
developments in computer vision such as ‘deep learning’, which attempts
to mimic the brain by using machine-learning algorithms to model high-
level abstractions in data, are heavily dependent on using similar types
and formats of images for analysis.14
A bird is not always a bird, at least to a computer. As we discovered
in our, admittedly more modest, experiments with image similarity algo-
rithms, the features of birds can resemble anything, from decorative
coats of arms to Victorian gentlemen’s facial hair. The Kestrel and the
Beard Problem, as I have come to think of it, suggests that the possibili-
ties of computer vision are tantalisingly close, but, at the moment, tech-
niques for making image content searchable need to be supplemented by
other manual processes in order to be effective across a range of images.
Mario Klingemann, an artist and practitioner of ‘computational craft’,
has adopted just such a mixed approach in his clustering of thousands
of illustrations from the British Library’s dataset. Klingemann’s work is
on display in The Illustration Archive, where typing almost any search
term is likely to return one of his collections, which include illustrations
of skulls, prams, shoes, maps (almost 40,000 of them), women with
books, artists, painters, nudity and pictorial letters (arranged by letter).
The brilliance of Klingemann’s identification of the content of the illus-
trations lies not in a reliance on a specific method, but on a combination
of approaches. He employs multiple data clustering algorithms alongside
a manual clustering of image training sets organised using human vision.
The most prevalent method for making image content searchable,
though, does not so much involve ‘vision’ as text. It is the spectral
presence of words that haunt these various techniques, from the ‘Bag
of Words’ model, and the textual metadata that underpins the pictures
 3 SEARCHABILITY  39

retrieved from Google Images, to the crowdsourced tags in ImageNet

that have been used in recent deep-learning experiments.15 This depend-
ence on words is a key component of large-scale databases contain-
ing illustrations that do not rely on computer vision, but instead use
the surrounding text to ‘explain’ and return relevant images, whether
this is the OCRed text (sometimes including the image caption), bib-
liographic and other metadata, or an amalgamation of different textual
information. This approach is typical of the majority of historic newspa-
per and periodical databases that include illustrations: Welsh Newspapers
Online (http://newspapers.library.wales), Gale Cengage’s The Times
Digital Archive, 19th Century British Newspapers, 19thCentury UK
Periodicals, The Illustrated London News Historical Archive, and Punch
Historical Archive, ProQuest’s British Periodicals and JISC’s Historical
Texts (which includes Early English Books Online, Eighteenth Century
Collections Online and the collection of British Library Books used in
The Illustration Archive). In the case of Historical Texts, the texts that are
used to render the illustrations searchable are the OCRed pages, infor-
mation about the book in the MARC record, and the occasional basic
description of an image (books containing the word ‘maps’ in the OCR
are indexed using this terms as an image descriptor).
The content of the illustrations in these resources is not searchable
in and of itself, but depends on textual information for its retrieval. The
words mediate and speak for the pictures. But is the voice of the text a
reliable one? To what extent can the text surrounding the image be used
to describe it? Illustrations are stubborn and words frequently fail them,
as their digital translation all too often exposes. The editors of the Marie
Duval Archive (http://www.marieduval.org), which brings together
images designed by this extraordinary late nineteenth-century cartoonist,
make a point of not utilising ‘thematic’ metadata in the form of image
tags on the grounds that ‘searches based on verbal tagging of visual
materials always generate exclusions and phantom sets of results, unless
the tags are verbally incontestable’.16 While this assessment is no doubt
a fair one (as I will go on to suggest), the vagaries of the ‘verbal’ are pre-
sent not just in tags but also in the bibliographic metadata, which is often
regarded as the most reliable—and in some cases, the only—information
that can make illustrations searchable (indeed, the bibliographic metadata
is used as the principle search mechanism in the Marie Duval Archive).
The reliance on bibliographic information instigates a certain type
of searchability that can influence and determine ideas of what an
40  J. THOMAS

illustration is ‘about’ or at what level it should be analysed. Book titles,

for instance, rarely describe the iconographic details of individual illus-
trations, drawing attention instead to the general content. A catalogued
title such as The Heart of Africa. Three years’ travels and adventures in
the unexplored regions of Central Africa from 1868 to 1871 … Translated
by E. E. Frewer. With an introduction by W. Reade … With maps and …
illustrations would return relevant illustrations for users searching for
the terms ‘Africa’ or ‘maps’ (although it is also likely, of course, that a
significant number of the images returned would not show ‘Africa’ and
would not be ‘maps’). However, the ellipses in this title indicate the fact
that the original title (as it appears on the titlepage of the published ver-
sion of the book) is frequently abridged in catalogue entries, making the
inherited bibliographic information less complete than it could be, and
less useful when it comes to searching for illustrations.17 And there are
also countless examples of titles, abridged or not, that reveal very little
about the content of the illustrations, like Mrs Alexander’s A Life Interest
(1888).
Although book titles may or may not overlap with the content of the
illustrations, they are notoriously inconsistent in the way that they refer
to and describe illustrated material. Indeed, this is the case more generally
across the bibliographic metadata of illustrated books. Historically, there
has been no standard convention for describing the presence of illustra-
tions in catalogue records, a factor that indicates and replicates the mar-
ginalisation of the genre. The presence of illustrations might be mentioned
arbitrarily (for example, if the illustrator is regarded as worthy of note), in
sparse details (‘3 plates’, ‘a frontispiece’), and sometimes not at all. A recent
study has concluded that the place of illustration in early modern books is
so complex that a bibliographic model needs to be developed that accounts
for the texts and the illustrations as two distinct, but intertwined, objects.18
The lack of a fixed bibliographic convention for describing illustration
presents logistical problems when it comes to developing computational
tools to retrieve information about the illustrations from the metadata.
These issues became apparent during the development of The Illustration
Archive when we attempted to devise an algorithm that would identify
and isolate illustrators’ and engravers’ names from the titles attached to
the books. A very rough version of the algorithm went like this:

• Select some word parts, which describe illustrative activities, e.g.

‘illust’, ‘etch’ and ‘draw’
 3 SEARCHABILITY  41

• Take the British Library’s designated title for a book: The Sports of
the Genii. [Etchings from drawings by Miss Susan Macdonald. With
verses by A. Home]
• Split this title into words, separated by spaces
• Look at each word in turn, checking for the occurrence of the
word-part: ‘draw’
• Count how many words into the title this word is
• Join the words before and after this word into two strings: The
Sports of the Genii. [Etchings from … by Miss Susan Macdonald. With
verses by A. Home]
• If the second string, which comes after the word-part ‘draw’, con-
tains the word ‘by’, continue
• Find the ‘by’ word in the second string (here it is the first word, but
this is not always the case)
• If the second string contains the word ‘author’, make a leap of faith
that it is the author who also designed the drawings and identify the
author’s name from the book’s metadata
• ‘Clean’ the data, removing any parentheses: Miss Susan Macdonald.
With verses by A. Home
• Take the start of the string up until the first word that is longer than
four characters (and so likely to be a name rather than a title) and
that ends in a full stop: ‘Miss Susan Macdonald’
• Remove any punctuation that may have been dragged along
• And this is the Eureka moment: the illustrator is ‘Miss Susan
Macdonald’

To say there is scope for error here is something of an understatement.19

There are countless examples where this technique would fail miserably,
such as the relatively common inclusion of information after the illustra-
tor’s name: Swiss Pictures drawn with Pen and Pencil. New edition, revised
and partly re-written. With several additional illustrations by Edward
Whymper and a map. In this case, the ‘map’ is identified as the illustra-
tor along with Edward Whymper. The same problem arises with a title
like The Months illustrated by pen and pencil where the illustrations seem
to have been created by a person with the name ‘pen and pencil’. Then,
there are those wonderfully long bibliographic records that make an algo-
rithm useless. A search for ‘embell’ (‘embellished’ is a typical indicator of
the presence of illustrations) would return this: The Beauties of England
and Wales; or, Delineations, topographical, historical, and descriptive, of
42  J. THOMAS

each country. Embellished with engravings. (vol. 1–6 by E. W. Brayley and

J. Britton; vol. 7 by E. W. Brayley; vol. 8 by E. W. Brayley; vol. 9 by J. Britton;
vol. 10, pt. 1, 2, by E. W. Brayley; vol. 10, pt. 3 by the Rev. Joseph Nightingale;
vol. 10, pt. 4 by J. Norris Brewer; vol. 11 by the Rev. J. Evans and J. Britton;
vol. 12, pt. 1 by the Rev. J. Hodgson and Mr. F. C. Laird; vol. 12, pt. 2 by
J. N. Brewer; vol. 13 by the Rev. J. Nightingale; vol. 14 by Frederic Shoberl;
vol. 15 by J. Britton, J. Norris Brewer, J. Hodgson, F. C. Laird; vol. 16 by
John Bigland; vol. 17 by the Rev. J. Evans; vol. 18 by Thomas Rees).
One of the main reasons for the ineffectiveness of this model, however,
is not so much the inconsistency of the book titles, but the slipperiness of
the language used to describe illustration, which indicates the extent to
which it is both a culturally dominant form and one that has been curi-
ously neglected. Although a search for ‘illustration’ across large reposi-
tories like the Hathi Trust Digital Library (https://www.hathitrust.org)
and the Internet Archive (https://archive.org) mines the bibliographic
metadata for the term ‘illustration’, there are, in fact, multiple terms for
describing the presence of illustrations: ‘embellished by’, ‘decorated by’,
‘illuminated by’, ‘designed by’, ‘adorned with’, ‘with etchings by’, ‘drawn
by’. This problem is intensified in The Illustration Archive because,
although the majority of books in the British Library dataset are written
in English, not all are. The multiplicity of terms used to describe illustra-
tion is a multi-lingual problem. Before one could even begin to extract
this information, therefore, one would need a dictionary of terms relat-
ing to illustration across different languages (needless to say, such a dic-
tionary does not exist). Indeed, this would provide an interesting starting
point for an analysis of the place of illustration in different geographic and
cultural contexts. The nineteenth century was a period when illustration
and illustrative techniques crossed borders and continents. Bibliographic
metadata may have something important to reveal about practices in dif-
ferent countries and the status given to certain methods of reproduction.
These variations bring their own meanings and values, some of which play
on notions of national identity and difference: the French ‘gravure’ is dis-
tinct from English ‘engraving’, with ‘wood engraving’ regarded as a pecu-
liarly English form well into the middle of the nineteenth century; what
is called a ‘mezzotint’ in English (a technique that is itself of German ori-
gin) is known in France as ‘la manière anglaise’; and William Vaughan has
noted that, in German, the word ‘Holzschnitt’ is used for both ‘wood-
cut’ and ‘wood-engraving’, while the precise terms for wood engraving,
‘Holzstich’ or ‘Xylographie’, rarely appear.20
 3 SEARCHABILITY  43

In The Illustration Archive, we did not overcome the issue of identify-

ing and extracting the names of the illustrators and engravers. Instead,
a search in the Archive for a specific individual interrogates all the bib-
liographic information from the British Library, however variable that
information may be. However, we did attempt to counter the marginali-
sation of the figure of the illustrator in a small way: by placing the ‘illus-
trator’ field at the top of the bibliographic search options. The fact that it
is so difficult to isolate information about illustration suggests the unsta-
ble presence (or absence) of the genre in bibliographic metadata, and the
arbitrary, even random, ways in which illustration is defined in relation
to the book. This is an issue that, to my knowledge, has not been dis-
cussed in scholarly work, but it has to be confronted in the development
of digital illustration resources. It could be argued, in fact, that the drive
to make illustration searchable in the digital archive reveals the extent to
which illustration is unsearchable in the physical archive.
If the bibliographic metadata fails to account fully for illustration,
there is other textual information that promises to make the content
of these pictures searchable online: the words that the images illustrate
and that appear alongside them. The ‘closest’ relation between text and
image in illustration seems to lie in the caption, the physical proxim-
ity of word and picture here suggesting their equivalence, a manifesta-
tion of Saussure’s illustration of the tree with the caption beneath. But,
although the caption is commonly regarded as a simple textual descrip-
tion of the image, a digital environment exposes how problematic this
assumption is. Captions occupy a contested borderland between word
and image, belonging to both or neither. This is exemplified when illus-
trations are extracted from the digitised and OCRed pages. At times, the
captions stick with the illustrations in a way that designates them part
of the ‘non-textual’ space of the image, while at other times the cap-
tions are cut off from the image and remain within the space of the text
proper. Even the notion of what constitutes a ‘caption’ is problematic, to
the extent that initial feedback from the users of The Illustration Archive
indicated that it should not be taken for granted and that all users would
know what a ‘caption’ was or where to find it. For this reason, we
included a very basic definition of a ‘caption’ on the site: ‘a title below
the image’. In DMVI, we used the word ‘title’ rather than caption on the
basis that ‘title’ is a more historically accurate term than ‘caption’, which
was not used in the context of an illustration until the early twentieth
century; in its early usage, ‘caption’ described part of a text—the heading
44  J. THOMAS

of a chapter or newspaper article. When applied to illustration, this text-

derived notion of a caption as a ‘heading’ is literally turned upside down,
with captions commonly placed at the foot rather than the head of an
image, as our definition of a caption as ‘a title below the image’ implies.
The caption is turned upside down in another sense, too. Whereas
the objective of a caption in the sense of a textual heading is to explain
or encapsulate what the article is about before it is read, the caption
of an illustration is often read alongside or even after a viewing of the
image. Moreover, the caption of an illustration does not necessarily play
an explicatory role. The nature of the interaction between caption and
illustration is, in some ways, determined by the genre of text. Captions
in geographical or ‘scientific’ books (in the widest sense) tend to have
descriptive functions, like ‘Fig. 128. The Boston Block, Minneapolis’
(from William H. Wiley and Sara King Wiley’s The Yosemite, Alaska,
and the Yellowstone, London, 1893); or ‘Cretaceous Bird: Hesperornis
Regalis of Marsh’ (from James Dwight Dana’s Manual of Geology, New
York, 1880). In illustrations accompanying literature, however, this rela-
tionship is often compromised. In DMVI, where all of the illustrations
are from literary texts, the title/caption does not necessarily describe in
words what the picture shows; rather, it can signify in an enigmatic and
allusive relationship to the image. An example of this is ‘True or False?’,
an illustration from DMVI that shows a woman holding a baby behind a
chair in which a man sits. The man reaches awkwardly above him to cup
the baby’s hand in his (Fig. 3.2).
Far from explaining the content of the illustration, the title, ‘True or
False?’, points to what is absent from this play of picture and caption: the
text itself. The cryptic question incites the viewer’s desire for a textual
resolution to the interaction between the illustration and its title, calling
on the missing words to explain the ambiguous meanings of the caption,
the image and the relationship between them. The ‘completed’ meaning
of the illustration, then, is seen to lie outside the apparent union of the
image and its caption: in the surrounding text.
A caption like ‘True or False?’ does little to add to the searchability
of the content of the illustration. It gives nothing away about what the
picture actually shows. But what of the textual domain that the caption
urges us toward? Does the missing text fulfil its promise to explain what
the illustration depicts? In some ways, the accompanying text can increase
the searchability of illustrations. The Cervantes Project has developed a
taxonomy in an XML schema of episodes, characters and adventures in
 3 SEARCHABILITY  45

Fig. 3.2  Arthur Boyd Houghton, ‘True or False?’, illustration for Adelaide

Anne Proctor, ‘True or False?’, Good Words 3: 12 (December 1862) p. 721.
Engraved by the Dalziel Brothers. Available on the Database of Mid-Victorian
Illustration: www.dmvi.org.uk
46  J. THOMAS

Don Quixote that is used to link the illustrations to the text, enabling
a search for illustrations such as ‘DQ knighted at inn by innkeeper’.
The use of this taxonomy to retrieve illustrations gives a fascinating insight
into the history of these images, indicating what scenes are most often
illustrated and what scenes are not (there are 250 of this scene alone
whereas there are no illustrations of ‘Night encounter with Maritornes’.
The taxonomy can also indicate at what historical point the illustration of
a particular scene becomes popular, how the images evolve, change and
differ from each other, and instances of their reuse and modification in
various editions (the presence of numerous illustrations in this archive that
are mirror images of each other suggests that these illustrations were cop-
ied from earlier woodblocks and reprinted). This search mechanism is an
effective way of connecting the illustrations to the text of Don Quixote,
but such an extraordinarily detailed taxonomy of episodes can only be
created for a single or small number of texts; it could not be scaled up
to dozens or hundreds of texts. Another limitation lies in the fact that,
because the taxonomy is based on episodes, adventures and scenes
from Don Quixote, it does not specify image content other than what is
included in the description of the episode. We can infer that there is prob-
ably an inn and an innkeeper in the illustration of ‘DQ knighted at inn
by innkeeper’, but the presence of other people, objects and animals is
detailed in a supplementary keyword list.
Although the taxonomy used in the Cervantes Project allows the
text and the illustrations to be brought together, it stands indepen-
dently from the text, as another layer of metadata. But there is always,
of course, the text ‘proper’. If the words of the text coincide with the
content of the illustrations, then the issue of making the images search-
able could be resolved. In the case of The Illustration Archive, all 68,000
volumes could potentially be mined to search for relevant images. This
idea that the text ‘maps’ onto the content of illustrations was fundamen-
tal to a recent project by Kalev Leetaru that used the accompanying text
to make searchable millions of illustrations from books in the Internet
Archive. The software that Leetaru developed extracted the caption from
each image and the paragraphs before and after the illustration in the
book and made this information searchable along with the bibliographic
metadata. When users type in search terms, the system interrogates these
texts in order to find relevant images.21
The searchability of the illustrations in Leetaru’s project, however,
relies on assumptions about how illustrations signify that are highly
 3 SEARCHABILITY  47

problematic. In particular, there is the idea that the proximity of words

to an illustration somehow guarantees the likelihood that these words
‘explain’ the image. But, just as a caption does not always explain what
is in an illustration, neither does the nearest text. In the adjoining ver-
bal context to an illustration, there are probably far more ‘redundant’
or even misleading words than there are ‘hits’ and this will result in false
negatives (missed images) and false positives (unrelated illustrations that
have to be manually ‘eliminated’). Even if some aspects of the text do
map onto the illustration, they are not necessarily to be found in the
words that wrap around the image, the paragraph before and the para-
graph after. An illustration can refer to words that cover several pages,
or it can be separated from the relevant text by many pages. A search
on the Flickr photostream where Leetaru uploaded the images demon-
strates that although his system does generate positive hits, there are also
a significant number of misses. In one of my searches for the term ‘boy’,
ten illustrations of boys were retrieved in the first 20 results (although
some of these were ambiguous). Some searches are inevitably more suc-
cessful than others and this seems to be dependent on the genre of the
illustrated book. A search for ‘bridge’, for example, retrieves illustrations
from the types of books (travel, engineering and geographical) that are
likely to have descriptive captions and relevant keywords in the surround-
ing texts, while a seemingly simple word like ‘boy’ is problematic pre-
cisely because the term can appear in a variety of textual contexts that do
not ‘match’ the illustration. Other potentially ambiguous search terms
like ‘laughing’ return a confusing array of results including a picture of a
butterfly from a Victorian novel, The Butterfly Hunters (1868), which is
retrieved because the text immediately prior to the image states, ‘All the
boys laughed’. More disturbing in the context of the search for ‘laugh-
ing’ is a photograph of two rows of solemn-faced Chinese children, some
with their heads bowed. This troubling image is taken from a social sur-
vey of Peking and the hit ‘laughing’ comes from the paragraph prior to
the illustration, which is a quotation from an advertisement for a young
female prostitute: ‘She is beautiful even when she does not laugh’. This
is just one example of the problem of assuming that the content of an
illustration re-produces in pictorial form the paragraphs that sandwich it.
There are no simple solutions to these issues. Until a Content-Based
Image Retrieval system is developed that is sophisticated enough to
cope with tens of thousands of diverse illustrations, the results of search-
ing big datasets of images will always be serendipitous. Indeed, as Viktor
48  J. THOMAS

Mayer-Schönberger and Kenneth Cukier have argued, the very condition

of big data is one in which we have to ‘loosen up our desire for exacti-
tude’.22 However, Leetaru’s answer to the searchability of illustrations is
interesting not so much because it prompts us to learn to live with error
rather than ‘exactitude’, but because of what it reveals about the relation
between word and image in illustration: that it does not function in the
transparent way suggested in the architecture of this project. Illustrations
do not simply replicate what is in the surrounding text, and this is even
assuming that a textual description of an image can be reduced to a list
of extractable keywords isolated from larger semantic units.
What also becomes apparent from this exercise is that different illus-
trations generate different interactions between word and image. Not all
illustrations function in the same illustrative way. As with captions, the
gap between the illustration and the surrounding text can become a gulf
in the case of literary works. This was a topic that we set out to ana-
lyse in more detail on an explorative project in which the illustrations
in DMVI were examined in terms of their correlation with the text.23
Focusing on illustrations in the periodicals London Society, Good Words
and Once a Week, the anthologies English Sacred Poetry and Favourite
English Poems, and the novels Orley Farm and Romola (as serialised in
the Cornhill Magazine), we devised a number of questions about the
relation between word and image, the answers to which were inputted
into three fields in an Access database. The first field contained details
about the physical relation of the image to the text (where and how the
illustration is positioned on the printed page); the second field included
information about the interaction between text and image (the relation
between the illustration and the caption, and whether or not the illustra-
tion refers to a specific textual moment or passage); the third field was
a text box for transcribing the textual episode identified in the second
field.
Results indicated that roughly 25% of the illustrations in our study
did not relate in any recognisable way to a specific part of the text they
accompanied. Perhaps more revealing is that a number of illustrations
failed even to conform to the rigid questions that we were asking, leav-
ing our researchers frantically filling in the ‘additional notes’ box. Some
illustrations, for example, have a tangential rather than a direct relation-
ship to the text and seem to depict the mood or tone of the text without
actually referring to particular words. Then there are those illustrations
that work in reverse, showing in concrete terms what are abstract ideas
 3 SEARCHABILITY  49

or concepts in the text. Such an ambiguous relation between word and

image is a typical feature of pictorial capital letters like those designed
by Frederic Leighton for Romola, which often set the scene rather than
show an identifiable textual episode. This is equally true of those other
pictorial capital letters that, aesthetically, stand almost in opposition to
Leighton’s classical ideal: the illustrated letters that appear in Thackeray’s
novels, which often engage with the text metaphorically rather than
depicting the words.24
We also found instances where there was a conflict between the infor-
mation given in the words and the information in the picture. In an illus-
tration for Robert Bloomfield’s ‘The Farmer’s Boy’ in Favourite English
Poems, the verse has the protagonist digging up ‘turnips’ to give to the
cattle (‘Beneath dread Winter’s level sheets of snow/The sweet nutri-
tious Turnip deigns to grow’25), but the figure in the illustration is carry-
ing sheaves of hay (Fig. 3.3).
This might seem like a trivial discrepancy, but it could be significant
for an understanding of the material and economic circumstances of
illustration because the generic ‘look’ of the picture, designed by Myles

Fig. 3.3  Detail from Myles Birket Foster, illustration for Robert Bloomfield,
‘The Farmer’s Boy’, Favourite English Poems (London: Sampson Low, 1862)
p. 174. Engraved by Edmund Evans. Available on the Database of Mid-Victorian
Illustration: www.dmvi.org.uk
50  J. THOMAS

Birket Foster, along with this incongruous detail, suggests that this could
be an illustration that has previously been used to illustrate a different
text. Certainly, Birket Foster was well known for mass-producing hun-
dreds of stock ‘rural’ illustrations that could potentially illustrate any
number of texts that happened to be set in the countryside. The dis-
crepancy between turnip and hay would have repercussions in a digi-
tal resource where the text is being used to search for images: if a user
searches for pictures containing ‘turnips’, the image returned via the text
would be wrong; and if the user searches for illustrations showing ‘hay’,
the picture would not be retrieved because the word ‘hay’ does not
appear in the text. In a similar, though more ethereal, example from ‘The
Rime of the Ancient Mariner’ in Favourite English Poems, a supernatural
figure appears in the illustration when the mariner shoots the albatross.
This ghost-like being would remain forever intangible if the user is reli-
ant on the words of the text to retrieve the images.26
These illustrations, which are by no means extraordinary, undermine
the notion of equivalence between word and image, an equivalence that
is necessary if the surrounding words are used to search for and retrieve
the content of pictures. Even in those cases where the text and image
do coincide in some way, the relevant text is not necessarily located in
proximity to the illustration. Although the actions and events in the text
might be physically and thematically ‘close’ to the actions and events
depicted in an illustration, the textual description of the setting for those
actions might be given pages earlier. ‘Peregrine’s Eloquence’, an illus-
tration from Trollope’s Orley Farm, appears facing page 289 in the first
book edition of the novel while the textual passage it illustrates is a short
section at the end of page 295 that spills over onto page 296 (this is the
only textual moment where the protagonist stands with his back against
the fire, the position in which he is pictured in the illustration). However,
the caption of the illustration, ‘Peregrine’s Eloquence’ relates to an ear-
lier moment on page 295, and the description of the room in which
the scene takes place is described a couple of pages before that, on page
293.27 In this example, the caption would be of little help in retrieving
the content of the illustration, unless the user was searching for pictures
of ‘Peregrine’ (in which case, numerous illustrations of falcons would be
returned, too!). ‘Peregrine’s Eloquence’ demonstrates that, in the case of
literary illustrations, the subject matter of an image can be taken from
multiple parts of the text, and these parts do not always surround the pic-
ture. In periodical publications and monthly instalments (Orley Farm was
 3 SEARCHABILITY  51

first published in monthly shilling parts), it was common for the illustra-
tions to appear at the front of the instalment, and in one such example
from London Society, a frontispiece illustration is separated from the poem
it illustrates by 65 pages.28 In cases like this, the caption of the illustra-
tion often points the reader in the direction of the text (here it states
‘see p. 162’), a device that presents its own problems if the caption is
employed as a simple descriptor of the image that can ‘find’ iconographic
content.
What all these examples suggest is that the relation between word
and image in illustration is a highly complex one. In terms of making
the content of illustrations searchable online, the textual information
surrounding the picture—whether this is the bibliographic metadata,
the caption or the words—is unreliable. The best option, perhaps, is to
let the illustrations speak for themselves. This, in theory at least, is the
approach adopted in what has become the most prevalent method for
making the content of images searchable online: keywording or tagging,
a process in which the images are looked at and described in words, and
users are able to find relevant pictures from this textual markup. Tagging
has been employed in a variety of digital contexts, from individual files
such as images and video to websites and collections of connected
resources. Search engines like Google rely heavily on tags, which are
often automatically retrieved from formalised data structures. As Mayer-
Schönberger and Cukier put it, ‘Tagging has emerged as the de facto
standard for content classification on the Internet’.29
In the case of illustration resources, tagging also has implications for
the relationship between word and image that defines illustration. As a
mechanism for describing the visual content of a picture in words, tag-
ging introduces another layer of text that comes to characterise illustra-
tion. In this sense, the marking up of image content could itself be called
an ‘illustrative’ act. Indeed, the activity of tagging illustrations, of add
ing words to pictures, is reminiscent of the ways in which nineteenth-
century texts were often ‘written to’ images.30 Traces of this prac-
tice, which was relatively commonplace in the case of albums and gift
books, can be found in the titles of books: Birket Foster’s Pictures of
English Landscape … with pictures in words by Tom Taylor (1862); Death’s
Doings; consisting of numerous original compositions, in prose and verse,
the … contributions of various writers; principally intended as illustra-
tions of 24 plates designed and etched by R. Dagley (1826). These titles
position the prose and verse compositions as ‘pictures’ or ‘illustrations’,
52  J. THOMAS

subverting what is commonly regarded as the conventional illustrative

relationship where the text precedes the illustration.
Like these ‘illustrated’ books, tagging or keywording also has the
effect of ‘writing’ the image, but the apparent textual excess of this activ-
ity is predicated on a lack or deficiency. While schema and guidelines like
the Text Encoding Initiative (TEI) exist for the marking up of texts that
allow them to be searched when those texts are transformed into human-
readable formats, there is no equivalent schema for illustration (or even
for images more generally31) that provides markup standards in terms
of what the illustrations show and how they show it, their attributes,
and their positioning on the page. As a result, illustrations have either
gone ‘un-marked up’ and, therefore, un-findable, or digital illustra-
tion resources have reinvented and refined existing classification systems
or developed their own bespoke keywording apparatus to describe the
content of the images. The Cervantes Project optimises the searchability
of its images by supplementing its taxonomy of chapters, episodes and
adventures with metadata that includes authors, artists, techniques and
types of illustration, free text information entered by the editors in the
Commentary and Notes fields, and a keywording structure that lists con-
tent tags for people (characters), places (geographical, physical), objects
and animals.
As the example of the Cervantes Project suggests, in order to ensure
the relevance of the keywords to the specific texts and images in the
digital archive, the keywording system often needs to be tailor-made;
the characters and locations in illustrations of Cervantes’ novels are not
the same as the characters and locations in illustrations of Shakespeare’s
plays. Such bespoke keywording systems are most effective in ‘closed’
datasets with a single text like Don Quixote or an individual author like
Blake. DMVI includes illustrations from multiple works and authors, but
the relatively small number of images (868) meant that we were able to
keyword the images using a structure that was developed in-house.32
Even with our comparatively small number of images, this was a labour
intensive process that involved examining each image in order to estab-
lish a vocabulary that could be applied across the corpus of images.33 We
were inspired in this exercise by other digital illustration resources like
the William Blake Archive, which employs an exhaustive list of keywords
in order to make the content of Blake’s illustrations searchable.34 One of
the closest neighbours to DMVI in terms of keywording infrastructure
is British Printed Images to 1700 (http://www.bpi1700.org.uk), which
 3 SEARCHABILITY  53

includes thousands of prints and book illustrations from the early mod-
ern period that have been marked up by hand and arranged in a hierar-
chical structure by topic. Although the terms of British Printed Images
are different to DMVI’s, reflecting the historical differences between the
two datasets, DMVI uses a similar tree-like structure that allows users
to drill down and across categories. The keywording system developed
on DMVI has since been adapted for use in The Yellow Nineties Online,
Illustrating Scott and Nineteenth-Century Serials Edition (NCSE)
(http://www.ncse.ac.uk).35
Keywording an illustration is not as straightforward or transparent as
it might seem, however. One problem that emerged during the develop-
ment of the keywording system for DMVI was the extent to which we
could, or should, look into the text to mark up the content of the illus-
tration. In some cases, this recourse to the text was difficult for practical
reasons (the illustrations from Aberystwyth and the Ashmolean museum
had already been extracted from the periodicals), but there was also a
critical issue at stake in using the text to describe the image. The activ-
ity of keywording those illustrations that had been divorced from the
text and then going back to the texts in order to shape and modify the
tags emphasised the complexities of the interaction between word and
image. To use Roland Barthes’ term, the words can ‘anchor’ the illus-
tration, steering an interpretation of the image or suggesting new and
different meanings that the image does not have independently.36 For
example, the text might specify a familial relationship between figures
that is not obvious from the picture alone: figures marked up as ‘boy’
and ‘girl’ become ‘brother’ and ‘sister’. Equally, the text can clarify what
is happening in an illustration or can fill in the details rendered by a few
scratched lines. When the text is read alongside the picture, an indistinct
snowy landscape turns into ‘Canada’. This textual ‘anchoring’ also means
that illustrations often marginalise or exclude details that are promi-
nent in the words (an illustration does not need to visually demarcate or
define a feature that is a textual given). In this sense, illustrations interact
with texts as much in what they do not show as what they do.
This is not a one-way ‘anchorage’, though: the illustration also
‘anchors’ the meanings of the words and determines how they are read.
A simple ‘drawing room’ in the text becomes furnished in the illustra-
tion, and these pictorial details generate their own meanings (about class,
gender and domesticity). An illustration can define what someone looks
like: the iconic figure of Sherlock Holmes, deerstalker cap and all, was
54  J. THOMAS

largely created by the illustrator, Sidney Paget. Peter Mendelsund has

remarked on how little information readers are given in texts about what
fictional characters look like. ‘It is precisely what the text does not eluci-
date’, he comments, ‘that becomes an invitation to our imaginations.’37
To emphasise this point, Mendelsund’s article in The Paris Review
includes numerous illustrations of figures with their faces and heads
erased or chopped off. The problem with illustrations, as Mendelsund’s
pictorial beheadings suggest, is that, by showing details that the texts
do not, they negate this imaginative interaction and instead direct and
manipulate our interpretations, ‘closing off’ the plurality of the text. This
is precisely the reason why illustration was so frequently criticised in the
nineteenth century. Gustave Flaubert admonishes illustrators in these
terms:

Du moment qu’un type est fixé par le crayon, il perd ce caractère de géné-
ralité, cette concordance avec mille objets connus qui font dire au lect-
eur: “J’ai vu cela” ou “Cela doit être”. Une femme dessinée ressemble à
une femme, voilà tout. L’idée est dès lors fermée, complète, et toutes les
phrases sont inutiles, tandis qu’une femme écrite fait rêver à mille femmes.
Donc, ceci étant une question d’esthétique, je refuse formellement toute
espèce d’illustration.38

The art critic, John Ruskin, reaches a similar conclusion, commenting

that ‘it is not well to make the imagination indolent, or take its work out
of its hands by supplying continual pictures of what might be sufficiently
conceived without pictures’.39 For Ruskin, the danger of illustration lies
in its impact on the reader, while for Flaubert it is the effect of illustra-
tion on the writer’s words that is of utmost concern. What both criti-
cisms imply, however, is that the threat of illustration exists not so much
in its suppression of the imagination as in its suppression of the multi-
plicity of the text.
In a digital environment, the multiplicity of language is exposed at the
very moment that it seems to be held in check. Keywording or tagging
an illustration depends on the fixity and stability of the relation between
word and image, the notion that a word can stand in for a feature of an
image. On the homepage of Your Paintings (now rebranded as Art UK:
http://artuk.org), which invited its visitors to tag its collection of over
200,000 paintings, luggage ‘tags’ attached themselves to the appropriate
part of the picture as a way of demonstrating the tagging process: a ‘hat’
 3 SEARCHABILITY  55

tag perched on a man’s bowler hat, a ‘duck’ tag impeded the progress of
a duck as it glided across a lake, and a young child could be forgiven for
looking rather red in the face as a ‘People>>Children’ tag attached itself
to the side of her head. The relationship between text and image here
was presented as an obvious and straightforward one. ‘All we ask’, Your
Paintings encouraged its potential taggers, ‘is that you look carefully at
the art and tell us what you see’.40
But ‘telling what you see’ is not a simple task, and this was exposed
in the tagging demonstration pages of the site that included sugges-
tions on how the images had been tagged ‘by the system’. ‘There is no
100% right answer to this’, the user was informed. ‘However, in this case
we would have chosen the following tags …’. Ironically, it is the very
requirement for this statement and the inclusion of the ‘recommended’
tags that reveals the fissure between words and images, the fact that they
are not the same, even though the act of tagging depends on the illusion
of their equivalence. As the ‘system’ and user-generated keywords appear
on the screen beside the image, this fissure becomes even more apparent.
Tags never fully succeed in labelling and identifying aspects of the pic-
ture. The luggage is always already lost.
The gap between word and image is as much the result of the visual
specificity of the image as of the instability of words. Gotthold Lessing
famously argued that images are spatial rather than temporal: they
exist as stationary objects in space rather than objects that progress in
time.41 A text has a discreet beginning and end according to the logic
of the sentence; images, on the other hand, have no such logic and this
becomes only too apparent in the ‘blind’ keywording with no prompts
from the system that is encouraged by Flickr and on other social media
sites. When it comes to keywording a picture, where does the tagger
even begin? The tagging structures on sites like Your Paintings impose
a temporality, a textual logic, on the image in the form of questions
about the image that the tagger is asked to answer. But even within the
constraints of these questions, there is no natural or obvious order in
which the features of an image should be tagged. Nor is there a predict-
able order in which these features are actually seen. This unpredictability
was the focus of First Impressions, a crowdsourced tagging project run
by Indianapolis Museum of Art, where users were asked to click the first
thing that caught their eye in a series of artworks. The results showed
some variation in what the visitors saw ‘first’, especially in those composi-
tions where single elements were not placed in the middle foreground.
56  J. THOMAS

Commenting on the results, the museum blog states that, ‘Through this
experiment it seems that individuals are viewing art however they want
to view art, which, is exactly correct’.42 Or not correct. What this experi-
ment indicates is that images are not viewed in the same way by all spec-
tators and the order in which things are seen is not self-evident, and this
is assuming that the activities of seeing and clicking/tagging are as simul-
taneous and as instantaneous as is implied here.
Some aspects of an image might not be seen at all. Keywording raises
the question of what is visible and invisible in an image. Paradoxically,
while digital image processing privileges the visuality of illustration (as
argued in Chap. 2), this is precisely the aspect of the image that tends to
be marginalised in keywording systems. The colour illustrations in The
Illustration Archive, for example, are rarely tagged as ‘colour’. Then,
there is the associated issue of what should or should not be marked up
in a picture. Are the keywords ‘man’ or ‘woman’ sufficient, or should
features and clothing also be described (a test tagger for The Illustration
Archive tagged ‘nose’, ‘mouth’ and ‘eyes’ for every portrait, a level of
detail that may seem excessive, but is by no means incorrect). And what
of the relationship between objects? This relationship cannot be articu-
lated in single keywords: the tags ‘woman’, ‘letter’, ‘cat’ and ‘sleep-
ing’ do not express how these features interact in an image. The Yellow
Nineties Online attempts to counteract this limitation by including a
prose description of each image alongside keywords. This combination of
approaches means that the illustrations are searchable using both sets of
markup (the keywords and the descriptions), with the prose markup hav-
ing the advantage, as The Yellow Nineties Online asserts, of describing the
image for the visually impaired.43
When employed as the sole mechanism for describing and searching
the content of images, however, prose descriptions can be as problematic
as keywords. The more ‘spontaneous’ nature of representing an image
in sentences rather than adhering to the rigid categories of a keywording
schema means that aspects of an image might be left out and the image
would not be retrieved if the end user were searching for the excluded
term.44 Prose descriptions, like keywords, give only a partial and frag-
mentary account of the visual impact of the picture. The illusion of
the equivalence of word and image on which both keywords and prose
descriptions depend is only ever an illusion: words do not seamlessly map
onto pictures. Indeed, in many ways, the activity of tagging is not actu-
ally one of ‘mapping’ at all, but of interpretation.
 3 SEARCHABILITY  57

This is suggested even in the scientific rigour of a tool like Iconclass

(http://www.iconclass.nl), an alpha-numeric system of classifying images
that is widely used in museums and galleries. Iconclass was set up in the
1950s by the art historian, Henri van de Waal, and has frequently been
updated to accommodate modern, abstract images and different artistic
genres like photography. It contains 28,000 hierarchically structured def-
initions within 10 main divisions that have an alpha-numeric code (nota-
tion) that equates to a description of the iconographic subject (textual
correlate). Despite its apparent precision, however, Iconclass is enmeshed
in ideas about what ‘art’ is and what it contains. It is heavily geared, for
instance, towards fine art as opposed to popular art forms, and to medi-
eval and renaissance religious art, in particular. Such images are highly
representative of the sort of gallery-based print and painting collections
that de Waal had in mind when he developed it. A browse through the
classification system begins with the category ‘Religion and Magic’, while
‘The Bible’ and ‘Classical Mythology and Ancient History’ feature as
part of the ten main iconographic categories.
We considered using Iconclass in the original incarnation of
DMVI, but decided against it on the basis that it did not fit our col-
lection of popular, mass-produced mid-Victorian illustrations as well
as a bespoke system would. We revisited Iconclass on a later project to
enhance the features of DMVI. In collaboration with the Arkyves pro-
ject (http://arkyves.org), the iconographic tags in DMVI were mapped
onto Iconclass codes. This enabled multi-lingual image retrieval, mak-
ing DMVI searchable in German, French and Italian, and also allowed
DMVI to be searched alongside other collections in the Arkyves data-
base.45 The act of ‘translation’ occurred not only at the level of the
multi-lingual searchability, however; it also characterised the mapping of
the DMVI iconography onto the Iconclass codes and the manual adjust-
ment to ‘tidy up’ and correct the automatic mapping. This mapping gen-
erally worked far better than we had anticipated, but occasionally there
were glitches in the process suggestive of the differences between the
Catholic fine art privileged by Iconclass and the largely Protestant illus-
trations in DMVI. An illustration showing the Scottish Presbyterian mar-
tyr, Margaret Wilson, as she stands tied to a palisade in the sea, caused
problems for Iconclass because the usual classification of such iconog-
raphy, ‘martyrdom of a saint’, was not appropriate in this context.46
With other illustrations, the Iconclass mapping alerted us to the preva-
lence of Christian themes and iconography that we had hardly noticed
58  J. THOMAS

Fig. 3.4  Arthur Boyd Houghton, illustration for J. Hamilton Fyfe, ‘About

Toys’, Good Words 3: 12 (December 1862) p. 753. Engraved by the Dalziel
Brothers. Available on the Database of Mid-Victorian Illustration: www.dmvi.
org.uk
 3 SEARCHABILITY  59

were there: an allusion from Exodus to the pillar of light that helps the
Israelites, which is depicted in an illustration by Myles Birket Foster47; a
toy ark in a picture by Arthur Boyd Houghton for Good Words (Fig. 3.4).
The keywords here are not merely adjuncts to the image, but consti-
tute its meanings, influencing how the illustration is viewed. The map-
ping of the Iconclass codes onto DMVI’s keywording system meant that
‘interpretations’ of the same image could be seen side by side, and, by
analysing these interpretations together, their cultural meanings became
apparent. The marking up of an image is always a culturally inflected
activity, and this takes on some political significance when decisions
have to be made about how illustrations should be keyworded. How,
for example, should a tagger describe an illustration of a fictional fig-
ure who is frequently described in the text in terms that would now be
regarded as racist? Should the keyworder use the racist term in fidelity
to the text, or omit it on the basis that it occurs in the text alone (even
though the illustration might include the stereotypical traits associated
with this signifier)?48 Such issues point towards the complex dynamics of
word and image in illustration, suggesting the ways in which keywording
exposes the values of previous historical moments as well as the ideologi-
cal assumptions of our own.
The methods for making the content of illustrations searchable online,
whether computer vision, the use of textual metadata or the activity of
keywording, are not simply ‘tools’ that are applied to the illustrations
from the outside. Rather, they are bound up in the relation between
word and image that defines illustration. The development of digital
illustration resources involves a negotiation of how illustration signifies,
raising the question of the extent to which an illustration can be said to
‘reflect’ or re-produce the text. Illustrations are distinctly visual entities:
they have their own modes of reproduction, their own tropes, allusions,
and traditions, but, like the digital methods that promise to make them
searchable, their visuality is far from unmediated. In illustrations, words
are never completely out of the picture.

Notes
1. Lorraine Janzen Kooistra, The Artist as Critic: Bitextuality in Fin-de-
Siècle Illustrated Books (Aldershot: Scolar, 1995), p. 3.
2. Julia Thomas, Pictorial Victorians: The Inscription of Values in Word and
Image (Athens, OH: Ohio University Press, 2004), p. 15, passim.
60  J. THOMAS

3. Kooistra, The Artist as Critic, p. 9.

4. When I began work on DMVI, Content-Based Image Retrieval was
in its infancy and was just beginning to be discussed in terms of how it
might function in digital image libraries. Margaret Graham, for exam-
ple, suggested that CBIR might have most potential when combined
with text-based image descriptors. Margaret E. Graham, ‘Enhancing
Visual Resources for Searching and Retrieval—Is Content-Based Image
Retrieval a Solution?’, Literary and Linguistic Computing 19: 3 (2004):
321–333. For another account of CBIR and its possibilities and limi-
tations, see Ritendra Datta, Dhiraj Joshi, Jia Li, and James Z. Wang,
‘Image Retrieval: Ideas, Influences, and Trends of the New Age’, ACM
Computing Surveys 40: 2 (April 2008). Available online at http://www.
ri.cmu.edu/publication_view.html?pub_id=6056.
5. See W. J. T. Mitchell, Iconology: Image, Text, Ideology (Chicago: University
of Chicago Press, 1986), p. 56, passim.
6. Matthew G. Kirschenbaum, ‘The Word as Image in an Age of Digital
Reproduction’, Eloquent Images: Word and Image in the Age of New
Media, ed. Mary E. Hocks and Michelle R. Kendrick (Cambridge, MA:
MIT Press, 2003), pp. 137–156.
7. The digitised illustrations that I am referring to in this book are pixel as
opposed to vector graphics (vector graphics are composed of geometric
features, such as lines, circles and curves, rather than individual picture
elements). It is interesting to note that the vector graphics editor, Adobe
Illustrator, uses ‘illustration’ as the title for the types of graphic design
that it makes possible (logos, typesetting, infographics).
8. Wolfgang Ernst, Digital Memory and the Archive (Minneapolis: University
of Minnesota Press, 2013), p. 136.
9. See J. S. Chung, R. Arandjelović, G. Bergel, A. Franklin and A. Zisserman,
‘Re-presentations of Art Collections’, paper presented for the Workshop
on Computer Vision for Art Analysis, ECCV, 2014. Available online at
http://www.robots.ox.ac.uk/~vgg/publications/2014/Chung14/.
10. John Resig, ‘Using Computer Vision to Increase the Research Potential
of Photo Archives’, Journal of Digital Humanities, 3: 2 (2014). Available
online at http://journalofdigitalhumanities.org/3-2/using-computer-
vision-to-increase-the-research-potential-of-photo-archives-by-john-
resig/.
11. Visual Geometry Group, ‘Exploring the British Library’s 1 Million
Images’, http://www.robots.ox.ac.uk/~vgg/research/BL/.
12. John Everett Millais, illustration for Anthony Trollope, Framley
Parsonage, Cornhill Magazine, 1 (April 1860), facing p. 449. Engraved
by the Dalziel Brothers.
 3 SEARCHABILITY  61

13. The ‘similar illustrations’ feature that we developed for The Illustration

Archive uses an algorithm to generate an ordered list of similar illustrations
based on the tags, the sequence in which the tags are entered into the sys-
tem, and whether or not the tag is generated by a user or WordNet. The
notion of image similarity here, then, is explicitly reliant on words.
14. Studies in deep learning and its applicability for computer vision have
tended to use datasets with relatively similar images. See Ji Wan, Dayong
Wang, Steven C. H. Hoi, Pengcheng Wu, Jianke Zhu, Yongdong Zhang
and Jintao Li, ‘Deep Learning for Content-Based Image Retrieval: A
Comprehensive Study’, 2014. Available online at https://pdfs.seman-
ticscholar.org/391b/86cf16c2702dcc4beee55a6dd6d3bd7cf27b.pdf.
Likewise, a recent project that uses computer vision to identify hand ges-
tures and associated objects in medieval illustrations depends on the fact
that these features are heavily standardised in the illustrations. See Peter
Bell, Joseph Schlecht and Björn Ommer, ‘Nonverbal Communication in
Medieval Illustrations Revisited by Computer Vision and Art History’,
Visual Resources, 29: 1–2 (2013): 26–37.
15. For an account of how ImageNet, which relies on tags generated via
Amazon Mechanical Turk, has been used in deep learning, see Ji Wan
et al., ‘Deep Learning for Content-Based Image Retrieval’.
16. ‘Academic Issues’, Marie Duval Archive, ed. Simon Grennan, Roger Sabin
and Julian Waite, http://www.marieduval.org/academic-issues.
17. For more information on abridged titles in The Illustration Archive, see
Ian Harvey and Nicky Lloyd, ‘Lost Visions: A Descriptive Metadata
Crowdsourcing and Search Platform for Nineteenth-Century Book
Illustrations’, Proceedings of the Digital Humanities Congress 2014, Studies
in the Digital Humanities (Sheffield: HRI Online Publications, 2014).
Available online at http://www.hrionline.ac.uk/openbook/chapter/
dhc2014-harvey.
18. John Bradley and Stephen Pigney, ‘Towards a Bibliographic Model of
Illustrations in the Early Modern Illustrated Book’, paper delivered at the
Digital Humanities conference, University of Hamburg, 2012. Available
online at https://lecture2go.uni-hamburg.de/konferenzen/-/k/13923.
I am grateful to John Bradley for sending me an extended version of this
paper.
19. Used in a subset of 30,000 images, the algorithm identified around
200 image creator substrings of which 30–40% were false positives. See
Harvey and Lloyd, ‘Lost Visions’.
20. William Vaughan, ‘Facsimile Versus White Line: An Anglo-German
Disparity’, Reading Victorian Illustration, 1855–1875: Spoils of the
Lumber Room, ed. Paul Goldman and Simon Cooke (Aldershot: Ashgate,
2012), pp. 33–52, p. 38.
62  J. THOMAS

21. For an account of this project, see Kalev Leetaru, ‘500 Years of Book
Images’, http://kalevleetaru.com.
22. Viktor Mayer-Schönberger and Kenneth Cukier, Big Data: A Revolution
that Will Transform How We Live, Work and Think (London: John
Murray, 2013), p. 13.
23. This project ran in 2008 and was funded by the Cardiff Undergraduate
Research Opportunities Programme. The team consisted of two final year
English Literature undergraduate students, Marianne Fisher and Simon
Eckstein.
24. For a discussion of the complexities of these illustrations, see Joan
Stevens, ‘Thackeray’s Pictorial Capitals’, Costerus: Essays in English and
American Language and Literature, 2 (1974): 113–140.
25. Robert Bloomfield, ‘The Farmer’s Boy’, Favourite English Poems
(London: Sampson Low, 1862), pp. 116–193, p. 175.
26. Edward Henry Wehnert, illustration for Samuel Taylor Coleridge, ‘The
Rime of the Ancient Mariner’, Favourite English Poems, p. 223. Available
on the Database of Mid-Victorian Wood-Engraved Illustration. www.
dmvi.org.uk.
27. These page references are for the first book edition of Anthony Trollope’s
Orley Farm, illustrated by John Everett Millais and engraved by the
Dalziel Brothers (London: Chapman and Hall, 1861–1862).
28. This is John Dawson Watson’s illustration, ‘A Summer’s Eve in a Country
Lane’ (p. 97), which illustrates a poem by T. H. (pp. 162–163) in London
Society, 2 (August 1862).
29. Mayer-Schönberger and Cukier, Big Data, p. 43.
30. In the case of their fine art books, the Dalziel brothers would commis-
sion a set of pictures from an artist and hire poets to write a page of
verse to accompany each plate. According to Kooistra, this led to ‘a
new kind of ekphrastic poetry for the gift-book market.’ See Lorraine
Janzen Kooistra, Poetry, Pictures, and Popular Publishing: The Illustrated
Gift Book and Victorian Visual Culture 1855–1875 (Athens, OH: Ohio
University Press, 2011), pp. 102–103.
31. Murtha Baca notes the ‘art information community’s very real need for
practical guidelines on how to lead end users to relevant images of art
and architecture online.’ Murtha Baca (ed.), Introduction to Art Image
Access: Issues, Tools, Standards, Strategies (Los Angeles: Getty Research
Institute, 2002), p. v.
32. For more details on the process of keywording DMVI and the issues it
raised, see Julia Thomas, ‘Getting the Picture: Word and Image in the
Digital Archive’, European Journal of English Studies, 11: 2 (August
2007): 193–206.
 3 SEARCHABILITY  63

33. It is because keywording is so time-consuming that projects and com-

panies are increasingly opting to outsource this work. The Church
Missionary Society Periodicals resource from Adam Matthew outsourced
the tagging of the illustrations to an external company and these tags
were then proof-read by the editorial team. Even this was only possible
because of the relatively ‘uniform nature’ of the images, which made
the task more ‘predictable’. I am grateful to Martha Fogg for the email
exchange about this process.
34. I was comforted to read Morris Eaves’ account of the issues involved in
the creation of the Image Search tool for the William Blake Archive.
His experiences coincide with my own on DMVI, even down to
the brief flirtation with Iconclass. Morris Eaves, ‘Picture Problems:
X-Editing Images 1992–2010’, DHQ: Digital Humanities Quarterly,
3: 3 (2009). Available online at http://digitalhumanities.org/dhq/
vol/3/3/000052/000052.html.
35.  For more information on the use of DMVI’s iconography in NCSE,
see James Mussell, The Nineteenth-Century Press in the Digital Age
(Basingstoke: Palgrave Macmillan, 2012), pp. 108–109.
36. Roland Barthes, ‘Rhetoric of the Image’, Image, Music, Text, ed. and
trans. Stephen Heath (London: Fontana, 1977), pp. 32–51.
37. Peter Mendelsund, ‘What We See When We Read’, The Paris Review, 14
August 2014. I am grateful to David Skilton for pointing me in the direc-
tion of this article. See also Mendelsund’s book, which deals with these
issues, What We See When We Read: A Phenomenology, with Illustrations
(New York: Vintage, 2014).
38.  Gustave Flaubert to Ernest Duplan, 12 June 1862, Flaubert
Correspondance, vol. III (janvier 1859–décembre 1868), ed. Jean
Bruneau (Paris: Gallimard, 1991), pp. 221–222.
39. John Ruskin, The Cestus of Alglaia, The Complete Works of John Ruskin,
ed. E. T. Cook and Alexander Wedderburn (London: George Allen,
1903–1912), vol. XIX, pp. 139–140. Originally given as a paper in the
Art Journal, January 1866.
40. http://artuk.org/tagger/faq/1175.
41. Gotthold Ephraim Lessing, Laocoőn: An Essay on the Limits of Painting
and Poetry, trans. Edward Allen McCormick (1766; Baltimore and
London: Johns Hopkins University Press, 1984), pp. 76–77.
42. ‘First Impressions’, Indianapolis Museum of Art blogpost, 4 May 2011,
http://www.imamuseum.org/blog/2011/05/04/first-impressions/.
43. ‘Editorial Principles’, The Yellow Nineties Online, ed. Dennis Denisoff and
Lorraine Janzen Kooistra, http://www.1890s.ca/Editorial_Principles.
aspx.
64  J. THOMAS

44. This is sometimes the case in the Rossetti Archive, which depends on prose
descriptions to make the content of images searchable. For a discussion of
these issues, see Thomas, ‘Getting the Picture’, p. 196.
45. For more information on this project, see Etienne Posthumus, ‘DMVI
Iconclass Mapping’, Cardiff Book History blogpost, 22 February 2011,
https://cardiffbookhistory.wordpress.com/tag/arkyves-project/.
46. John Everett Millais, illustration for Anon. ‘Margaret Wilson’, Once a
Week 7 (5 July 1862), p. 42. Available online at www.dmvi.org.uk.
47. Myles Birket Foster, illustration for Thomas Campbell, ‘The Pleasures of
Hope’, Favourite English Poems, p. 280. Available online at www.dmvi.
org.uk.
48. This particular example came out of an email correspondence with Kate
Holterhoff, the editor of Visual Haggard, http://www.visualhaggard.
org, an archive of illustrations of Rider Haggard’s novels. As Holterhoff
points out, there may be an ‘ethics of access’ in making some of these
racially discomfiting illustrations accessible. Digital accessibility raises
the question of ‘whether digitization recapitulates an image’s nine-
teenth-century ideology, or whether the online image has become some-
thing altogether new.’ Kate Holterhoff, ‘Ethics and the Digital Archive:
The Case for Visualizing H. Rider Haggard’, Journal of Visual Culture
Online, 21 October 2013. Available online at http://blogs.tandf.co.uk/
jvc/2013/10/21/ethics-and-the-digital-archive-the-case-for-visualizing-
h-rider-haggard/.
 CHAPTER 4

Crowdsourcing

Abstract  This chapter moves away from discussions of crowdsourc-

ing that have focused on its evolution and efficacy to identify this
activity as inherently political. Thomas argues that the crowdsourced
tagging of illustration is a textually constituted process in which the
meanings of illustration refuse to be fixed. The discussion focuses on
three intersecting areas which, together, reveal what is at stake in the
new access to pictures offered by crowdsourced image tagging: the
construction of the tagger and tagging ‘community’ in a crowdsourc-
ing environment; the infrastructure of digital resources and attempts
to control the ‘messiness’ of taxonomies; and the definitions of
‘types’ of illustration that emerge from the development of crowd-
sourcing tasks.

Keywords  Nineteenth century · Illustration · Digital

Crowdsourcing · Tagging

Making the content of images searchable might be a task that is fraught

with issues, but at least it does not have to be a lonely one (Fig. 4.1).
Crowdsourcing, the enlisting of members of the public to perform
online tasks, has become a preferred method for adding value to digital
resources, especially across the galleries, libraries and museums (GLAM)
sector.1 This activity has been employed on text-based initiatives such as
Transcribe Bentham (http://www.transcribe-bentham.da.ulcc.ac.uk)2

© The Author(s) 2017 65

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4_4
66  J. THOMAS
 4 CROWDSOURCING  67

t Fig. 4.1  Charles Bennett, illustration for ‘Paterfamilias Reading the “Times”’,

London Society 2 (September 1862) facing p. 233. Engraved by the Dalziel
Brothers. Available on the Database of Mid-Victorian Illustration www.dmvi.org.uk

and the National Archive’s Citizen Archivist project (http://www.

archives.gov/citizen-archivist/). But it has come into its own in the
domain of image archives, where members of the public are petitioned,
cajoled and enticed into looking at images and describing what they see.
In early image crowdsourcing projects, this petitioning often took the
form of inviting participants to play a game. The ESP Game and its later
incarnation, Google Image Labeler, aimed to markup the content of the
images on Google by asking users to match tags with other users, leading
to a consensus of descriptors that increased the searchability of the mate-
rial.3 This method has carried through to more recent image tagging
ventures like ARTigo (http://www.artigo.org) and the Metadata Games
suite (http://www.metadatagames.org).
As the use of gaming structures imply, there is more to tagging images
than simply tagging images: crowdsourcing is a vehicle for actively
engaging users and promoting image collections.4 Crowdsourcing has
other advantages, too. It is often seen as a solution to image searchability
in big visual datasets because it is one of the only methods that can go
some way towards marking up content that could not be viewed in its
entirety by a limited number of curators. Thus, while keywording was
an effective mechanism for making the content of illustrations searchable
in a closed and manageable corpus of images such as DMVI, the num-
ber and variety of illustrations in The Illustration Archive (over a million
images from diverse books, covering several centuries, and using numer-
ous reproductive techniques) meant that it was impossible to tag these
images ourselves or to mark them up according to a predetermined key-
wording system. We decided instead to crowdsource this task, inviting
users of the Archive to help tag the illustrations.
The experience of developing the crowdsourcing mechanism for The
Illustration Archive led to a stark realisation of the politics of crowdsourced
image tagging, which lies at the heart of this chapter. With the many
advantages that it offers to developers and users, crowdsourcing appears
as a straightforward and neutral activity, a simple crowd-pleaser. However,
this marginalises the extent to which the meanings and implications of
crowdsourced image tagging are textually embodied, generated not only in
the process of tagging, of adding words to images, but also in the way that
68  J. THOMAS

this activity is articulated and described. The aim of this chapter, then, is to
move away from discussions of the history, logistics and benefits of crowd-
sourcing5 and instead to identify this activity as an inherently problematic
one. The discussion focuses on three intersecting areas, which, together,
reveal what is at stake in the new access to images offered by crowdsourced
image tagging: the construction of the tagger and tagging ‘community’ in
a crowdsourcing environment; the infrastructure of digital resources and
attempts to control the ‘messiness’ of taxonomies; and the definitions of
‘illustration’ that emerge from the development of crowdsourcing tasks.
The politics of crowdsourced image tagging is apparent in one of the
earliest and most successful image tagging initiatives: the mounting of
a corpus of images on Flickr by the Library of Congress and the launch
of ‘The Commons’, a designated area of Flickr where cultural heritage
institutions share photographs that are tagged by visitors to the site.6
Central to this project was the strategic objective to increase awareness
of the collection and ‘to expand outreach’, an objective that seems to
have taken priority over making the images searchable, which is why the
tags are ‘free’, that is, without what prompts requisite in crowdsourc-
ing projects that aim for increased image searchability.7 For Mia Ridge,
crowdsourcing is ‘a powerful platform for audience engagement with
museums, offering truly deep and valuable connection with cultural her-
itage through online collaboration around shared goals or resources’.8
Moreover, the audience that is being engaged here is not just the typical
demographic, with a concerted effort being made to reach those who
are not conventionally interested or invested in art. In the case of Your
Paintings, another crowdsourced image tagging project, one-fifth of tag-
gers reported that they had never, or very rarely, visited art galleries.9
There are, of course, any number of ‘external’ reasons why users
become engaged in such resources, and the collaborative and pleasurable
aspects of image tagging might be some of them. But I want to suggest
that at least part of the success (or otherwise) of crowdsourced image
tagging lies in the textual strategies of the resources themselves: the
ways in which they address, construct and enlist the users, or how they
‘play the crowd’. It is noticeable, for instance, that the activity of tag-
ging images is often articulated as way of giving taggers (an illusion of)
‘ownership’ of the works of art: these are ‘Your paintings’. This strategy
has historical precedents in the Victorian period, where numerous meth-
ods were employed to try and interest the working classes in the arts,
from the offer of cheap rail fares to places of cultural significance, to the
 4 CROWDSOURCING  69

opening of galleries at times that fitted in with the working week. Art
was regarded as an ennobling force, able to imbue the working classes
with the values of the middle classes. A century on and John Berger was
lamenting the fact that the history of art was too elitist and appealing to
the public to reclaim it.10 While such ideas seem far removed from the
space of the digital archive, the terms of the debate are strikingly similar.
Crowdsourced image tagging plays a key role in outreach programmes
precisely because it is driven by the assumption that anyone—and every-
one—can tag a picture.11 Its attempts to address and engage the crowd
take place in the context of a concomitant attempt to break down barri-
ers, whether these barriers are intellectual, geographic, social, political or
cultural. Crowdsourced image tagging is defined as a democratic act.
Where crowdsourcing differs from Victorian attempts to engage the
public, however, is in its move away from the notion that cultural institu-
tions can benefit the individual to the idea that the individual can ben-
efit the institution. It is not what your gallery can do for you, but what
you can do for your gallery. As Ridge argues, ‘GLAM projects provide
an opportunity for altruistic acts’.12 This ‘opportunity’ is intensified in
cultural heritage crowdsourcing projects, which, according to Ridge, ‘are
well positioned for appeals to altruism’.13 In fact, several studies have
indicated that an emphasis on altruistic motivations in crowdsourcing
initiatives increases levels of participation.14 This debate has been taken
up by Trevor Owens, who raises the status of the ‘crowd’ by defining
them as ‘amateurs’ in the sense that they are part of the history of ‘con-
tributors to the public good’.15 Owens advocates a model of crowd-
sourcing that empowers a community of viewers, and where tagging is a
valued activity: ‘We are in a position to let the users of these collections
leave a mark on the collections. Instead of browsing through a collec-
tion they literally become the authors of our historical record’.16 What
Owens’ comments reveal is that the role of visitors to digital archives has
radically changed with the onset of crowdsourcing. Crowdsourcing pro-
vides an opportunity for viewers to become active partners rather than
passive recipients in their exchange with the image, to ‘write’ the picture
as they tag it, and this activity compromises the seemingly fixed distinc-
tion between the ‘expert’ and the ‘amateur’.
There is a fundamental sense, though, in which the appeal to altru-
ism that has come to characterise crowdsourcing projects is embedded
in the textual interfaces of these sites. Johanna Drucker has drawn atten-
tion to the fact that ‘An interface is a space in which a subject, not a user,
70  J. THOMAS

is invoked. Interface is an enunciative system’.17 In crowdsourced image

tagging, the motivation to tag does not only (or even necessarily) emerge
from the innate altruism of the tagger, but lies on a more integral level
in the language employed in the interfaces of these digital resources. To
use Louis Althusser’s term, the tagger is ‘hailed’ as an altruistic subject
by the language that pervades these resources and their surrounding
texts, a language that has become so naturalised that it is barely notice-
able. When the Library of Congress mounted its images on Flickr, the
press release stressed the fact that the Library was ‘asking the public for
help’.18 The original Flickr blog post and the announcement of The
Commons came with the exclamation, ‘This is for the good of humanity,
dude!!’19 It is revealing that one of the most frequent words to appear
on sites that include a tagging facility is ‘help’, and this signifier has very
different connotations to its more familiar usage in a digital context: it
is a rallying cry to the user to ‘help’ the resource rather than advice on
where the user can find ‘help’ on how to use that resource. In an analysis
of tagger motivations, 80% of the taggers on Your Paintings identified
‘helping’ as the key incentive for tagging, and this came second only to
interest in the paintings.20
Lev Manovich has coined the phrase ‘Cultural Interfaces’ to describe
how interfaces distribute and allow us to interact with cultural data,21
but interfaces are also culturally significant in and of themselves: they
are modes of representation that generate cultural meanings. This is sug-
gested in the research of Aaron Marcus and Emilie West Gould, who use
Geert Hofstede’s analysis of cultural values in organisations (Hofstede’s
foundational research was conducted in IBM) to argue that Web inter-
face designers should take account of cultural differences if their sites
are to have global impact, whether this is by adapting the use of col-
our (sacred colours differ in various countries), using different count-
ing/dating systems, or ordering the data differently.22 The focus here is
primarily on web design (Aaron Marcus and Associates, Inc. is a user
interface development company), but the implication is that cultural val-
ues are always present in interfaces, whether they are consciously embed-
ded there or not. The interface, then, addresses specific cultures or user
groups that understand its mode of address and enter into dialogue with
it, thereby enabling the chaotic, faceless ‘crowd’ to be harnessed into a
group, or, perhaps more significantly, a ‘community’. In their findings on
how crowdsourcing has been used in humanities projects, Stuart Dunn
and Mark Hedges suggest that the motivations to get involved with such
 4 CROWDSOURCING  71

projects are ‘intimately linked with notions of community, and the sense
of community felt by participants’.23 Certainly, crowdsourced image tag-
ging projects often take their lead from the so-called ‘Citizen Science’
initiatives, which construct their anonymous users as nameable and
acknowledged ‘citizens’. Such a ‘community’ is organised and known,
and in a literal sense, too, since user profiles are frequently stored. Even
as they label pictures, taggers are themselves labelled.
The forging of a ‘community’ with a specific identity was central to
Brooklyn Museum’s Tag! You’re It! game that launched in 2008 with the
objective of keywording art objects in the collection. Taggers here were
described collectively as a ‘posse’, and suitably diverse members of this
posse demonstrated the tagging process in a short video set to music in
which they held up handwritten keywords. The altruistic appeal of this
project was inseparable from the creation of a community, objectives that
were built into the tagging game, where users played against each other:
‘Help us out by playing tag with other posse members. Your tags will
help everyone find objects in our collections’. At the end of the game,
users were told the names of the posse members with whom their tags
matched. Tag! You’re It! was as much about interacting with other posse
members as interacting with the art objects, even though this interac-
tion was illusory since the game was not conducted in real time and the
tags of other participants had already been collected and entered into the
system. The attempt to personalise the posse of taggers was accompanied
by an attempt to personify the museum, to give it a face, or faces. There
were ‘thank you’ videos in which museum professionals addressed the
taggers and told them how well they had done. A prolific tagger could
‘collect’ 12 such videos.
Tag! You’re It! addressed the tagger directly as both a member of
the posse and a ‘subject’. As such, it attempted to dismantle the barrier
between the tagger and the museum. The 12 videos were so appeal-
ing and collectible because of their informality: the tagger encountered
museum workers at the photocopier, on a bench eating lunch, or sat
at a paper-strewn desk. The implication was that these were real peo-
ple, just like the tagger, and that these curators truly valued the tagger’s
input. This informality was mirrored by the informality of the language
of the interface. There were no detailed instructions on how to tag or
any ‘Frequently Asked Questions’, but there was constant encourage-
ment throughout the tagging process: ‘Blast! You just passed’, ‘Work
it! You just passed’, ‘Chirp! You just passed’. The voice of the interface,
72  J. THOMAS

moreover, was a ventriloquised one that also spoke for the tagger: ‘What
else ya got?’, ‘Nah, skip this one’. This linguistic register emulates the
language of the street, of youth culture or an idea of youth culture. It
suggests an attempt to render art more ‘real’ and relevant, less ‘grown
up’. This language was in opposition to the rest of the website, which
adopted the more formal discourse typical of museums: ‘The Museum’s
collections were initially developed, in the early decades of the twenti-
eth century, by such outstanding curators as Stewart Culin, Herbert
Spinden, and William Henry Goodyear, with the generous sup-
port of collectors and donors from Brooklyn and around the country.
Continuing to build upon their pioneering work, the Brooklyn Museum
has amassed one of the largest and most diverse collections in the
United States’.24 It is hard to imagine Stewart Culin, Herbert Spinden
and William Henry Goodyear, the great curators of the past, giving the
thumbs up to prolific taggers.
In July 2014, Tag! You’re It! was decommissioned. A blogpost
by Shelley Bernstein, the Vice Director of Digital Engagement and
Technology at Brooklyn Museum, explained what had gone wrong.25
The problem seems not to have been the number of tags generated (over
230,000), but the fact that this was, indeed, an intimate ‘posse’: the tag-
ging community consisted of only 1100 taggers in the 6 years that the
game was running and the most consistently engaged taggers turned out
to be members of Brooklyn Museum staff. As Bernstein writes, ‘If our
institutional mission centers around community with the aim to engage
a broad audience, are the Posse and our tagging games doing that effec-
tively?’26 The answer, according to Bernstein, was a resounding ‘no’.
Tag! You’re It! was retired on the basis of the limited size and breadth
of the tagging community and the ineffectiveness of the gaming struc-
ture at engaging a wider user group (Bernstein admitted that more than
half of the tags were actually generated directly via the object page rather
than through the game).
Where Tag! You’re It! failed, Your Paintings seems to have succeeded,
having accumulated over six million tags inputted by 12,000 taggers
before the site was rebranded as Art UK in 2016. There was no gam-
ing incentive here, although there were different coloured paintbrushes
that indicated the number of paintings tagged by the user. In terms of
the motivational techniques that the site used to increase and sustain
participation, there was the familiar use of the language of altruism,
but here it came with an added dimension that may help to explain the
 4 CROWDSOURCING  73

project’s impressive results. In Your Paintings, the focus was on the own-
ership and national identity of the user: these are ‘your paintings’, and
the ‘you’ to whom these images ‘belong’ was addressed specifically as a
British citizen. The appeal to altruism, then, was couched in nationalist
terms: ‘help tag the nation’s art!’ the site implored its visitors.27 Visitors
to Your Paintings were constructed, and actively engaged, as British
Citizens, who would be even better citizens if they tagged the images on
display. Paradoxically, however, at the very moment that Your Paintings
addressed its ‘British’ users, it also revealed the ambiguous notion of the
‘crowd’ or ‘community’ that underlies such tagging projects: whether it
is a posse or a nation, the ‘crowd’ is always exclusive as well as inclusive.
And just as the taggers are characterised and labelled by the digital
environment with which they interact, so are the tags themselves. The
steve.museum, a collaboration of museums and galleries that was estab-
lished to examine the possibilities of user-generated descriptions of art,
acknowledged that ‘the anarchy of emergent folksonomy seemed a cause
for concern’.28 Of particular ‘concern’ was the apparent gap between
user-generated tags and the descriptions of the images provided by
experts. It was estimated that up to 90% of the terms users contributed
were not present in the museum documentation for the art object.29 The
language employed in such accounts of crowdsourced tagging suggests
that the problematics of this activity extend to the terms in which the
tags themselves are described. Tags are never merely tags. They acquire
a political momentum: they can be good or bad, fixed or unruly, ‘eve-
ryday’ or expert, valuable or valueless, anarchic or democratic.30 The
Powerhouse Museum in Sydney, Australia, advocates the use of crowd-
sourcing to mark up its collections (users can delete as well as add key-
words) but the resultant tags are defined as informal and potentially
unreliable, their difference from the more formal tags stressed in their
visual incarnation as a word cloud.31 Such a tactic has parallels in archi-
tectural restoration projects that make visible the new additions to build-
ings so that they can never be mistaken for the original. Powerhouse
even provides a ‘warning’ to visitors: ‘Sometimes museums describe
objects in language that is highly specialist and user added keywords
are useful in bridging the “semantic gap” between the language of the
museum and that of the user…. Please note that this is an experimental
way of navigating the collection and that the “accuracy” of these key-
words is not verified by the Museum. We recommend their use primarily
as exploration and navigation tools’.32
74  J. THOMAS

The language of the museum and the tagger is strictly demarcated

here, with that of the museum regarded as ‘highly specialist’ and, by
implication, authoritative. This distinction between expert and crowd-
generated tags frequently revolve around a hierarchical notion of the
‘quality’ or ‘value’ of the markup: if user-generated tags are ‘added-
value’ at best, then the implication is that originary (that is, ‘true’) value
resides with the ‘expert’ tags. Some critics have argued the opposite.33
A recent study asserted that there are no significant differences between
professionals and external participants in terms of tagging patterns and
accuracy levels. Again, though, the focus of this study is on what it calls
the ‘value and accuracy of folksonomies’ (my emphasis).34 In the unfor-
tunate case of Tag!You’re It!, there was the conviction that the tags of
the crowd were actually more ‘valuable’ than those of the experts.
Indeed, it was the game’s failure to attract taggers outside the museum
that led to its demise. As Bernstein notes, ‘If tagging is meant to democ-
ratize collections by applying everyday words instead of specialized ones,
you have to wonder how much traction we were getting if the majority
of tags were coming from specialized voices’.35
The potential for the democratisation of art, it seems, resides not so
much in the collections or even the institutions, but in the language
used to tag the pictures, the ‘everyday words’ of the crowd. According
to Bernstein, tagging practices have changed dramatically in recent years
with the emphasis on ‘a more social language, not a descriptive one’
and taggers moving away from organised (and controlling) tagging ini-
tiatives: ‘The good news is that people are finally psyched about tagging
stuff. On their own. Without institutional prompting’.36 Such ‘inde-
pendent’ tagging has been spurred on by social media platforms like
Twitter, Instagram and Facebook, and finds its origin in Flickr, where
the Library of Congress took a deliberate ‘hands off’ approach to the
tagging of its images.37 At the time of original upload, each photograph
included only three tags: one ‘regular’ tag (‘Library of Congress’) and
two machine tags to correlate the Library of Congress and Flickr pho-
tographs through identification numbers. All other tags were added by
users. Flickr set limits on the number of tags that could be added to each
photo at 75, but they could be in any alphabet, in lowercase or upper-
case, single words or characters or multi-word phrases. No distinctions
were made between ‘types’ of tags such as those that identify creators,
places, time periods, genre, format, subject (of and about) or general
associations. Apart from being checked for ‘blatantly inappropriate con-
tent’, the tags were left to their own devices.38
 4 CROWDSOURCING  75

There is a sense, of course, in which this laissez-faire approach to

image tagging does enrich the material and its searchability. Using Flickr
as an example, Viktor Mayer-Schönberger and Kenneth Cukier identify
such tagging as indicative of the move from ‘clean taxonomies’ struc-
tured according to a pre-set taxonomy, to ‘messy’ ones.39 This move,
they argue, is a beneficial one: ‘in return for messiness in the way we
organize our photo collections, we gain a much richer universe of labels,
and by extension, a deeper, broader access to our pictures’.40 There are
circumstances, however, when tags cannot be ‘messy’, when they have
to be curtailed and controlled. Perhaps in some ways they always are
controlled. The monitoring of ‘blatantly inappropriate content’ by the
Library of Congress, for example, compromises the extent to which it
actually adopts a ‘hands off’ approach. Certainly, this control becomes
a necessity in those tagging initiatives where the objective of making
images searchable takes priority over attempts at public outreach. Such
projects have to limit the ‘messy’ taxonomies associated with social
media platforms and aim for a structured and consistent tagging that
allows the images to be searched more effectively. This is the goal of
commercial image banks like photographic archives, which have an eco-
nomic interest in ensuring that the pictures are as searchable as possi-
ble. As Doireann Wallace points out, such archives formulate keywording
clusters that aim ‘to stand in for’ or ‘be synonymous with the image’.41
One method for achieving this is to incorporate tagging by consensus
into the infrastructure of the project, like the gaming environment, ESP,
which validated tags by agreement. The objective of consensual tagging
is to reach the definitive keyword, to subsume multiple, plural readings
of the image in favour of a singular, unified one. Any incongruous (or
unique) tags are discarded along the way. In Your Paintings, 15 view-
ings of each image were required to reach a consensus of tags before the
image was regarded as complete and the tags were loaded into the data-
base and went live on the site.42
In The Illustration Archive, the tagging process was structured using
an auto-correct feature connected to WordNet (https://wordnet.prince-
ton.edu/), a lexical database of the English language that groups words
together as synonyms or ‘synsets’ to create a network of related terms
and concepts. WordNet standardises the tags inputted by the user, while
also enabling the storage of semantically linked tags that can themselves
be searched.43 The process for retrieving and storing the tags in the
Archive works like this: the user types a word (or partial word) into the
76  J. THOMAS

auto-completing text field, which offers multiple meanings of that word

from the WordNet database (the word ‘book’, for example, has 15 pos-
sible meanings); the user selects the most appropriate meaning of the
word she/he has typed. At the back-end of the Archive, the set of syno-
nyms (or ‘synset’) for the entered word is recorded along with a hierar-
chy of up to three hypernyms (that is, the broader, general categories
into which the word falls: ‘animal’, for example, is a hypernym of ‘dog’);
for each of these hypernyms, up to two other synonyms are retrieved;
and the tags, as well as a reference to the ‘distance’ of the synonym from
the user-entered tag, are stored in the database for future searches.
WordNet does more than complete the tag: it directs the tagger,
indicating the most ‘appropriate’ meaning of the tag. It can even offer
a way of dealing with the issue of politically inflected mark up, so that,
for example, a racially provocative term could be entered into the data-
base, but the dictionary definition provided by WordNet places it in the
context of ‘an offensive and insulting term’. What these ‘problem tags’
and the qualifiers provided by WordNet suggest, of course, is that lan-
guage is always political, bound up in the values of its cultural moment.
However systematic WordNet seems, it is not an impartial or objective
system, and its own place in the generation of cultural meanings is hinted
at in the order in which the system presents its list of possible meanings.
According to the Princeton University site, ‘WordNet senses are ordered
using sparse data from semantically tagged text. The order of the senses
is given simply so that some of the most common uses are listed above
others (and those for which there is no data are randomly ordered). The
sense numbers and ordering of senses in WordNet should be considered
random for research purposes’.44 But this largely ‘random’ order can
have some interesting results. In The Illustration Archive, we overrode
the sense order of the word ‘woman’ because the first meaning to appear
on our drop-down list was ‘a human female employed to do housework’.
While the vocabulary of user-generated tags is curtailed using fea-
tures like WordNet, the nature of the tags, what aspects of the image
they actually describe, is systematised by asking the tagger to respond to
specific questions about the picture. The Illustration Archive follows the
model of projects like Your Paintings and those hosted on Zooniverse,
which ask a set of clearly defined questions. In Your Paintings, the tagger
was required to look at the painting displayed and answer these ques-
tions: What things or ideas can you see in this painting? Can you name
any people in this painting? What places are shown in the painting? Does
 4 CROWDSOURCING  77

this painting relate to any event? What subjects do you see in this paint-
ing? What type of painting is this? Very little is left to chance in terms of
how the tagger responds to these questions. The keywords entered for
the ‘things and ideas’ question were connected to definitions from the
Oxford English Dictionary and the tagger selects the most appropriate
term. Likewise, the subject categories for the painting (including ‘Home
and family’, ‘Ideas and Emotions’, ‘Religion and belief’) appeared as a
drill-down list, and the options for the ‘type’ of painting were also pre-
sented to the tagger in list form.
According to Dunn and Hedges, the editorial control in crowdsourc-
ing projects is ‘reassuring’ to users: ‘There also needs to be some ele-
ment of peer review/quality control to reassure the community that
their efforts create academically viable outputs’.45 Such control, how-
ever, exposes an implicit uneasiness about unruly tags embedded in
these systems, a suspicion that, instead of calling a spade a spade, taggers
might call it a shovel, or they might upset the entire system by labelling
it a hoe. The activity of tagging is reined in, so much so that, to some
extent, the resultant descriptions of the images derive as much from the
structures and methodologies of the tagging systems as from the tags
themselves. The architecture of crowdsourced image tagging quite liter-
ally provides its own added value.
In The Illustration Archive, targeted questions about the illustrations
function as cues to the user, adding a formal structure to the tags that,
in turn, optimises the searchability of the images. The questions asked of
the tagger are, therefore, crucial because they determine how the digital
archive can be interrogated and what information can be gleaned about
the illustrations. Initial attempts to adapt the tagging questions used in
Your Paintings for The Illustration Archive led to an awareness of the
very differences between these artistic genres. While a relatively open
question like ‘What things or ideas can you see in this painting?’ could
usefully be adapted for illustration, other questions are not so relevant.
The question ‘Can you name any people in this painting?’ (the examples
given on the site were those heavyweights of British history, Winston
Churchill, Florence Nightingale and Horatio Nelson) suggests the prom-
inence of portraiture; while the question, ‘Does this painting relate to an
event?’ (the examples given were the Battle of Trafalgar, the First World
War and the Coronation of Queen Elizabeth II) refers to the tradition of
history painting. While illustrations do, of course, include portraits and
represent historic events, these are not distinct genres that they are in
78  J. THOMAS

painting. Even an open question like ‘What subjects do you see in this
painting?’ raises questions about how illustration signifies. Sara Shatford
Layne offers useful advice to art cataloguers on how the ‘subjects’ of
works of art might be classified. As she acknowledges, subject access is
‘One of the most important means of enabling users to locate art images’
but it is also ‘a complex and sometimes messy process’.46 Identifying the
subject of a picture, it seems, can be as ‘messy’ in formal classifications
as it is in user-generated folksonomies. This is exacerbated in the case of
illustration because the ‘subject’ or content of the picture is not neces-
sarily contained within the frame of the image as it is in painting, but is
generated in the interaction between the image and the text.
Illustration, then, needs to be defined in its own terms rather than
by the definitions of fine art. One of the most straightforward tagging
questions in terms of painting, ‘What type of painting is this?’, is actu-
ally one of the most problematic when it is asked of illustration. Your
Paintings offered its taggers suggestions on how to answer this question:
a painting might be ‘abstract’, a ‘landscape’, ‘portrait’, ‘seascape’ or ‘still
life’. Again, however, while an illustration may appear ‘abstract’ (like the
images in some twentieth-century artists’ books, for instance), this is not
the same thing as an ‘abstract’ painting; neither is an illustrated landscape
the same as a painted one. And there are ‘types’ of image peculiar to
illustration that do not have equivalents in painting: illustrated advertise-
ments, vignettes, decorations, pictorial capitals, titlepages, frontispieces
and tailpieces.
Our objective in asking taggers in The Illustration Archive to spec-
ify the ‘type’ of illustration was not to formulate a definitive, classifica-
tory list of illustrations, but, rather, to make some sense of the diverse
illustrations contained in the British Library dataset and to allow the
end user to search the Archive for broader categories of illustration as
well as particular iconographic features or concepts, to be able to find
a ‘scientific’ or ‘photographic’ illustration alongside those illustrations
that depict ‘cats’ or ‘bridges’. But just as there is no standard formula
for bibliographically describing an illustration, neither is there an estab-
lished typology of illustrations. The very notion of a ‘type’ of illustra-
tion is fragmented and multifaceted. ‘Type’ itself has meanings that cross
the boundary between the visual and textual: a type in its earliest deriva-
tion is a picture of something, a representation, an image, but the word
also resonates with letters and printers’ type. Likewise, there is a slippage
between the typology of an illustration and its iconographic content. The
 4 CROWDSOURCING  79

‘type’ of illustration might be the ‘mode’ of illustration, which impacts

on where and how the image appears in relation to the text (for example,
plate, vignette, titlepage, frontispiece and pictorial capital), or it could be
the reproductive technique by which an illustration is reproduced (etch-
ing, engraving, lithography, etc.), its generic category (advertisement,
portrait, literary, etc.), or its subject matter (images depicting buildings,
flora, fauna, etc.). While ‘types’ of illustration do not necessarily coin-
cide with types of painting, neither do they coincide with textual genres.
The ‘type’ of illustration is not tied to the type of book in which the
illustration appears. Books are classified according to established conven-
tions, albeit conventions that have changed over time, but illustrations
do not easily conform to the same classifications. This is strikingly appar-
ent in the case of the illustrated books in the British Library dataset,
which are broadly categorised as ‘literature’, ‘philosophy’, ‘history’ and
‘geography’. While an idea of what constitutes a ‘philosophical’ book
might be relatively straightforward, even in its wider nineteenth-century
sense, it is difficult to imagine what a ‘philosophical’ illustration would
look like. And where do the boundaries lie between a ‘historical’, ‘liter-
ary’ or ‘geographical’ illustration? An illustration of the location of the
Battle of Bosworth Field could well be ‘literary’, ‘geographical’ and/or
‘historical’. Illustrations jump between textual genres (an illustration of
Falstaff can appear in a book about the Thames47) and there are often
multiple ‘types’ of illustration within a single book. Moreover, there is
no standard vocabulary for describing even those ‘types’ of illustration
that are specific to this mode of representation: illustrated letters are
known as ‘pictorial capitals’, ‘decorated initials’, ‘historiated letters’, ‘illu-
minated letters’; and the illustrative ‘flourishes’ that appear in books and
periodicals are ‘decorations’, ‘embellishments’, ‘fleurons’, ‘motifs’ ‘orna-
ments’.48 Such plural terms are not necessarily synonymous, but can have
different meanings and values.
As part of a text and separate from it, spoken for and unclassified,
illustration is stateless and nameless. The development of the mecha-
nisms for making the images searchable in The Illustration Archive meant
that we had to overcome this lack of vocabulary and produce a work-
able, even if slippery, definition of illustration, which enabled us to ask
the question, ‘What type of illustration is this?’ We began by examining
the tools available for subject and genre indexing of images. The Art and
Architecture Thesaurus Online (AAT) (http://www.getty.edu/research/
tools/vocabularies/aat/), developed by the Getty Research Institute,
80  J. THOMAS

is a hierarchically structured controlled vocabulary and classification

scheme ‘intended to provide terminology and other information about
the objects, artists, concepts, and places important to various disciplines
that specialize in art, architecture, and material culture’.49 AAT emerged
in the 1970s as a response to the growing automation of indexing, but
for all its structures, hierarchies and semantic linking, the AAT, like many
other indexing systems that we investigated, is astonishingly arbitrary
when it comes to illustration. The entries on ‘illustration’ are scattered
across various ‘trees’: ‘image-making processes and techniques’ leads to
‘Illustration’, ‘Fashion illustration’, ‘Scientific Illustration’ and ‘Technical
Illustration’; the broad category of ‘visual works by subject type’ leads
to ‘Topographical Illustration’; ‘illustrations by form’ includes ‘figures’,
‘frontispieces’, ‘marginal illustrations’ and ‘plates’; and ‘illustrations by
subject type’ consists of ‘botanical’, ‘courtroom’, ‘fashion’, ‘scientific’
and ‘technical’ illustrations.
Like the AAT, the Library of Congress Subject Headings (LCSH)
(http://id.loc.gov/authorities/subjects.html) also slips between the
vague and specific when it comes to classifying illustration, although
this tool is more extensive in its list of types. In terms of what might be
defined as a ‘type’ or ‘subject’ of illustration (what LCSH calls a ‘topic’),
LCSH lists ‘scientific illustration’ (to include anthropological, biologi-
cal, medical, natural history), ‘botanical illustration’ (to include flower
and fruit illustration), ‘technical illustration’, ‘biological illustration’ (to
include botanical and zoological, medical), ‘natural history illustration’
(to include biological, botanical, paleontological and zoological), ‘orni-
thological illustration’, ‘fashion illustration’, ‘archaeological illustration’,
‘animal illustration’ and ‘courtroom illustration’. (The apparently incon-
gruous category of ‘courtroom illustration’, which also features in the
AAT, seems to reflect the presence of images of court scenes in nine-
teenth-century American newspapers.) LCSH appears as a catch-all sys-
tem, but it is heavily weighted towards scientific illustrations. ‘Literary
illustration’ is not listed here and these images can only be found under
bibliographic information attached to specific authors (there is a head-
ing for ‘Shakespeare, William, 1564–1616—Illustrations’), countries
(‘English Literature’—Illustrations’) and titles (‘Beowulf—Illustrations’).
While we were able to take our lead from these indexing systems, we
had to adjust and supplement them because they did not always include
illustrations that we knew were in our dataset, like illustrated advertise-
ments, titlepages and various modes of decoration such as decorative
 4 CROWDSOURCING  81

borders and pictorial capital letters. Our list of types also needed to take
into account the demands and protocols of crowdsourced image tagging,
which were often at odds with the ‘expert’ curatorial tagging assumed
by subject indexes. For example, we took the decision to visually define
our illustrative types using images from the Archive because our tagging
‘community’ was a potentially diverse one that might include anyone
from school-going children to practising illustrators. Each ‘type’ of illus-
tration is displayed on the site in the form of a thumbnail illustration as
this screenshot shows (Fig. 4.2):
Although many of the ‘types’ of illustration that are displayed in
this screenshot drill down to other ‘types’ (choosing ‘a decoration’,
for example, leads to ‘decorative border’, ‘decorative letter’, ‘coat of
arms’ and ‘decorative motif’), the main categories needed to be concise
enough to be displayed on a single screen, which meant that we were
highly selective in what we included as ‘types’. A key part of this selec-
tion process lay in determining if the aspect of an illustration that might
be called a ‘type’ was likely to be tagged in subsequent questions (the
marking up of an aspect of an image in The Illustration Archive only
requires inputting once for the image to be retrieved using this term).
The second question we asked the tagger, ‘Are there any other things or
ideas in this illustration?’, mops up the majority of iconographic informa-
tion about the image, meaning that we could remove certain categories
from our list of types, such as ‘animal illustration’ (if the tagger input-
ted ‘dog’ in the second question, the hypernyms for this term would
automatically be added by WordNet).50 Another determinant in what
was included and excluded from our list of illustrative types was the rela-
tive ease with which these types could be identified by the tagger. This
applied particularly to types of reproductive technique, which are diffi-
cult to identify, especially in digitised versions of images where the qual-
ity is not always consistent. We included ‘photography’ on the basis that
it is easier to recognise than most of the other processes, and the system
allows details of reproductive techniques or, indeed, any other additional
information to be inputted by the tagger in a free text box at the end of
the tagging task. The titles of the books, which would be interrogated in
a general search, also often include information about the reproductive
technique, especially if it was regarded as a marketable feature.51
The question ‘What type of illustration is this?’ involves a confron-
tation with ideas of ‘illustration’ that are determined by the develop-
mental processes that have gone into articulating this tagging question
82  J. THOMAS

Fig. 4.2  Screenshot from the ‘Tag Illustrations’ page, The Illustration Archive:
http://illustrationarchive.cardiff.ac.uk
 4 CROWDSOURCING  83

in the first place. To put it simply, there is a sense in which the defini-
tions of ‘illustration’ are created by the digital archive itself, and these
definitions are multiple and unstable, transformed in the click of an
illustrated button. This comes to the fore in Science Gossip (https://
www.sciencegossip.org), a Zooniverse tagging project that crowdsources
information about the illustrations in nineteenth-century science peri-
odicals, including Hardwicke’s Science-Gossip, the journal that gives
the project its name. Part of the AHRC-funded project, ‘Constructing
Scientific Communities: Citizen Science in the nineteenth and twenty-
first Centuries’, Science Gossip makes parallels between the current infor-
mation revolution and the nineteenth-century revolution in print and
journal publishing when the number of science periodicals jumped from
around 100 titles in 1800 to 10,000 by the end of the century.52 This
crowdsourcing project has two main objectives: it is intended to sup-
plement metadata on the illustrations by adding information about who
created the images, their subject matter and any particular species they
portray; and it sets out to make an explicit connection between nine-
teenth-century citizen scientists who contributed to these periodicals
and the twenty-first century citizen scientists doing the tagging. As the
site states, ‘This is the first Zooniverse project where citizen scientists are
both the researchers and the subject of the research’.53
As an investigation into the communication of science in the Victorian
period and today, the project is extremely valuable, but its negotiation
of illustration is more problematic. This is exposed in the seemingly
transparent question that Science Gossip begins by asking its taggers, ‘Are
there any illustrations on this page?’ Like The Illustration Archive, this
site is forced to confront the difficulty of defining ‘illustration’, a dif-
ficulty that is compounded in the absence of an established vocabulary
around illustration. The ‘types’ of scientific illustration given as exam-
ples here are ‘drawing/painting/diagram, chart/table, photograph and
map’; but what is most revealing are those illustrations that are excluded
from this definition. The pages of these science journals are full of illus-
trated advertisements, decorated titlepages and pictorial capital letters,
but these are not defined as illustrations and, as such, are not identified
as ‘illustrations’ by the taggers. As Science Gossip openly declares, ‘we’re
only looking for certain types of illustrations’. The identification of only
‘certain types of illustrations’ is the source of some anxiety for its ‘citizen
scientists’, though. In one discussion thread, opening with the subject
line, ‘When is an illustration an illustration?’, the tagger writes, ‘On this
84  J. THOMAS

page there is a rather nice geometric design header. Is this an illustra-

tion? I have decided not, but would like to know. The drawing is actu-
ally an “illuminated” capital T. It is beautiful but doesn’t actually depict
a Sunflower which is the title of the piece. Is this an illustration? I have
assumed Yes…but would like to know.’ At this point, a moderator inter-
venes and comments, ‘I’ve not been marking these as the instructions say
not to mark illustrated letters—but they are rather nice! Hope someone
clarifies for us’.54
So, when is an illustration an illustration? The Help Board on Science
Gossip includes definitions of these ‘non-illustrations’ and how they
should be dealt with: ‘EXAMPLES OF PAGES WHICH SHOULD
BE MARKED AS HAVING “NO ILLUSTRATION”. We do not need
these types of page to be classified (e.g. text, blank pages, title pages, book-
plates, book covers, indexes, tables of content, decorative letters or stamps and
advertisements). Please mark as no illustrations on the page’.55 To reiterate
the point, these instructions are given for a second time a little further
down the Help Board:

DO I MARK ILLUSTRATIONS THAT HAVE NO RELATION TO

THE TEXT?

Decorative letters - We are not asking for these to be classified as illustra-

tions, as they were purely decorative. Treat them the same way that you
would classify a title page. Science Gossip (the journal), after its first cou-
ple years, started to place these ornamental initials at the start of each new
issue. The interesting part is that they made a number of different blocks
(wood blocks that is) for each initial letter - and would reuse those blocks
whenever a capital ‘T’ - for instance - was needed. From what I have seen,
this was only a practice for Science Gossip, and not for the other journal…

Illustrations purely for decoration - Illustrations that serve no other pur-

pose than decoration are marked ‘no illustrations’ on page.56

But can ‘decorative letters’ be disregarded on the basis that they ‘have
no relation to the text?’ The very presence of these images on the page
means that they are, inevitably, bound up in their interaction with the
words, even if this interaction is one of disjunction. The example given
on the site, a vignette of shells, crabs and seaweed that appears on a
page about astronomy, is a case in point. The image might not depict
anything explicitly ‘astronomical’, but this very deviation signifies.
 4 CROWDSOURCING  85

Even leaving aside their relation to the text, such images make mean-
ings in other, distinctly visual, ways that call into question their defini-
tion as ‘purely decorative’. Decorations and ‘ornaments’ actually have a
key place in the history of the book, constituting some of the earliest
examples of graphic design. This significance has been attested in another
digital project, Fleuron (https://fleuron.lib.cam.ac.uk/), which contains
thousands of images of printers’ ornaments extracted from the pages of
eighteenth-century books. A century on from these fleurons and con-
temporaneous with the journals in Science Gossip, the role and aesthetic
importance of decorative and ‘ornamental’ illustrations was being revived
and redefined. Owen Jones’s classic book on the subject, The Grammar
of Ornament, was published in 1856 and stands alongside the ‘ornamen-
tal’ philosophy of the Kelmscott Press. Walter Crane uses the signifier
‘ornament’ in preference to ‘illustrate’ because of its associations with
the ideals of fine art.57 James Mussell has made a convincing case for the
recognition, in both theoretical and digital terms, of all the ‘visual codes’
that exist on the pages of nineteenth-century newspapers and periodi-
cals.58 Lorraine Janzen Koositra has similarly attested to the important
place of ornamentation in the periodical press in this period. Referring to
the Scottish periodical, The Evergreen, she comments on ‘the structural
patterns that emerge from the cumulative effect of the designs across
the periodical’s pages’.59 It is easy to envisage that the ornaments on the
pages of Hardwicke’s Science-Gossip function in the same way, but the
refusal to acknowledge their status as illustrations means that they could
be lost in the very digital space that promises to ‘unlock’ so many ‘illus-
trative treasures’.60
What is also lost in the definition of these ‘decorations’ as ‘not-
illustrations’ is the opportunity to explore the practice of recycling and
modifying illustrations to which the editor of Science Gossip refers in the
comments about the reuse of ‘decorative letters’. Indeed, the very pres-
ence of such reused images in Hardwicke’s Science-Gossip attests to the
ubiquity of a practice that crossed the boundaries of publications. The
editor, Charles Knight, for instance, often reused illustrations across his
versions of Shakespeare’s Works (which included a Stratford edition,
Pictorial edition, Cabinet edition, National edition and Imperial edition).
Even the illustrations in a seemingly stable form like the Victorian gift
book are frequently re-mixed. As Kooistra notes, ‘publishers increasingly
took advantage of the gift book’s modular form and recycled poems and
pictures from old stock in different combinations, tricked out in new
86  J. THOMAS

bindings’61 Indeed, this was an activity that crossed continents as well

as publications. John Cassell, the publisher responsible for some of the
most popular illustrated periodicals and books in the period, acquired
electrotypes of pictures from French illustrated journals suitable for
whatever publication he had planned.62 These electrotypes were used
liberally, including in works of natural history and travel, in periodicals
like the Quiver and the Family Paper, and in books for children.63 It was
actually through the reuse of electrotypes that GustaveDoré’s reputa-
tion was established in Britain: when Cassell discovered Doré’s edition
of Don Quixote in progress in France, he published a large number of
his illustrations from the French electrotypes.64 Certainly, it was common
for publishers of illustrated works to build up stockpiles of electrotypes
and used wood engravings, which would then be passed on and sold
to other printers and publishers. Cassell’s stock rose from 20,000 in the
1860s to 180,000 at the end of the 1880s. When George Manville Fenn
resigned the editorship of Cassell’s Magazine to work on Once a Week, he
frequently used Cassell’s stockpile to supplement his illustrations.65
Although the reuse of illustrations might seem a dubious practice, the
biographer of Cassell’s publishing house points out that it was ‘legitimate
enough’, except in those cases where a publisher infringed the right of
an English counterpart, who had acquired the same electrotypes first.66
But, however ‘legitimate’ or otherwise, the reuse of illustrations caused
some concern for authors. W.H.G. Kingston, who had been commis-
sioned in 1870 to write an adventure story of a South Polar exploration,
had to make some major adjustments to his text, including ditching the
first three chapters, when he saw the illustrations that he was expected
to write to. The pictures had been recycled from an old book of voy-
ages and were ‘miserably poor’.67 While he simply could not use some
of these pictures, Kingston commented sardonically that he might be
able to justify the use of others by introducing into his narrative ‘the sup-
posed discovery of an ancient ship, cast away a century before’.68
The young readers of Kingston’s illustrated book probably did not
notice the reused illustrations, but there were attempts on the part of
publishers to cover up this practice. In 1871, the novelist Thomas
Mayne Reid wrote to the publisher George William Petter advising him
to remove the French artists’ and engravers’ names on illustrations that
were being reused in a book of illustrated travels:
 4 CROWDSOURCING  87

Don’t you think that it would be judicious to take the French artists’ and
gravers’ names from the pictures in Illustrated Travels? They take away the
idea of originality as your readers may easily know them to be French, and
there is no reason why they should know it. I adopted the plan myself with
some of your cuts which I purchased from Mr Turner in New York for my
magazine there. The least touch of the chisel will do it without damaging
the picture in any way.

A reader who suspects an article to have been written to an illustration,

and not vice versa, will doubt its genuineness; and many readers know the
names of these French fellows familiarly enough.69

This is an extraordinary statement, not so much because of the air of

duplicity that hangs over this advice, although this is striking enough
(‘The least touch of the chisel will do it’), or the hint of anti-French sen-
timent, but in the assumption that knowing that the illustrations have
been used before will influence how the text is read. Reused illustra-
tions undermine the ‘originality’ of the publication and subvert what
is increasingly regarded as the normative illustrative hierarchy in which
an illustration illustrates an anterior text.70 The very idea that a text has
been ‘written to an illustration’ compromises its ‘genuineness’, its claims
to authenticity and authority, and questions the notion that ‘illustration’
is the relation between an image and a single, identifiable and static text.
The reuse of illustrations is not a marginal aspect of the history of
illustration, either in terms of its frequency or its significance. On the
contrary, this widespread practice has implications for how illustration is
understood in its economic and material context as well as in its gen-
eration of meanings. This practice has been little understood, primar-
ily because it is so difficult to trace reused illustrations across thousands
of publications, but digital resources make it possible for the first time
to examine this reuse in a more systematic way. This is only possible, of
course, if these images are actually defined as ‘illustrations’. As a ‘not-
illustration’, the ‘beautiful’ illuminated capital T identified by the tagger
in Science Gossip will be lost to the end users of the resource, along with
its numerous reiterations and incarnations.
Science Gossip is indicative of the extent to which digital illustration
projects have to confront the problematic definition of illustration. This
issue is intensified because, while this definition has not been fixed by
88  J. THOMAS

established conventions, the demands of crowdsourced image tagging

necessitate that it is fixed, or that it seems to be. In The Illustration
Archive, we decided that musical scores and images of handwriting were
less likely to be types of illustration than titlepages, advertisements, maps,
pictorial letters and other ‘decorations’. On this basis, we excluded these
categories from our list of types of illustration, although they would still
be identified in the second tagging question.71 Science Gossip, on the
other hand, follows what may be described as a ‘scientific’ rationale that
holds that the ‘purely decorative’ is not illustrative. Crowdsourced illus-
tration tagging is a space where these meanings are generated and come
into conflict not only in the activity of tagging the images, but also in
the way that this activity is defined and articulated, from the construc-
tion of the tagger and the tagging ‘community’ to the language used to
describe the tags. Ultimately, the question ‘What type of illustration is
this?’ might be impossible to answer, but it is a question that opens up
the unstable and plural meanings of this mode of representation and of a
digital space where the notion of ‘illustration’ is always in process.

Notes
1. For a sense of the range of these projects, see Mia Ridge (ed.),
Crowdsourcing our Cultural Heritage (Aldershot: Ashgate, 2014).
2. For an outline of this project, see Tim Causer and Valerie Wallace, ‘Building
A Volunteer Community: Results and Findings from Transcribe Bentham’,
DHQ: Digital Humanities Quarterly 6:2 (2012). Available online at
http://www.digitalhumanities.org/dhq/vol/6/2/000125/000125.
html. The crowdsourcing aspect of Transcribe Bentham is discussed in Tim
Causer and Melissa Terras, ‘“Many Hands Make Light Work. Many Hands
Together Make Merry Work”: Transcribe Bentham and Crowdsourcing
Manuscript Collections’, Crowdsourcing our Cultural Heritage, ed. Mia
Ridge (Aldershot: Ashgate, 2014), pp. 57–88.
3. See Luis von Ahn and Laura Dabbish, ‘Labeling images with a com-
puter game’, Proceedings of the SIGCHI conference on Human Factors
in Computing Systems (Vienna, Austria: ACM, 2004), pp. 319–326.
Available online at http://ael.gatech.edu/cs6452f13/files/2013/08/
labeling-images.pdf.
4. A recent study has identified three main aims of crowdsourcing: to explore
new forms of public engagement, to enrich institutional resources,
and to build novel resources. Laura Carletti, Derek McAuley, Dominic
 4 CROWDSOURCING  89

Price, Gabriella Giannachi and Steve Benford, ‘Digital Humanities and

Crowdsourcing: An Exploration’, paper delivered at MW2013: Museums
and the Web conference, 2013. Available online at http://mw2013.
museumsandtheweb.com/paper/digital-humanities-and-crowdsourcing-
an-exploration-4/.
5. An example of this approach can be seen in Melissa Terras,
‘Crowdsourcing in the Digital Humanities’, A New Companion to Digital
Humanities, ed. Susan Schreibman, Ray Siemens and John Unsworth
(Chichester: Wiley-Blackwell, 2016), pp. 420–438.
6. Michelle Springer, Beth Dulabahn, Phil Michel, Barbara Natanson, David
Reser, David Woodward and Helena Zinkham, For the Common Good:
The Library of Congress Flickr Pilot Project, report (30 October 2008),
p. iii. Available online at http://www.loc.gov/rr/print/flickr_report_
final.pdf.
7. Springer et al., For the Common Good, p. iii.
8. Ridge (ed.), Crowdsourcing our Cultural Heritage, p. 1.
9. Kathryn Eccles and Andrew Greg, ‘Your Paintings Tagger: Crowdsourcing
Descriptive Metadata for a National Virtual Collection’, Crowdsourcing
our Cultural Heritage, ed. Ridge, pp. 185–208, pp. 195–196.
10. John Berger, Ways of Seeing (London: Penguin and BBC, 1972).
11. Your Paintings followed its claim that ‘Tagging is easy!’ with the state-
ment ‘Anyone can do it’.
12. Mia Ridge, ‘From Tagging to Theorizing: Deepening Engagement with
Cultural Heritage through Crowdsourcing’, Curator: The Museum
Journal, 56:4 (2013): 435–450, p. 438.
13. Ridge, ‘From Tagging to Theorizing’, p. 441.
14. See, for example, Oomen et al., who argue that ‘The current research
on tagging shows that taggers that are explicitly invited to help an
institution by tagging, are notably more active.’ Johan Oomen,
Lotte Belice Baltussen, Sander Limonard, Maarten Brinkerink, Annelies
Van Ees, Lora Aroyo, Just Vervaart, Kamil Afsar and Riste Gligorov,
‘Emerging Practices in the Cultural Heritage Domain: Social Tagging
of Audiovisual Heritage’, Proceedings of the WebSci10: Extending the
Frontiers of Society On-Line (Raleigh, N.C., 2010). Available online at
http://journal.webscience.org/337/.
15. Trevor Owens, ‘Digital Cultural Heritage and the Crowd’, Curator: The
Museum Journal 56:1 (2013): 121–130, p. 122.
16. Owens, ‘Digital Cultural Heritage and the Crowd’, p. 128.
17. Johanna Drucker, Graphesis: Visual Forms of Knowledge Production
(Cambridge, MA: metaLABprojects and Harvard University Press,
2014), p. 177.
90  J. THOMAS

18. Springer et al., For the Common Good, p. 1.

19. Springer et al., For the Common Good, p. 15.
20. Eccles and Greg, ‘Your Paintings Tagger’, Crowdsourcing our Cultural
Heritage, ed. Ridge, pp. 196–200.
21. Lev Manovich, The Language of New Media (Cambridge, MA: MIT,
2001), pp. 69–73.
22. Aaron Marcus and Emilie W. Gould, ‘Cultural Dimensions and Global
Web User-Interface Design: What? So What? Now What?’, White Paper
(2000). Available online at http://www.amanda.com/cms/uploads/
media/AMA_CulturalDimensionsGlobalWebDesign.pdf.
23. Stuart Dunn and Mark Hedges, Crowd-Sourcing Scoping Study: Engaging
the Crowd with Humanities Research, AHRC, 12 November 2012,
p. 9. Available online at http://crowds.cerch.kcl.ac.uk/wp-content/
uploads/2012/12/Crowdsourcing-connected-communities.pdf.
24. Recent changes to Brooklyn Museum’s website have diluted the more
‘formal’ language of the museum. Visitors to the site are now greeted
with this mission statement: ‘Dedicated to the primacy of the visitor
experience, committed to excellence in every aspect of its collections and
programs, and drawing on both new and traditional tools of communi-
cation, interpretation, and presentation, the Museum aims to serve its
diverse public as a dynamic, innovative, and welcoming center for learn-
ing through the visual arts.’ https://www.brooklynmuseum.org/about.
25. Shelley Bernstein, ‘Clear Choices in Tagging’, Brooklyn Museum blog-
post, 22 July 2014, https://www.brooklynmuseum.org/community/
blogosphere/2014/07/22/clear-choices-in-tagging/.
26. Bernstein, ‘Clear Choices in Tagging’.
27. At the time of writing, this appeal to the public has been retained on the
homepage of Art UK, http://artuk.org/.
28. Jennifer Trant, David Bearman and Susan Chun, ‘The Eye of the
Beholder: Steve. Museum and Social Tagging of Museum Collections’,
proceedings of ICHIM, 24–26 October 2007, Toronto, Canada.
Available online at http://www.archimuse.com/ichim07/papers/trant/
trant.html.
29. Trant et al, ‘The Eye of the Beholder’. See also Jennifer Trant, ‘Tagging,
Folksonomy and Art Museums: Early Experiments and Ongoing
Research’, Journal of Digital Information 10:1 (2009). Available online
at https://journals.tdl.org/jodi/index.php/jodi/article/view/270.
30. See, for example, the description of the ESP game, which was articulated
in terms of ‘good’ or ‘robust’ tags. Ahn and Dabbish, ‘Labeling images
with a computer game’.
31. http://www.powerhousemuseum.com/collection/database/browsekey-
words.php. Philadelphia Museum of Art also presents its user-generated
 4 CROWDSOURCING  91

tags as a red word cloud clearly marked as ‘social tagging’: http://www.

philamuseum.org/collections/socialTagging.html.
32. http://www.powerhousemuseum.com/collection/database/browsekey-
words.php.
33. For a neat summary of the debate between expert ontologies and folkson-
omies, see Arjun Sabharwal, Digital Curation in the Digital Humanities:
Preserving and Promoting Archival and Special Collections (Waltham,
MA: Chandos, 2015), pp. 64–65.
34. Christina Manzo, Geoff Kaufman, Sukdith  Punjasthitkul and Mary
Flanagan, ‘“By the People, For the People”: Assessing the Value of
Crowdsourced, User-Generated Metadata’, DHQ: Digital Humanities
Quarterly 9:1 (2015): 1–16. Available online at http://digitalhumani-
ties.org:8081/dhq/vol/9/1/000204/000204.html.
35. Bernstein, ‘Clear Choices in Tagging’.
36. Bernstein, ‘Clear Choices in Tagging’.
37. Springer et al., For the Common Good, p. 18.
38. Springer et al., For the Common Good, p. 18.
39. Viktor Mayer-Schönberger and Kenneth Cukier, Big Data: A Revolution
That Will Transform How We Live, Work and Think (London: John
Murray, 2013), p. 43.
40. Mayer-Schönberger and Cukier, Big Data, p. 43.
41. Doireann Wallace, ‘Words as Keys to the Image Bank’, Revisualizing
Visual Culture, ed. Chris Bailey and Hazel Gardiner (Aldershot: Ashgate,
2010), pp. 83–96, p. 88.
42. See Eccles and Greg, ‘Your Paintings Tagger’, Crowdsourcing our
Cultural Heritage, ed. Ridge.
43. We have also supplemented our tags by exporting crowdsourced tags from
the British Library Flickr page: https://www.flickr.com/photos/britishli-
brary/.
44. https://wordnet.princeton.edu/wordnet/frequently-asked-questions/
for-linguists/.
45. Dunn and Hedges, Crowd-Sourcing Scoping Study, p. 20.
46. Sara Shatford Layne, ‘Subject Access to Art Images’, Introduction to Art
Image Access: Issues, Tools, Standards, Strategies, ed. Murtha Baca (Los
Angeles: Getty Research Institute, 2002), pp. 1–19, p. 1.
47. A pictorial capital featuring Falstaff appears in Charles Mackay, The
Thames and its Tributaries; or, Rambles Among the Rivers (London:
Richard Bentley, 1840), vol. 1, p. 26. Available on The Illustration
Archive, http://illustrationarchive.cardiff.ac.uk.
48. Recognising this issue, Lorraine Janzen Kooistra is creating a Database
of Ornament, which will develop a vocabulary for dealing with these
92  J. THOMAS

illustrations. Lorraine Janzen Kooistra, ‘The Politics of Ornament:

Remediation and/in The Evergreen’, ESC 41:1 (March 2015): 105–128,
p. 121.
49. ‘About the AAT’, http://www.getty.edu/research/tools/vocabularies/
aat/about.html.
50. An exception to this rule were ‘maps’ and ‘locations’, which we included
as ‘types’ even though these terms would probably be captured in ques-
tion 2 because it allowed us to link the tags to the GIS coordinates of the
places depicted. As a genre, illustration depicts more locations, inhabit-
ants, customs and cultures than any other mode of representation.
51. See, for example, the multiple artistic processes mentioned in this title:
The Riviera: Pen and Pencil Sketches From Cannes to Genoa … with
Twelve Chromo-Lithographic Illustrations and Numerous Woodcuts from
Drawings by the Author (London, 1870).
52. https://conscicom.org.
53. https://www.sciencegossip.org/#/about.
54. ‘When is an illustration an illustration?’, Help Board/The Objects
https://talk.sciencegossip.org/#/boards/BSC0000008/discussions/
DSC00000ex.
55. https://talk.sciencegossip.org/#/boards/BSC0000005/discussions/
DSC00000wh.
56. https://talk.sciencegossip.org/#/boards/BSC0000005/discussions/
DSC00000wh.
57. Walter Crane, Of the Decorative Illustration of Books Old and New (1896;
London: Bracken, 1984), p. 157.
58. James Mussell, The Nineteenth-Century Press in the Digital Age
(Basingstoke: Palgrave Macmillan, 2012), pp. 69–113.
59. Kooistra, ‘The Politics of Ornament’, p. 108. For another example of the
importance of ‘ornamental’ illustrations in the nineteenth-century peri-
odical press, see Alison Chapman’s discussion of the illustrated poetry of
M.C. Gillington, which appeared in Woman’s World. Alison Chapman,
‘Virtual Victorian Poetry’, Virtual Victorians: Networks, Connections,
Technologies, ed. Veronica Alfano and Andrew Stauffer (New York and
Basingstoke: Palgrave Macmillan, 2015), pp. 145–166, pp. 154–157.
60. ‘Zooniverse releases Science Gossip’, Biodiversity Heritage Library blog-
post, 4 March 2015, http://blog.biodiversitylibrary.org/2015/03/
zooniverse-releases-science-gossip.html.
61. Lorraine Janzen Kooistra, Poetry, Pictures, and Popular Publishing: The
Illustrated Gift Book and Victorian Visual Culture 1855–1875 (Athens,
OH: Ohio University Press, 2011), p. 243.
62. Simon Nowell-Smith, The House of Cassell 1848–1958 (London: Cassell,
1958), p. 37.
 4 CROWDSOURCING  93

63. Nowell-Smith, The House of Cassell, p. 106.

64. Nowell-Smith, The House of Cassell, pp. 82–83.
65. Nowell-Smith, The House of Cassell, p. 105.
66. Nowell-Smith, The House of Cassell, p. 106.
67. W.H.G. Kingston qtd. in Nowell-Smith, The House of Cassell, p. 106.
68. Kingston qtd. in Nowell-Smith, The House of Cassell, p. 106.
69. Nowell-Smith, The House of Cassell, pp. 105–106.
70. One commentator criticised illustrated annuals because ‘Many of the
annuals were little more than showy pictures, with trashy, insipid stories
to explain them, thus reversing the proper order, in which the painter
should follow the writer, and not precede him.’ ‘Book-Prints’, Chambers’s
Journal of Popular Literature, Science and Arts, 30 August 1862, pp.
135–137, p. 136.
71. In some cases, however, musical scores do function as illustrations. See,
for example, the music reproduced in George Moore’s In Single Strictness
(London: William Heinemann, 1922), p. 27. I am grateful to Ann
Heilmann for this reference. Likewise, there are problems with the types
of images that we have identified as illustrations. Patricia Seed has made
the case that maps are not illustrations. Patricia Seed, ‘A Map is Not a
Picture: How the Digital World Threatens the Validity of Printed Maps’,
Between Humanities and the Digital, ed. Patrik Svensson and David Theo
Goldberg (Cambridge, MA and London: MIT Press, 2015), pp. 401–409.
 CHAPTER 5

Tailpiece

Abstract  The conclusion outlines the conceptual paradigms opened up

by digital modes of visibility and searchability. Thomas suggests that a
digital environment contravenes analogue limitations by enabling mul-
tiple illustrations from different pages or publications to be displayed
alongside each other. This results in a radically new way of viewing illus-
tration that is defined here as ‘affillustration’, a term that captures the
interplay or ‘kinship’ between illustrations, their allusions and references
to other illustrations, and their place in a distinct iconographic tradition.

Keywords  Nineteenth century · Illustration · Digital · Intertextuality

This tailpiece is not the end of the tale. So far, this book has focused
on the issues of making illustrations visible and searchable online, iden-
tifying the development of the digital resource as a process that engages
with notions of what an ‘illustration’ is and how it makes its meanings. I
want to shift the emphasis in these concluding pages to outline the con-
ceptual paradigms opened up when these new modes of visibility and
searchability are operational.
Digital archives offer a new accessibility for historic illustrations, mak-
ing available images housed in disparate publications and collections and
allowing them to be retrieved and viewed in diverse ways. Such features
are of benefit to researchers well beyond those working in the immedi-
ate field of illustration studies. Indeed, the identification of the ‘academic

© The Author(s) 2017 95

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4_5
96  J. THOMAS

beneficiaries’ of The Illustration Archive (an all-too familiar phrase for

grant writers in the UK) is difficult because the size and breadth of the
dataset makes it hard to imagine a field of humanities in which these
images would not be useful. There is also the potential for digital illustra-
tion resources to help raise public awareness of a genre that is a largely
forgotten yet highly significant part of cultural heritage. Having created
DMVI primarily with academic researchers in mind, we were gratified by
the demand for these images from very different quarters. Illustrations
from the database have been used by graphic designers, heritage and
community groups, on websites, in newspapers, on book covers, in tel-
evision programmes and films. These different constituencies have been
drawn to DMVI primarily because the database enables illustrations to be
searched in multiple ways. It is important to recognise the significance of
this searchability because it indicates how an encompassing digital frame-
work can extend the ‘life’ of illustrations and allow them to circulate
across centuries and sectors, while delimitations in how an illustration is
understood (by excluding ‘ornaments’ from its definition, for example)
constrain the apparent relevance of the genre.
Although digital illustration archives make thousands of historic illus-
trations available, however, they are fundamentally distinct from image
banks. The driving-force behind these archives is the research questions
about illustration that they articulate and answer. Diverse modes of ana-
lysing illustrations are made possible in the digital mechanisms that ren-
der illustrations visible, enable their content to be searched and retrieved,
and allow them to be displayed and grouped together. Indeed, the ability
to look at images alongside each other is crucial for an understanding of
illustration, particularly as this process is severely hampered in analogue
formats. The turning of a page in a book, magazine or newspaper hides
the illustrations on that page, and, even when several texts are opened
simultaneously, a viewing of the illustrations is curtailed by the size of
the surface on which they are placed (and I do not recommend laying
illustrated books side by side on the floor of the British Library reading
rooms).
A digital viewing in which multiple illustrations can be seen together
entails a ‘looking across’ or ‘scanning’ of images that stands as the visual
equivalent of what Andrew Piper has termed ‘roaming’.1 According to
Piper, digital texts transgress the horizontal limitations of the page, cre-
ating a model where ‘the textual surface is roamable rather than turn-
able’ and where the ‘adjacency’ of text becomes paramount.2 Likewise,
 5 TAILPIECE  97

the digital illustration archive creates an environment that contravenes

the analogue limitations of the genre and allows illustrations from dif-
ferent pages or publications to be displayed alongside each other. Such
a ‘scanning’ of multiple illustrations brings with it a theoretical engage-
ment that reveals the affiliative interplay between illustrations, their
allusions and references to other illustrations, or what might be called
affillustration.
This book has sought to conceptualise what a practice of affillustration
might involve. It is a practice, indeed, that has applicability beyond the
nineteenth century, with the focus of this book providing a lens through
which illustrations from other times and cultures might be viewed in dig-
ital contexts. But while the reach of images embraced by the notion of
affillustration might be wide, there is another sense in which affillustra-
tion stresses particularity, situating illustration as part of a distinct tra-
dition and iconography. As Stephen C. Behrendt remarks, ‘we need to
consider more carefully the tendency of the literary illustrator to bring
to the text an iconographic tradition that often derives not from the
work being illustrated but rather from other sources entirely, sources that
sometimes are quite at variance with the text’.3 Behrendt recognises that
illustrations can refer to other illustrations as part of the ‘iconographic
tradition’ within which the illustrator works. But as a form of meaning
production, affillustration is not always conscious or deliberate. Just as
every text is a ‘tissue of quotations’, so every illustration is constituted
by its references to other illustrations.4 Affillustration, then, resembles
its ‘intertextual’ counterpart, but it is distinct from intertextuality in its
focus on the specificity of illustration and the kinds of associations it puts
into play. The ‘kinship’ between illustrations that is implied in the term
‘affillustration’ draws attention to the fact that illustration is an eminently
social genre. It makes meanings not just in its (conscious and uncon-
scious) references to other illustrations, but also in the groupings and
clusters that it generates, the ‘networks’ that exist within and across the
boundaries of the illustrated text.
In its emphasis on these networks, affillustration is both a prod-
uct and signifier of the centrality of illustration in nineteenth-century
culture and offers a way of analysing the genre that takes account of,
and is predicated on, this position. The affiliative relationship between
illustrations is apparent across a range of nineteenth-century publica-
tions: in fiction, where illustrations repeat iconographic details of set-
tings and figures, and in illustrated newspapers and periodicals, where
98  J. THOMAS

the illustrations create a recognisable and consistent ‘identity’ for the

publication.5 The illustrations here signify in their relation to each other,
relying on the viewer’s growing familiarity with shared pictorial styles
and features. Affillustration does not always follow the lead of the text,
though. It can stand independently, creating its own narrative and, in
turn, influencing the meanings of the words. David Skilton has drawn
attention to the similarities between an illustration that Millais designed
for Trollope’s Framley Parsonage and one by John Leech that appeared
in the Illuminated Magazine 17 years earlier, an association between
images that adds to their meanings and the meanings of the texts they
accompany.6
Likewise, illustrations can be placed in new affiliative relation-
ships. The concept of affillustration incorporates the practice of ‘extra-
illustration’ in which readers exploit the signifying interaction between
illustrations by using the text as a starting point to make associations
between illustrations and words, and between illustrations and other
illustrations.7 Extra-illustration came to prominence after the publica-
tion of James Granger’s Biographical History of England (1769), which
gave the activity the name ‘grangerizing’. Granger’s book was not illus-
trated in its original form, but readers ‘illustrated’ their copies, remov-
ing pictures from different sources and adding them to Granger’s text.
This activity remained popular throughout the nineteenth century
when Shakespeare’s plays, some of the most common extra-illustrated
texts, were illustrated with pictures from different editions, photographs
of actors and images of landscapes or flowers mentioned in the words.
The forging of affiliative links between illustrations was also a principle
underlying the early collection of illustrations, with collectors striving to
organise and view these images in multiple ways. One method for achiev-
ing this was to remove illustrations from their textual confines altogether.
Gleeson White advocates extracting illustrations from periodicals and
mounting them on brown paper so that the images can be distributed
among boxes with different subdivisions and ‘examined easily’.8 Forrest
Reid also points to the benefits for the collector of ‘clipping’ illustrations
from their sources in order to ‘classify’ them.9
The collection principles put into practice by White and Reid advocate
new groupings of illustrations and the ability to look across them in a
way that is impossible when they are enclosed in the pages of a periodi-
cal, or, for that matter, a book. Their methods are inherently bound up
in the materiality of nineteenth-century illustration; but the environment
 5 TAILPIECE  99

they describe, where images can be moved around from box to box and
viewed in the context of other images, anticipates the modes of display
enabled by the digital, which emphasises and exposes affillustration. In
The Illustration Archive, the illustrations can be seen in a ‘list’ or ‘gal-
lery’ format (with or without metadata), viewed alongside illustrations
from the same book, viewed with ‘similar’ images, or in the context of
user-generated exhibitions. In DMVI, we added a ‘lightbox’ function
when we enhanced the site so that users can make their own collections
of illustrations. Digital resources can even emulate the practice of extra-
illustration. The Victorian Illustrated Shakespeare Archive brings together
illustrations from four nineteenth-century editions of Shakespeare and, at
the click of a button, reconfigures them in new ways, generating numer-
ous extra-illustrated versions.
Recognition of affillustration is not simply the result of being able
to view the images together, however. The research potential of digital
illustration resources lies in a combination of visibility and searchability,
of making illustrations available and allowing them to be interrogated
across bibliographic and iconographic fields. These features vary from
archive to archive, but, as a general rule, a bibliographic search mines
the metadata associated with the illustration, allowing users to find illus-
trations by publication title, author, publisher, date, illustrator and (in
the case of DMVI) engraver. This bibliographic search option offers up
illustrations to different modes of analysis. The diverse dataset in The
Illustration Archive, for example, makes it possible to analyse illustra-
tions by date range or decade, to look at illustrations of specific works,
or to gauge what texts were illustrated when. Such bibliographic searches
of illustrations lend themselves to multiple critical approaches, from
a ‘close’ viewing that analyses a single or small group of images to a
‘distant’ viewing, or what Matthew L. Jockers calls ‘a bird’s-eye view’,
which looks for overarching patterns.10
Whereas the bibliographic metadata consists of, and supplements,
information found in library catalogues, the iconographic details of illus-
trations are not so easily identifiable from catalogues, metadata or shelves
full of illustrated books. The ability to search the content of illustra-
tions and to identify the interrelations between images in terms of what
they depict is a direct result of digital intervention, having emerged
from the methodologies of digital illustration resources. In the absence
of a sophisticated method of Content-Based Image Retrieval, the icon-
ographic markup of illustrations is generated through the intermediary
100  J. THOMAS

of language and is inevitably inflected by our own historical moment.

Nevertheless, it is this search mechanism, borne out of the digital, that
signals a new direction for illustration studies by establishing a mode of
analysing illustration that identifies affillustration as part of and constitu-
tive of cultural values. It is the articulation of this new theoretical para-
digm for illustration studies that has been the main objective of this book
and its underpinning assertion that digital methods for making illustra-
tions visible and searchable are critical interventions.
There are, inevitably, limitations to the research potential of digi-
tal resources, even with the bibliographic and iconographic markup in
place. Digital illustration archives are discreet entities with defining
principles that, by necessity, demarcate the illustrations they contain
and how they can be analysed: the William Blake Archive is delimited
by author, as is the Rossetti Archive and Illustrating Scott; the Victorian
Illustrated Shakespeare Archive is delimited by author and editions; The
Yellow Nineties Online by publication and date; DMVI by year, technique
(wood engraving) and textual genre (‘literary’). Such organising princi-
ples form the scaffolding of the resource and determine how the illustra-
tions can be interrogated (for example, although DMVI can be searched
for illustrations of women looking out of windows published in 1862,
these illustrations cannot be compared, at least within the parameters
of this database, to illustrations with similar iconographic features that
appeared in 1892). A solution to this problem lies in the potential for
cross searchability between archives that allows for an analysis of the rela-
tion between illustrations in different datasets and between illustrations
and other cultural forms. The possibilities of this cross searchability have
been established in resources like NINES (http://www.nines.org) and
Connected Histories (http://www.connectedhistories.org), which bring
together a range of textual and visual digital resources with a single fed-
erated search.
An infinite archive of searchable illustrations, however, is still more
of an ideal than a reality, and in the absence of extensive cross search-
ability, new pathways for examining illustrations have emerged in The
Illustration Archive, which abandons the conventional foundational prin-
ciples of digital archives. Here, there is no clearly defined set of authors
or works; rather, the images are taken from out of copyright books cov-
ering history, geography, philosophy and literature that were housed in
 5 TAILPIECE  101

the basement storage of the British Library and collected between 1850
and 1975. In terms of digitisation, the only selection criteria concerned
the size of the books, which were between 18.5 and 30.5 cm in height
and no thicker than 10 cm because of the capabilities of the scanners.
The sheer variety of this dataset means that the material is characterised
more by its diversity than uniformity and it is the very randomness or
‘unpredictability’ of the illustrations that has proved most fruitful. While
the 68,000 volumes limit the maximum number of authors to 68,000
(give or take multi-authored volumes), in any practical sense, the names
of the authors are ‘unpredictable’. This is also the case with the date
range: the collection is known as ‘The Nineteenth-Century Printed
Books Dataset’, but there are a significant number of eighteenth-century
books and we have identified books published as early as 1511 and as late
as 1946. It goes without saying that the iconographic features of these
illustrations are unquantifiable and unknown. As I have suggested in pre-
vious chapters, the unpredictability of the illustrations in The Illustration
Archive presents numerous challenges in terms of the development of the
archive, but there is a sense in which it also liberates the research poten-
tial of the digital resource, making it possible for the first time to identify
and explore the interplay between illustrations in ways that might com-
plement but are not limited to bibliographic imperatives.
With nineteenth-century illustrations now in the sight, if not in the
hands, of the user, it is time to draw a line under this tailpiece. This book
has focused on nineteenth-century material and archives, but its aim has
been to develop conceptual models for digital illustration studies that
might inform future practice more broadly, both in terms of how we
conceive of illustration and the relation between word and image, and
in how digital resources might be constructed to take account of and
expose the multiplicity and plurality of illustration. As I have argued,
the interaction between the digital and illustration is a dynamic one in
which the meanings of illustration are negotiated and played out. A fun-
damental aspect of these meanings is the visual dialogue between illus-
trations that I have called affillustration, which situates these images as
part of a network that shapes and is shaped by culture. I want to con-
clude by making the case that digital archives are uniquely positioned to
reveal these affiliative relations and to enable new and nuanced analyses
of how illustrations signify. The potential for this analysis of illustration,
102  J. THOMAS

moreover, lies as much in the development of the digital archive and the
modes of visibility and searchability that it institutes as in the tools and
features made available in the ‘live’ resource. The digital, then, is never
just a means of bringing illustration to light. Illustration might be allu-
sive, as the affiliative relations between these images suggest, but it is also
elusive, an ‘immense continent’ that is still largely unexplored. A digital
environment illuminates the way—it allows illustrations to be seen and
traversed; but it also gestures to hidden territories—to those areas that
continue to keep us in the dark (Fig. 5.1).

Fig. 5.1  ‘The End’, tailpiece from Edward Lamplough, Yorkshire Battles (Hull:
William Andrews and Co., 1891). Available on The Illustration Archive: http://
illustrationarchive.cardiff.ac.uk
 5 TAILPIECE  103

Notes
1. Andrew Piper, Book Was There: Reading in Electronic Times (Chicago and
London: University of Chicago Press, 2012), pp. 55–56.
2. Piper, Book Was There, pp. 55, 56.
3. Stephen C. Behrendt, ‘The Functions of Illustration—Intentional
and Unintentional’, Imagination on a Long Rein. English Literature
Illustrated, ed. Joachim Möller (Marburg: Jonas Verlag, 1988),
pp. 29–44, p. 29.
4. Roland Barthes, ‘The Death of the Author’, Image Music Text, ed. and
trans. Stephen Heath (London: Fontana, 1977), pp. 142–148, p. 146.
5. See James Mussell, The Nineteenth-Century Press in the Digital Age
(Basingstoke: Palgrave Macmillan, 2012), pp. 73–74.
6. David Skilton, ‘The Centrality of Literary Illustration in Victorian Visual
Culture: The Example of Millais and Trollope from 1860 to 1864’,
Journal of Illustration Studies, 1 (December, 2007). Available online at
http://jois.uia.no/articles.php?article=30.
7. For a history of extra-illustration, see Erin C. Blake and Stuart Sillars,
Extending the Book: The Art of Extra-Illustration (Washington D.C.:
Folger Shakespeare Library, 2010).
8. Gleeson White, English Illustration, ‘The Sixties’: 1855–1870 (1897; Bath:
Kingsmead Reprints, 1970), p. 7.
9. Forrest Reid, Illustrators of the Eighteen Sixties (1928; Toronto: Dover,
1975), p. 11.
10. Jockers is referring to a way of analyzing literature that he calls ‘macro-
analysis’ and that draws on Franco Moretti’s concept of ‘distant’ read-
ing. The ‘scanning’ or ‘looking across’ of illustration that I propose,
although embedded in viewing rather than reading, is equally indebted
to the possibilities opened up by large-scale datasets. Matthew L. Jockers,
Macroanalysis: Digital Methods and Literary History (Urbana: University
of Illinois Press, 2013), p. 19. See also Franco Moretti, Distant Reading
(London and New York: Verso, 2013).
 Bibliography

Ahn, Luis von and Laura Dabbish, ‘Labeling images with a computer game’,
Proceedings of the SIGCHI conference on Human Factors in Computing Systems
(Vienna, Austria: ACM, 2004) pp. 319–326. Available online at http://ael.
gatech.edu/cs6452f13/files/2013/08/labeling-images.pdf.
Allington, Daniel, Sarah Brouillette and David Golumbia, ‘Neoliberal Tools (and
Archives): A Political History of Digital Humanities’, Los Angeles Review of
Books, 1 May 2016.
Baca, Murtha (ed.), Introduction to Art Image Access: Issues, Tools, Standards,
Strategies (Los Angeles: Getty Research Institute, 2002).
Barthes, Roland, ‘The Death of the Author’, Image Music Text, ed. and trans.
Stephen Heath (London: Fontana, 1977) pp. 142–148.
Barthes, Roland, ‘Rhetoric of the Image’, Image, Music, Text, ed. and trans.
Stephen Heath (London: Fontana, 1977) pp. 32–51.
Beegan, Gerry, The Mass Image: A Social History of Photomechanical
Reproduction in Victorian London (Basingstoke: Palgrave Macmillan, 2008).
Behrendt, Stephen C., ‘The Functions of Illustration—Intentional and
Unintentional’, Imagination on a Long Rein. English Literature Illustrated,
ed. Joachim Möller (Marburg: Jonas Verlag, 1988) pp. 29–44.
Bell, Peter, Joseph Schlecht and Björn Ommer, ‘Nonverbal Communication in
Medieval Illustrations Revisited by Computer Vision and Art History’, Visual
Resources 29:1–2 (2013): 26–37.
Berger, John, Ways of Seeing (London: Penguin and BBC, 1972).
Bernstein, Shelley, ‘Clear Choices in Tagging’, Brooklyn Museum blogpost,
22 July 2014, https://www.brooklynmuseum.org/community/blogo-
sphere/2014/07/22/clear-choices-in-tagging/.

© The Editor(s) (if applicable) and The Author(s) 2017 105

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4
106  Bibliography

Blake, Erin C. and Stuart Sillars, Extending the Book: The Art of Extra-Illustration
(Washington D.C.: Folger Shakespeare Library, 2010).
Bloomfield, Robert, ‘The Farmer’s Boy’, Favourite English Poems (London:
Sampson Low, 1862) pp. 116–193.
Bolter, J. David, and Richard Grusin, Remediation: Understanding New Media
(Cambridge, MA: MIT Press, 1999).
Bolter, J. David, Writing Space: The Computer, Hypertext, and the History of
Writing (New Jersey and London: Lawrence Erlbaum Associates, 1991).
‘Book-Prints’, Chambers’s Journal of Popular Literature, Science and Arts, 30
August 1862, pp. 135–137.
Bradley, John and Stephen Pigney, ‘Towards a Bibliographic Model of
Illustrations in the Early Modern Illustrated Book’, paper delivered at the
Digital Humanities conference, University of Hamburg, 2012. Available
online at https://lecture2go.uni-hamburg.de/konferenzen/-/k/13923.
Burdick, Anne, Johanna Drucker, Peter Lunenfeld, Todd Presner, Jeffrey
Schnapp, Digital_Humanities (Cambridge, MA: MIT Press, 2012).
Carletti, Laura, Derek McAuley, Dominic Price, Gabriella Giannachi and Steve
Benford, ‘Digital Humanities and Crowdsourcing: An Exploration’, paper
delivered at MW 2013: Museums and the Web conference, 2013. Available
online at http://mw2013.museumsandtheweb.com/paper/digital-humanities-
and-crowdsourcing-an-exploration-4/.
Causer, Tim and Melissa Terras, ‘“Many Hands Make Light Work. Many Hands
Together Make Merry Work”: Transcribe Bentham and Crowdsourcing
Manuscript Collections’, Crowdsourcing our Cultural Heritage, ed. Mia Ridge
(Aldershot: Ashgate, 2014) pp. 57–88.
Causer, Tim and Valerie Wallace, ‘Building A Volunteer Community: Results
and Findings from Transcribe Bentham’, DHQ: Digital Humanities Quarterly
6:2 (2012). Available online at http://digitalhumanities.org:8081/dhq/
vol/6/2/000125/000125.html.
Chapman, Alison, ‘Virtual Victorian Poetry’, Virtual Victorians: Networks,
Connections, Technologies, ed. Veronica Alfano and Andrew Stauffer (New
York and Basingstoke: Palgrave Macmillan, 2015) pp. 145–166.
Chung, J. S., R. Arandjelović, G. Bergel, A. Franklin and A. Zisserman,
‘Re-presentations of Art Collections’, paper presented for the Workshop on
Computer Vision for Art Analysis, ECCV, 2014. Available online at http://
www.robots.ox.ac.uk/~vgg/publications/2014/Chung14/.
Cooke, Simon, ‘A Bitter After-Taste: The Illustrated Gift Book of the 1860s’,
Reading Victorian Illustration, 1855–1875: Spoils of the Lumber Room, ed.
Paul Goldman and Simon Cooke (Aldershot: Ashgate, 2012) pp. 53–78.
Crane, Walter, Of the Decorative Illustration of Books Old and New (1896;
London: Bracken, 1984).
 Bibliography   107

Darnton, Robert, ‘What is the history of books?’, Daedalus 111:3 (1982):

65–83.
Datta, Ritendra, Dhiraj Joshi, Jia Li, and James Z. Wang, ‘Image Retrieval: Ideas,
Influences, and Trends of the New Age’, ACM Computing Surveys 40:2 (April
2008).
Drucker, Johanna, Graphesis: Visual Forms of Knowledge Production (Cambridge,
MA: metaLABprojects and Harvard University Press, 2014).
Dunn, Stuart and Mark Hedges, Crowd-Sourcing Scoping Study: Engaging the
Crowd with Humanities Research, AHRC, 12 November 2012. Available
online at http://crowds.cerch.kcl.ac.uk/wp-content/uploads/2012/12/
Crowdsourcing-connected-communities.pdf.
Eaves, Morris, ‘Picture Problems: X-Editing Images 1992–2010’, DHQ: Digital
Humanities Quarterly 3:3 (2009). Available online at http://digitalhumani-
ties.org:8081/dhq/vol/3/3/000052/000052.html.
Eccles, Kathryn and Andrew Greg, ‘Your Paintings Tagger: Crowdsourcing
Descriptive Metadata for a National Virtual Collection’, Crowdsourcing our
Cultural Heritage, ed. Mia Ridge (Aldershot: Ashgate, 2014) pp. 185–208.
Ernst, Wolfgang, Digital Memory and the Archive (Minneapolis: University of
Minnesota Press, 2013).
‘First Impressions’, Indianapolis Museum of Art blogpost, 4 May 2011, http://
www.imamuseum.org/blog/2011/05/04/first-impressions/.
Flaubert, Gustave, Flaubert Correspondance Vol. III (janvier 1859–décembre
1868), ed. Jean Bruneau (Paris: Gallimard, 1991).
Galloway, Alexander R., The Interface Effect (Cambridge: Polity, 2012).
Genette, Gérard, Paratexts: Thresholds of Interpretation, trans. Jane E. Lewin
(1987; Cambridge: Cambridge University Press, 1997).
Goldman, Paul, ‘Defining Illustration Studies’, Reading Victorian Illustration,
1855–1875: Spoils of the Lumber Room, ed. Paul Goldman and Simon Cooke
(Aldershot: Ashgate, 2012) pp. 13–32.
[Gore, Catherine], ‘The New Art of Printing. By a Designing Devil’, Blackwood’s
Edinburgh Magazine 55:339 (January 1844): 45–49.
Graham, Margaret E., ‘Enhancing Visual Resources for Searching and
Retrieval—Is Content-Based Image Retrieval a Solution?’, Literary and
Linguistic Computing 19:3 (2004): 321–333.
Harpring, Patricia, ‘The Language of Images: Enhancing Access to Images by
Applying Metadata Schemas and Structured Vocabularies’, Introduction to
Art Image Access: Issues, Tools, Standards, Strategies, ed. Murtha Baca (Los
Angeles: Getty Research Institute, 2002) pp. 20–39.
Harvey, Ian and Nicky Lloyd, ‘Lost Visions: A Descriptive Metadata
Crowdsourcing and Search Platform for Nineteenth-Century Book
Illustrations’, Proceedings of the Digital Humanities Congress 2014, Studies in
108  Bibliography

the Digital Humanities (Sheffield: HRI Online Publications, 2014). Available

online at http://www.hrionline.ac.uk/openbook/chapter/dhc2014-harvey.
Hayles, N. Katherine, My Mother Was a Computer: Digital Subjects and Literary
Texts (Chicago and London: University of Chicago Press, 2005).
Hocks, Mary E. and Michelle R. Kendrick (eds.), Eloquent Images: Word and
Image in the Age of New Media (Cambridge, MA: MIT Press, 2003).
Holterhoff, Kate, ‘Ethics and the Digital Archive: The Case for Visualizing
H. Rider Haggard’, Journal of Visual Culture Online, 21 October 2013.
Available online at http://blogs.tandf.co.uk/jvc/2013/10/21/ethics-and-
the-digital-archive-the-case-for-visualizing-h-rider-haggard/.
Ionescu, Christina (ed.), Book Illustration in the Long Eighteenth Century:
Reconfiguring the Visual Periphery of the Text (Cambridge: Cambridge
Scholars, 2011).
Jackson, John, A Treatise on Wood Engraving, Historical and Practical (London:
Charles Knight, 1839).
James, Henry, The Golden Bowl, ed. Virginia Llewellyn Smith (1909; Oxford and
New York: Oxford University Press, 1963).
Jessop, Martyn, ‘Digital visualization as a scholarly activity’, Literary and
Linguistic Computing 23:3 (2008): 281–293.
Jockers, Matthew L., Macroanalysis: Digital Methods and Literary History
(Urbana: University of Illinois Press, 2013).
Jung, Sandro (ed.), British Literature and Print Culture (Cambridge: D. S. Brewer,
2013).
Kim, Sung-Do, ‘La Raison Graphique de Saussure’, Cahiers Ferdinand de
Saussure 61 (2008): 23–42.
Kirschenbaum, Matthew G., ‘The Word as Image in an Age of Digital
Reproduction’, Eloquent Images: Word and Image in the Age of New Media,
ed. Mary E. Hocks and Michelle R. Kendrick (Cambridge, MA: MIT Press,
2003) pp. 137–156.
Kittler, Friedrich, Optical Media: Berlin Lectures 1999, trans. Anthony Enns
(Cambridge: Polity, 2010).
Kooistra, Lorraine Janzen, The Artist as Critic:Bitextuality in Fin-de-Siècle
Illustrated Books (Aldershot: Scolar, 1995).
Kooistra, Lorraine Janzen, Poetry, Pictures, and Popular Publishing: The
Illustrated Gift Book and Victorian Visual Culture 1855–1875 (Athens, OH:
Ohio University Press, 2011).
Kooistra, Lorraine Janzen, ‘The Politics of Ornament: Remediation and/in The
Evergreen’, ESC 41:1 (March 2015): 105–128.
Layne, Sara Shatford, ‘Subject Access to Art Images’, Introduction to Art Image
Access: Issues, Tools, Standards, Strategies, ed. Murtha Baca (Los Angeles:
Getty Research Institute, 2002) pp. 1–19.
 Bibliography   109

Lessing, Gotthold Ephraim, Laocoőn: An Essay on the Limits of Painting and

Poetry, trans. Edward Allen McCormick (1766; Baltimore and London: Johns
Hopkins University Press, 1984).
Maidment, Brian, ‘The Illuminated Magazine and the Triumph of Wood
Engraving’, The Lure of Illustration in the Nineteenth Century: Picture
and Press, ed. Laurel Brake and Marysa Demoor (Basingstoke: Palgrave
Macmillan, 2009) pp. 17–39.
Mandal, Anthony, ‘Digital Humanities’, The Year’s Work in Critical and
Cultural Theory, ed. Neil Badmington and David Tucker, vol. 24 (Oxford:
Oxford University Press, 2016) pp. 317–342.
Manovich, Lev, ‘How to Compare One Million Images?’, Understanding Digital
Humanities, ed. David M. Berry (Basingstoke: Palgrave Macmillan, 2012)
pp. 249–278.
Manovich, Lev, The Language of New Media (Cambridge, MA: MIT Press,
2001).
Manzo, Christina, Geoff Kaufman, Sukdith Punjasthitkul and Mary Flanagan,
‘“By the People, For the People”: Assessing the Value of Crowdsourced, User-
Generated Metadata’, DHQ: Digital Humanities Quarterly 9:1 (2015). Available
online at http://digitalhumanities.org:8081/dhq/vol/9/1/000204/000204.
html.
Marcus, Aaron and Emilie W. Gould, ‘Cultural Dimensions and Global Web
User-Interface Design: What? So What? Now What?’, White Paper (2000).
Available online at http://www.amanda.com/cms/uploads/media/AMA_
CulturalDimensionsGlobalWebDesign.pdf.
Maxwell, Richard (ed.), The Victorian Illustrated Book (Virginia: University of
Virginia Press, 2002).
Mayer-Schönberger, Viktor and Kenneth Cukier, Big Data: A Revolution that
Will Transform How We Live, Work and Think (London: John Murray, 2013).
McGann, Jerome, Radiant Textuality: Literature after the World Wide Web (New
York and Basingstoke: Palgrave, 2001).
McPherson, Tara, ‘Post-Archive: The Humanities, the Archive, and the
Database’, Between Humanities and the Digital, ed. Patrik Svensson and
David Theo Goldberg (Cambridge, MA and London: MIT Press, 2015)
pp. 483–502.
Mendelsund, Peter, ‘What We See When We Read’, The Paris Review, 14 August
2014.
Mendelsund, Peter, What We See When We Read: A Phenomenology, with
Illustrations (New York: Vintage, 2014).
Meyrick, Robert, ‘“Spoils of the lumber-room”: Early collectors of wood-
engraved illustrations from 1860s periodicals’, Reading Victorian Illustration,
1855–1875: Spoils of the Lumber Room, ed. Paul Goldman and Simon Cooke
(Aldershot: Ashgate, 2012) pp. 179–199.
110  Bibliography

Miller, J. Hillis, Illustration (London: Reaktion, 1992).

Mitchell, W. J. T., Iconology: Image, Text, Ideology (Chicago: University of
Chicago Press, 1986).
Mitchell, W. J. T., ‘There Are No Visual Media’, Journal of Visual Culture 4:2
(August 2005): 257–266.
Moretti, Franco, Distant Reading (London and New York: Verso, 2013).
Mussell, James, The Nineteenth-Century Press in the Digital Age (Basingstoke:
Palgrave Macmillan, 2012).
Nowell-Smith, Simon, The House of Cassell 1848–1958 (London: Cassell, 1958).
Nowviskie, Bethany, ‘Resistance in the Materials’, Between Humanities and the
Digital, ed. Patrik Svensson and David Theo Goldberg (Cambridge, MA and
London: MIT Press, 2015) pp. 383–389.
Nurse Rockbaby’s Easy Reading, and Pretty Pictures (London: Dean and Son,
1853).
Oomen, Johan, Lotte Belice Baltussen, Sander Limonard, Maarten Brinkerink,
Annelies Van Ees, Lora Aroyo, Just Vervaart, Kamil Afsar and Riste Gligorov,
‘Emerging Practices in the Cultural Heritage Domain: Social Tagging of
Audiovisual Heritage’, Proceedings of the WebSci10: Extending the Frontiers
of Society On-Line (Raleigh, N.C., 2010). Available online at http://journal.
webscience.org/337/.
Owens, Trevor, ‘Digital Cultural Heritage and the Crowd’, Curator: The
Museum Journal 56:1 (2013): 121–130.
Piper, Andrew, Book Was There: Reading in Electronic Times (Chicago and
London: University of Chicago Press, 2012).
Posthumus, Etienne, ‘DMVI Iconclass Mapping’, Cardiff Book History blogpost,
22 February 2011, https://cardiffbookhistory.wordpress.com/tag/arkyves-
project/.
Price, Leah, How to Do Things with Books in Victorian Britain (Princeton and
Oxford: Princeton University Press, 2012).
Rajewsky, Irina O., ‘Intermediality, Intertextuality, and Remediation: A Literary
Perspective on Intermediality’, Intermédialités: histoire etthéorie des arts,
des lettres et des techniques / Intermediality: History and Theory of the Arts,
Literature and Technologies 6 (2005): 43–64.
Reid, Forrest, Illustrators of the Eighteen Sixties (1928; Toronto: Dover, 1975).
Resig, John, ‘Using Computer Vision to Increase the Research Potential of
Photo Archives’, Journal of Digital Humanities 3:2 (2014).
Ridge, Mia (ed.), Crowdsourcing our Cultural Heritage (Aldershot: Ashgate,
2014).
Ridge, Mia (ed.), ‘From Tagging to Theorizing: Deepening Engagement with
Cultural Heritage through Crowdsourcing’, Curator: The Museum Journal,
56:4 (2013): 435–450.
 Bibliography   111

Ruskin, John, The Cestus of Alglaia, The Complete Works of John Ruskin, ed.
E. T. Cook and Alexander Wedderburn (London: George Allen, 1903–1912)
vol. XIX, pp. 139–140.
Sabharwal, Arjun, Digital Curation in the Digital Humanities: Preserving and
Promoting Archival and Special Collections (Waltham, MA: Chandos, 2015).
Saussure, Ferdinand de, Course in General Linguistics, ed. Charles Bally and
Albert Sechehaye, trans. Wade Baskin (New York: McGraw-Hill, 1959).
Seed, Patricia, ‘A Map is Not a Picture: How the Digital World Threatens the
Validity of Printed Maps’, Between Humanities and the Digital, ed. Patrik
Svensson and David Theo Goldberg (Cambridge, MA and London: MIT
Press, 2015) pp. 401–409.
Shillingsburg, Peter L., ‘Practical Editions of Literary Texts’,Variants 4: The
Book as Artefact: Text and Border, ed. Anne MetteHansen, Roger Lüdeke,
Wolfgang Streit, Cristina Urchuaguía and Peter Shillingsburg (Amsterdam
and New York: Rodopi, 2005) pp. 29–55.
Skilton, David, ‘The Centrality of Literary Illustration in Victorian Visual
Culture: The Example of Millais and Trollope from 1860 to 1864’, Journal
of Illustration Studies 1 (December 2007). Available online at http://jois.uia.
no/articles.php?article=30.
Skilton, David, ‘Illustration Studies and the Infinite Archive’, Journal of
Illustration Studies 2 (December 2013). Available online at http://jois.uia.
no/articles.php?article=50.
Springer, Michelle, Beth Dulabahn, Phil Michel, Barbara Natanson, David Reser,
David Woodward and Helena Zinkham, For the Common Good: The Library
of Congress Flickr Pilot Project, report (30 October 2008). Available online at
http://www.loc.gov/rr/print/flickr_report_final.pdf.
Stauffer, Andrew, ‘Evidence and Interpretation in the Digital Age: Searching
Engines, Reading Machines’, Victorian Studies 54:1 (2011): 63–68.
Stevens, Joan, ‘Thackeray’s Pictorial Capitals’, Costerus: Essays in English and
American Language and Literature 2 (1974): 113–140.
Terras, Melissa, ‘Crowdsourcing in the Digital Humanities’, A New Companion
to Digital Humanities, ed. Susan Schreibman, Ray Siemens and John
Unsworth (Chichester: Wiley-Blackwell, 2016) pp. 420–438.
Terras, Melissa, ‘Cultural Heritage Information: Artefacts and Digitization
Technologies’, Cultural Heritage Information: Access and Management, ed.
Ian Ruthven and G. G. Chowdhury (London: Facet, 2015) pp. 63–88.
Terras, Melissa, ‘Reuse of Digitised Content (1)’, blogpost, 6 October 2014,
http://melissaterras.blogspot.co.uk/2014/10/reuse-of-digitised-content-
1-so-you.html.
Thackeray, William Makepeace, Vanity Fair. A Novel Without a Hero (London:
Bradbury and Evans, 1848).
112  Bibliography

Thomas, Julia, ‘Getting the Picture: Word and Image in the Digital Archive’,
European Journal of English Studies 11:2 (August 2007): 193–206.
Thomas, Julia, Pictorial Victorians: The Inscription of Values in Word and Image
(Athens, OH: Ohio University Press, 2004).
Thomas, Julia, ‘Reflections on Illustration: The Database of Mid-Victorian Wood-
Engraved Illustration’, Journal of Illustration Studies 1 (December 2007).
Available online at http://jois.uia.no/articles.php?article=37.
Trant, Jennifer, David Bearman and Susan Chun, ‘The Eye of the Beholder:
Steve.Museum and Social Tagging of Museum Collections’, proceedings of
ICHIM, 24–26 October 2007, Toronto, Canada. Available online at http://
www.archimuse.com/ichim07/papers/trant/trant.html.
Toronto, Jennifer, ‘Tagging, Folksonomy and Art Museums: Early Experiments
and Ongoing Research’, Journal of Digital Information 10:1 (2009).
Trollope, Anthony, Orley Farm, illustrated by John Everett Millais and engraved
by the Dalziel Brothers (London: Chapman and Hall, 1861–1862).
Trollope, Anthony, Orley Farm, ed. David Skilton (1861–1862; Oxford: Oxford
University Press, 1985).
Urbina, Eduardo, ‘Visual Knowledge: Textual Iconography of the Quixote’,
Don QuixoteIllustrated: Textual Images and Visual Readings/Iconografía del
Quijote, ed. Eduardo Urbina and Jesús G. Maestro, Biblioteca Cervantes 2
(Pontevedra: Mirabel Editorial, 2005) pp. 15–38.
Vandendorpe, Christian, From Papyrus to Hypertext: Toward the Universal Digital
Library, trans. Phyllis Aronoff and Howard Scott (1999; Urbana: University
of Illinois Press, 2009).
Vaughan, William, ‘Facsimile Versus White Line: An Anglo-German Disparity’,
Reading Victorian Illustration, 1855–1875: Spoils of the Lumber Room,
ed. Paul Goldman and Simon Cooke (Aldershot: Ashgate, 2012) pp. 33–52.
Wallace, Doireann, ‘Words as Keys to the Image Bank’, Revisualizing Visual
Culture, ed. Chris Bailey and Hazel Gardiner (Aldershot: Ashgate, 2010)
pp. 83–96.
Wan, Ji, Dayong Wang, Steven C. H. Hoi, Pengcheng Wu, Jianke Zhu,
Yongdong Zhang and Jintao Li, ‘Deep Learning for Content-Based Image
Retrieval: A Comprehensive Study’, 2014. Available online at https://pdfs.
semanticscholar.org/391b/86cf16c2702dcc4beee55a6dd6d3bd7cf27b.pdf.
White, Gleeson, English Illustration, ‘The Sixties’: 1855–1870 (1897; Bath:
Kingsmead Reprints, 1970).
‘Zooniverse releases Science Gossip’, Biodiversity Heritage Library blogpost,
4 March 2015, http://blog.biodiversitylibrary.org/2015/03/zooniverse-
releases-science-gossip.html.
 Bibliography   113

Web Resources
Arkyves http://arkyves.org.
Art and Architecture Thesaurus Online www.getty.edu/research/tools/vocabu-
laries/aat.
ARTigo www.artigo.org.
British Museum Collection Online http://www.britishmuseum.org/research/
collection_online/search.aspx.
British Printed Images to 1700 www.bpi1700.org.uk, dir. Michael Hunter.
Broadside Ballads Online http://ballads.bodleian.ox.ac.uk, dir. Alexandra
Franklin.
Brooklyn Museum www.brooklynmuseum.org.
Cervantes Project http://cervantes.tamu.edu, dir. Eduardo Urbina.
Citizen Archivist www.archives.gov/citizen-archivist.
Connected Histories www.connectedhistories.org.
Database of Mid-Victorian Illustration www.dmvi.cardiff.ac.uk, dir. Julia Thomas.
Database of Mid-Victorian Illustration (enhanced site) www.dmvi.org.uk, dir.
Julia Thomas.
Fleuron: A Database of Eighteenth-Century Printers’ Ornaments https://fleuron.
lib.cam.ac.uk/, dir. Hazel Wilkinson.
Flickr: The Commons www.flickr.com/commons.
HathiTrust Digital Library www.hathitrust.org.
Iconclass www.iconclass.nl.
Illustration Archive http://illustrationarchive.cardiff.ac.uk, dir. Julia Thomas.
Illustrating Scott http://illustratingscott.lib.ed.ac.uk, dir. Peter Garside.
Internet Archive https://archive.org.
Library of Congress Subject Headings http://id.loc.gov/authorities/subjects.
html.
Marie Duval Archive http://www.marieduval.org, dir. Simon Grennan.
MatchEngine https://services.tineye.com/MatchEngine.
Metadata Games www.metadatagames.org.
NINES www.nines.org.
Nineteenth-Century Serials Edition www.ncse.ac.uk, dir. Laurel Brake.
Philadelphia Museum of Art www.philamuseum.org.
Powerhouse Museum https://maas.museum/powerhouse-museum.
Rossetti Archive www.rossettiarchive.org, dir. Jerome McGann.
Science Gossip www.sciencegossip.org, dir. Sally Shuttleworth.
Transcribe Bentham www.transcribe-bentham.da.ulcc.ac.uk, dir. Philip Schofield.
Ukiyo-e http://ukiyo-e.org, dir. John Resig.
Victorian Illustrated Shakespeare Archive www.shakespeareillustration.org, dir.
Michael John Goodman.
114  Bibliography

Visual Geometry Group, University of Oxford, ‘Exploring the British Library’s 1

Million Images’ http://www.robots.ox.ac.uk/~vgg/research/BL.
Visual Haggard http://www.visualhaggard.org, dir. Kate Holterhoff.
Welsh Newspapers Online http://newspapers.library.wales.
William Blake Archive www.blakearchive.org, dir. Morris Eaves, Robert Essick
and Joseph Viscomi.
WordNet https://wordnet.princeton.edu.
The Yellow Nineties Online http://www.1890s.ca, dir. Dennis Denisoff and
Lorraine Janzen Kooistra.
Your Paintings, succeeded byArt UK http://artuk.org.
 Index

A B
Aberystwyth University, School of Art Bag of Words, 36. See also Content-
Gallery and Museum, 4 Based Image Retrieval (CBIR)
Adobe illustrator, 60 Baltussen, Lotte Belice, 89
Adobe Photoshop, 25, 26, 28 Barthes, Roland, 53
Advertisements, 78, 80, 83, 88 Bearman, David, 90
Affillustration, 12, 97–101 Beegan, Gerry, 31
Afsar, Kamil, 89 Behrendt, Stephen C., 97
Ahn, Luis von, 88 Bell, Peter, 61
Alexander, Mrs., A Life Interest, 40 Benford, Steve, 89
Allington, Daniel, 14 Bennett, Charles, 67
Althusser, Louis, 70 Beowulf, 80
Apple, 10 Bergel, Giles, 60
Aquatint, 37 Berger, John, 69
Arandjelović, Relja, 60 Bernstein, Shelley, 72, 74
Arkyves, 57 Bible, 17, 57
Aroyo, Lora, 89 Bibliographic metadata, 5, 21, 39, 40,
Art and Architecture Thesaurus Online 42, 43, 46, 51, 99
(AAT), 79, 80 Big Data, 5, 8, 47, 48
Art UK, 54. See also Your Paintings Blake, Erin C., 103
ARTigo, 67 Bloomfield, Robert‚ ‘The Farmer’s
Arts and Humanities Research Council Boy’, 49
(AHRC), 3, 83 Bolter, Jay David, 9, 29
Ashmolean Museum, Oxford, 4, 53 Bradley, John, 61

© The Editor(s) (if applicable) and The Author(s) 2017 115

J. Thomas, Nineteenth-Century Illustration and the Digital,
The Digital Nineteenth Century, DOI 10.1007/978-3-319-58148-4
116  Index
Brinkerink, Maarten, 89
British Library, 5, 22, 23, 35, 37–39,
42, 43, 78, 79, 96, 101
British Museum, 36
British Periodicals, 39
British Printed Images to 1700, 52
Broadside Ballads Online, 37
Brooklyn Museum. See Tag! You’re It!
Brouillette, Sarah, 14
Burdick, Anne, 13
C (DMVI), 3
Cameron, Julia Margaret, 31
Campbell, Sir Gilbert Edward,
Mysteries of the Unseen, 17, 18
Campbell, Thomas, 64
Captions, 10, 11, 43, 44, 47
Carletti, Laura, 88
Carroll, Lewis, Alice’s Adventures in
Wonderland, 21
Cassell, John, 86
Causer, Tim, 88
Cervantes Project, 2, 44, 46, 52
Chun, Susan, 90
Chung, J. S., 60
Churchill, Winston, 77
Cinema, 9
Citizen Science, 71, 83
Coburn, Alvin Langdon, 24
Coleridge, Samuel Taylor, ‘The Rime
of the Ancient Mariner’, 50
Computer vision, 11. See also Content-
Based Image Retrieval (CBIR)
Connected Histories, 100
Content-Based Image retrieval
(CBIR), 36
Cooke, Simon, 24
Cornhill Magazine, 20, 48
Courtroom illustration, 80
Crowdsourced image tagging, 11,
67–71, 77, 81, 88
Crowdsourcing, 65. See also
Crowdsourced image tagging
Cruikshank, George, 15
Cukier, Kenneth, 48, 75
Culin, Stewart, 72
D
Dabbish, Laura, 88
Dalziel Brothers, 45, 58, 69
Darnton, Robert, 18
Database of Mid-Victorian Illustration
Datta, Ritendra, 60
Decorations, 85. See also Ornaments
Deep Learning, 38
Denisoff, Dennis, 26. See also Yellow
Nineties Online
Dickens, Charles, 7, 21
Digital Humanities, 6–8, 18
Digitisation, 3–5, 21, 25, 35, 101
Don Quixote, 2, 46, 52, 86
Doré, Gustave, 86
Drucker, Johanna, 10, 69
Dulabahn, Beth, 89
Dunn, Stuart, 70
E
Eaves, Morris, 2. See also William
Blake Archive
Eccles, Kathryn, 89
Ees, Annelies Van, 89
Eliot, George; Romola, 4, 21
English Sacred Poetry, 48
Ernst, Wolfgang, 37
ESP game, 67
Essick, Robert N., 2. See also William
Blake Archive
Etching, 15, 23, 37, 42, 79
Evergreen: A Northern Seasonal, 85
Extra-Illustration, 98, 99
 Index   117

F Greg, Andrew, 89
Facebook, 74 Grennan, Simon, 39. See also Marie
Falstaff, 79. See also Shakespeare, Duval Archive
William Grusin, Richard, 29
Family Paper, 86
Favourite English Poems, 48–50
Fenn, George Manville, 86 H
First Impressions, 55 Hardwicke’s Science-Gossip, 83, 85. See
Flanagan, Mary, 91 also Science Gossip
Flaubert, Gustave, 7, 54 Hardy, Thomas, 21
Fleuron, 79, 85 Harpring, Patricia, 22
Flickr, 22, 23, 47, 55, 67, 70, 74, 75 Harvey, Ian, 61
Fogg, Martha, 63 HathiTrust Digital Library, 42
Folksonomies, 74, 78 Hayles, Katherine N., 27
Foster, Myles Birket, 49, 51, 59 Hedges, Mark, 70
Franklin, Andrew, 60 Heilmann, Ann, 93
Frick Photoarchive, 37 Historical Texts, 39
Frontispieces, 78, 80 Hockney, David, 22
Fyfe, J. Hamilton, ‘About Toys’, 58 Hocks, Mary E., 9
Hofstede, Geert, 70
Hoi, Steven C.H., 61
G Holterhoff, Kate, 21. See also Visual
Galloway, Alexander R., 14 Haggard
Genette, Gérard, 20 Home, Anne, The Sports of the Genii,
Getty Research Institute, 79 41. See also Macdonald, Susan
Giannachi, Gabriella, 89 Hood’s Own: or, Laughter from Year to
Gift books, 21, 51 Year, 22
Gligorov, Riste, 89 Houghton, Arthur Boyd, 59
Goldman, Paul, 5, 18, 25 Hypermedia, 11
Golumbia, David, 14
Good Words, 4, 20
Goodman, Michael, 28. See also I
Victorian Illustrated Shakespeare IBM, 70
Archive Iconclass, 57, 59
Goodyear, William Henry, 72 Idyllic School, 20
Google, 3, 4, 10, 37, 39, 51, 67 Illuminated Magazine, 98
Google Image Labeler, 66 Illustrated London News Historical
Gore, Catherine, 29 Archive, 39
Gould, Emilie West, 70 Illustrating Scott, 21, 53, 100
Graham, Margaret E., 60 Illustration Archive, 2, 3, 5, 8, 21, 22,
Granger, James, Biographical History 25, 28, 35, 37, 38, 40, 42, 43,
of England, 98
118  Index

46, 56, 67, 75–78, 81, 83, 88, Layne, Sara Shatford, 78
96, 97, 99–101 Leech, John, 98
Illustration Studies, 5–9, 12, 20, 100, Leetaru, Kalev, 46, 47
101 Leighton, Frederic, 4, 49
ImageNet, 39 Lessing, Gotthold, 55
Indianapolis Museum of Art. See First Li, Jia, 60
Impressions Li, Jintao, 61
Instagram, 74 Library of Congress Subject Headings
Interface, 9, 10, 69–71 (LCSH), 80
Intermediality, 13 Library of Congress, 68. See also
Internet Archive, 42, 46 Flickr; Library of Congress Subject
Intertextuality, 97 Headings (LCSH)
Ionescu, Christina, 8 Limonard, Sander, 89
Lithography, 37, 79
Lloyd, Nicky, 61
J London Society, 4, 20, 51
Jackson, John, 32 Los Angeles Review of Books, 8
James, Henry, 7, 24 Lunenfeld, Peter, 63
Jessop, Martyn, 6, 7
Jockers, Matthew L., 99
Jones, Owen, 85 M
Joshi, Dhiraj, 60 Macdonald, Susan, The Sports of the
Jung, Sandro, 7 Genii, 41
Mackay, Charles, 91
Maidment, Brian, 31
K Mandal, Anthony, 14
Kaufman, Geoff, 91 Manovich, Lev, 27, 28, 70
Kelmscott Press, 85 Manzo, Christina, 91
Kendrick, Michelle R., 9 Maps, 38–40, 46, 88
Keywording, 4. See also Tagging Marcus, Aaron, 70
Kim, Sung-Do, 30 Marie Duval Archive, 39
Kingston, W. H. G., 86 Martineau, Harriet, The Anglers of the
Kirschenbaum, Matthew, 37 Dove, 34
Kittler, Friedrich, 10, 19 Maxwell, Richard, 29
Klingemann, Mario, 38 Mayer-Schönberger, Viktor, 48, 51, 75
Knight, Charles, 85 McAuley, Derek, 88
Kooistra, Lorraine Janzen, 34. See also McGann, Jerome, 2. See also Rossetti
Yellow Nineties Online Archive
McPherson, Tara, 21
Mendelsund, Peter, 54
L Metadata Games, 67
Landon, Letitia Elizabeth, 2 Meyrick, Robert, 13

Michel, Phil, 89
Microsoft, 5, 21, 35
Millais, John Everett, 4, 17
Miller, J. Hillis, 22
Mitchell, W. J. T, 3, 36
Moore, George, 93
Moretti, Franco, 103
Mosaic, 10
Music, 71
Moxon, Edward, 21
Mussell, James, 29, 85
N False?’, 45
Natanson, Barbara, 89
Nelson, Horatio, 77
Newspapers, 2, 24, 39, 80, 85, 96, 97
Nightingale, Florence, 77
NINES, 100
Nineteenth- Century British
Newspapers, 39
Nineteenth-Century Serials Edition
(NCSE), 53
Nineteenth-Century UK Periodicals, 39
Nowell-Smith Simon, 92
Nowviskie, Bethany, 5
O Reser, David, 89
OCR (Optical Character Recognition),
34, 39, 43
Ommer, Björn, 61
Once a Week, 4, 20, 48, 86
Oomen, Johan, 89
Ornaments, 24. See also Decorations
Owens, Trevor, 69
P Archive
Paget, Sidney, 54
Paratext, 20, 26
Petter, George William, 86
Index   119
Photography, 9, 23, 57, 81
Pictorial capital letters, 10, 49, 81, 83
Pigney, Stephen, 61
Piper, Andrew, 96
Portraits, 77
Posthumus, Etienne, 64
Powerhouse Museum, Sydney,
Australia, 73
Pre-Raphaelites, 20
Presner, Todd, 13
Price, Dominic, 88–89
Price, Leah, 23
Proctor, Adelaide Anne, ‘True or
Punch Historical Archive, 39
Punjasthitkul, Sukdith, 91
Q
Queen Elizabeth II, 77
Quiver, 86
R
Rajewsky, Irina O., 3, 6
Reid, Forrest, 4, 20, 98
Reid, Thomas Mayne, 86
Remediation, 9, 12, 29
Resig, John, 37. See also Ukiyo-e.org
Ridge, Mia, 68
Rossetti Archive, 2, 21, 27, 100
Ruskin, John, 54
S
Sabharwal, Arjun, 91
Sabin, Roger, 39. See also Marie Duval
Saussure, Ferdinand de, 19, 43
Scanning, 4. See also Digitisation
Schlecht, Joseph, 61
120  Index
Schnapp, Jeffrey, 13
Science Gossip, 83–85, 87, 88
Scientific illustration, 80, 83
Seed, Patricia, 93
Shakespeare, William, 80
Shillingsburg, Peter L., 29
Sillars, Stuart, 103
Skilton, David, 6, 98
Spinden, Herbert, 72
Springer, Michelle, 89
Stauffer, Andrew, 12, 92
Steve.museum, 73
Stevens, Bethan, 32
Stevens, Joan, 62
Swain, Joseph, 34
T Visual Haggard, 21
Tag! You’re It!, 71, 72, 74
Tagging, 5. See also Keywording,
Crowdsourced image tagging
Tenniel, John, 15, 21
Tennyson, Alfred Lord, 21
Terras, Melissa, 22, 25
Text Encoding Initiative (TEI), 52
Thackeray, William Makepeace; Vanity
Fair, 17
The Times Digital Archive, 39
Thomas, Julia, 3. See also Database
of Mid-Victorian Illustration;
Illustration Archive
Titlepages, 78, 80, 83, 88
Transcribe Bentham, 65
Trant, Jennifer, 90
Trollope, Anthony; Framley Parsonage;
Orley Farm, 4, 38, 50, 98
Tumblr, 10
Twitter, 74
U
Ukiyo-e.org, 37
Urbina, Eduardo, 2. See also Cervantes
Project
V
Vandendorpe, Christian, 11
Vaughan, William, 28, 42
Vervaart, Just, 89
Victorian Illustrated Shakespeare
Archive, 26, 28, 99, 100
Vignettes, 24, 78
Viscomi, Joseph, 2. See also William
Blake Archive
Visual Geometry Group, University of
Oxford, 37
Visualisation, 27
W
Waal, Henri van de. See Iconclass, 57
Waite, Julian, 39. See also Marie Duval
Archive
Wallace, Doireann, 75
Wallace, Valerie, 88
Wan, Ji, 61
Wang, Dayong, 61
Wang, James Z, 60
Watson, John Dawson, 62
Wehnert, Edward Henry, 62
Welsh Newspapers Online, 39
White, Gleeson, 20, 24, 99
Wilde, Oscar, 21
William Blake Archive, 2, 21, 26, 27,
52, 100
Wilson, Margaret, 57
 Index   121

Wood engraving, 10, 17, 23, 28, 42, Z

86, 100 Zhang, Yongdong, 61
Woodward, David, 89 Zhu, Jianke, 61
WordNet, 75, 76, 81 Zinkham, Helena, 89
Wu, Pengcheng, 61 Zisserman, Andrew, 60
Zoomify, 28, 29
Zooniverse, 76, 83
Y
Yellow Nineties Online, 2, 21, 25, 26,
53, 56, 100
Your Paintings, 54. See also Art UK