Untitled


University of the Arts London

Transforming structured descriptions to visual
representations. An automated visualization

of historical bookbinding structures

Alberto Campagnolo

Thesis for the degree of Doctor of Philosophy

Volume 1

May 2015

About this document
This document has been generated with RenderX XEP.
					Visit http://www.renderx.com/ to learn more about
					RenderX family of software solutions for digital
					typography.



Abstract

In cultural heritage, the documentation of artefacts can be both iconographic

and textual, i.e. both pictures and drawings on the one hand, and text and words

on the other are used for documentation purposes.

This research project aims to produce a methodology to transform automatically

verbal descriptions of material objects, with a focus on bookbinding structures,

into standardized and scholarly-sound visual representations.

In the last few decades, the recording and management of documentation data

about material objects, including bookbindings, has switched from paper-based

archives to databases, but sketches and diagrams are a form of documentation

still carried out mostly by hand. Diagrams hold some unique information, but

often, also redundant information already secured through verbal means within

the databases. This project proposes a methodology to harness verbal information

stored within a database and automatically generate visual representations.

A number of projects within the cultural heritage sector have applied semantic

modelling to generate graphic outputs from verbal inputs. None of these has

considered bookbindings and none of these relies on information already recorded

within databases. Instead they develop an extra layer of modelling and typically

gather more data, specifically for the purpose of generating a pictorial output. In

these projects qualitative data (verbal input) is often mixed with quantitative data

(measurements, scans, or other direct acquisition methods) to solve the problems

of indeterminateness found in verbal descriptions. Also, none of these projects

has attempted to develop a general methodology to ascertain the minimum amount

ii



of information that is required for successful verbal-to-visual transformations for

material objects in other fields. This research has addressed these issues.

The novel contributions of this research include: (i) a series of methodological

recommendations for successful automated verbal-to-visual intersemiotic transla-

tions for material objects — and bookbinding structures in particular — which

are possible when whole/part relationships, spatial configurations, the object’s

logical form, and its prototypical shapes are communicated; (ii) the production

of intersemiotic transformations for the domain of bookbinding structures; (iii)

design recommendations for the generation of standardized automated prototyp-

ical drawings of bookbinding structures; (iv) the application — never considered

before — of uncertainty visualization to the field of the archaeology of the book.

This research also proposes the use of automatically generated diagrams as data

verification tools to help identify meaningless or wrong data, thus increasing data

accuracy within databases.

iii



Table of contents

Volume 1

Abstract...................................................................................................................ii
Table of contents....................................................................................................iv
Table of figures.......................................................................................................x
Table of tables......................................................................................................xiv
Table of listings.....................................................................................................xv
Acknowledgements.............................................................................................xvii

Chapter 1. Introduction..........................................................................................1

1.1. The cognitive importance of diagrams...........................................................4
1.1.1. Diagrams as visual proofs.............................................................................5
1.2. Problems with the adoption of diagrams........................................................6
1.3. Verbal to visual: intermediation & intersemiotic translations........................7
1.3.1. The essential elements of the communication cycle....................................8
1.4. Bookbinding descriptions and the generation of diagrams...........................8
1.5. Terminological issues.....................................................................................11
1.6. Aims and objectives.......................................................................................12
1.7. Chapter overview............................................................................................13

Chapter 2. Modelling & related work..................................................................16

2.1. Modelling.......................................................................................................16
2.1.1. Modelling of and modelling for..................................................................18
2.2. From verbal to visual: related projects and approaches...............................19
2.2.1. Hardware description languages................................................................19
2.2.2. Natural language processing......................................................................20
2.2.3. Cultural heritage modelling........................................................................22
2.3. From verbal to visual through models of......................................................25
Summary...............................................................................................................28

iv



Chapter 3. Signs & material objects.....................................................................30

3.1. Signs, signification, and communication......................................................31
3.1.1. Signs............................................................................................................32
3.1.2. Reading and remembering iconic signs.....................................................34
3.1.2.1. Reading modalities of icons.....................................................................34
3.1.2.2. Formalization of hypoicons.....................................................................36
3.1.3. Visual texts..................................................................................................38
3.2. Multimodal communication..........................................................................40
3.2.1. Dual-coding representation........................................................................41
3.2.2. Verbal recoding of ambiguous visual stimuli.............................................42
3.3. Categorization and prototypification.............................................................43
3.3.1. Prototypes...................................................................................................43
3.3.2. Basic level of abstraction............................................................................45
3.4. Objects: their form and their shape..............................................................50
3.4.1. The form of objects....................................................................................50
3.4.1.1. Material components and form...............................................................51
3.4.1.2. Logical form and expression...................................................................51
3.4.2. The shape of objects...................................................................................53
3.4.2.1. The nature of shape.................................................................................54
3.4.2.2. Regularities for the prototypification of shape.......................................55
3.4.2.3. Line drawings..........................................................................................56
3.4.2.4. Perceptual biases.....................................................................................58
Summary...............................................................................................................59

Chapter 4. Bookbinding descriptions...................................................................61

4.1. Books as artefacts...........................................................................................61
4.1.1. Binding structures and decoration.............................................................62
4.1.1.1. The evolution of bookbinding studies....................................................63
4.1.2. Terminology................................................................................................67
4.2. Verbal and visual descriptions of bindings...................................................69
4.2.1. A description of Coptic bindings................................................................70
4.2.2. A particular sewing pattern........................................................................71
4.2.3. A controlled vocabulary description..........................................................75
4.2.4. Structured and controlled vocabulary descriptions..................................77
4.3. Prototypical visualization of bookbinding elements....................................81
4.3.1. Gathering assembly diagrams....................................................................82
4.3.1.1. Naturalistic representations of gatherings..............................................83
4.3.1.2. Schematic representations of gatherings.................................................84
4.4. A categorization of bookbinding illustrations..............................................87
4.4.1. Archaeological-style drawings....................................................................87
4.4.2. Naturalistic drawings..................................................................................88
4.4.3. Line drawings.............................................................................................90
4.4.4. Generic-shape drawings.............................................................................92

v



4.4.5. Scenes..........................................................................................................93
4.5. Clarity of information in bookbinding line drawings...................................95
4.5.1. Sketchiness and detail views.......................................................................97
Summary.............................................................................................................101

Chapter 5. Communicating & translating objects..............................................103

5.1. Communicating material objects through verbal means.............................104
5.1.1. Communicating through controlled vocabularies....................................105
5.1.1.1. Names and prototypical material components.....................................107
5.1.1.2. Sewing structure controlled vocabulary...............................................108
5.1.2. Communicating through structured descriptions....................................109
5.1.3. Diagrams as support for natural language descriptions..........................111
5.2. Automated intersemiotic translations.........................................................112
5.2.1. Translating material components, their form, and shape........................112
5.2.2. Translating material objects’ shapes.........................................................113
5.2.3. Translating an object’s form.....................................................................114
5.2.4. The reference to the logical form.............................................................114
5.3. Communicating material objects through visual means.............................115
5.3.1. Marks on a background............................................................................116
5.3.2. Continuity of the plane.............................................................................117
5.3.3. Visualization conventions.........................................................................117
5.3.3.1. Fixing meaning through a reference system.........................................118
5.3.3.1.1. Titling and verbal labelling.................................................................119
5.3.3.1.2. Graphic reference elements...............................................................119
5.3.3.2. Separation of information.....................................................................119
5.3.3.3. Use of colour..........................................................................................120
5.3.3.4. Two-dimensional vs. three-dimensional visualizations.........................121
5.3.4. Structured visualizations as output..........................................................121
5.4. The uncertainty of reality............................................................................122
5.4.1. Uncertainty in structured descriptions....................................................125
5.4.2. Uncertainty in visualizations....................................................................126
5.4.2.1. Archaeological visualizations................................................................126
5.4.2.1.1. An example from archaeology of the book.......................................127
5.4.2.1.2. Uncertainty clues in archaeological visualizations.............................128
5.4.2.2. Irregularities in visual propositions.......................................................129
5.4.2.2.1. Visual cues for uncertainty visualization............................................129
5.4.2.3. Imprecision as uncertainty....................................................................132
5.4.2.4. Uncertainty inherent in the visualization process.................................132
5.4.3. Human error in the data...........................................................................133
Summary.............................................................................................................135

Chapter 6. Transformation framework...............................................................138

vi



6.1. The technology and the description schema..............................................139
6.1.1. Extensible Markup Language..................................................................139
6.1.2. Extensible Stylesheet Language Transformations...................................140
6.1.3. Scalable Vector Graphics.........................................................................141
6.1.4. The description schema and the dataset..................................................141
6.1.4.1. The Ligatus schema for the description of bookbinding structures....142
6.1.4.1.1. The manuscript collection survey......................................................142
6.1.4.1.2. The printed book collection survey...................................................143
6.1.4.1.3. Free-hand drawings............................................................................145
6.1.4.2. The project dataset................................................................................146
6.2. The transformations....................................................................................147
6.2.1. Knowledge graph schemas.......................................................................147
6.2.1.1. Graphic conventions for knowledge graph schemas...........................149
6.2.1.2. Reading the visual description graphs..................................................151
6.2.2. Number of possible visualizations...........................................................152
6.2.2.1. Board marker visualizations..................................................................153
6.2.2.2. Endleaf structure visualizations............................................................154
6.2.3. Visualizations............................................................................................157
6.2.3.1. Level of abstraction and semantic entry point.....................................158
6.2.3.2. Diagram elements..................................................................................161
6.2.3.3. Spatial arrangement...............................................................................163
6.2.3.4. Thread paths..........................................................................................165
6.2.3.5. Uncertainty............................................................................................166
6.2.3.6. Accuracy and imprecision of measurements........................................172
6.2.3.7. Place-holders for uncertainty inherent in the visualization process....174
Summary.............................................................................................................176

Chapter 7. Transformations & analysis...............................................................177

7.1. A complete visualization example: endleaves.............................................178
7.1.1. Visual description framework..................................................................178
7.1.2. Shapes: standard endleaf components.....................................................179
7.1.2.1. Reference elements................................................................................181
7.1.2.2. Endleaf components..............................................................................182
7.1.2.2.1. Pastedowns in use visualizations........................................................182
7.1.2.3. Attachment modalities...........................................................................184
7.1.2.3.1. Sewn....................................................................................................184
7.1.2.3.2. Glued..................................................................................................185
7.1.2.4. Component material...............................................................................185
7.1.2.5. Spatial arrangement...............................................................................186
7.1.3. Shapes: different types of endleaves........................................................186
7.1.3.1. Integral endleaves..................................................................................186
7.1.3.2. Separate endleaves.................................................................................187
7.1.3.2.1. Single leaf............................................................................................187
7.1.3.2.2. Fold.....................................................................................................188

vii



7.1.3.2.3. Hooks..................................................................................................188
7.1.3.2.3.1. Endleaf hook....................................................................................188
7.1.3.2.3.2. Text hook.........................................................................................189
7.1.3.2.4. Outside hook......................................................................................189
7.1.3.2.5. Guard..................................................................................................190
7.1.4. Visualization problems.............................................................................190
7.1.4.1. Schema limitations.................................................................................191
7.1.4.2. Schema misinterpretations....................................................................196
7.2. The set of transformations...........................................................................202
7.2.1. Markers: page markers, board markers, bookmarks...............................202
7.2.2. Sewing.......................................................................................................208
7.2.3. Boards.......................................................................................................213
7.2.4. Spine shape and lining..............................................................................216
7.2.5. Endbands..................................................................................................223
7.2.6. Coverings...................................................................................................224
7.2.7. Furniture...................................................................................................236
7.3. Dealing with erroneous data.......................................................................237
7.3.1. Error examples.........................................................................................239
7.3.2. The feedback loop....................................................................................242
Summary.............................................................................................................244

Chapter 8. Recommendations.............................................................................245

8.1. Intersemiotic translations: methodological recommendations..................245
8.1.1. Intersemiotic translation elements...........................................................246
8.1.2. Functional verbal descriptions.................................................................246
8.1.3. The nature of sememes in diagrams.........................................................248
8.1.4. Prototypicality of information..................................................................249
8.1.5. Graphic prototypes..................................................................................251
8.1.6. Fixing the meaning of visual sememes.....................................................252
8.1.7. Uncertainty...............................................................................................252
8.2. Recommendations for the communication of bookbinding structures.....252
Summary.............................................................................................................254

Chapter 9. Conclusions........................................................................................255

9.1. Successful automated intersemiotic translations........................................256
9.1.1. Summary of the recommendations..........................................................257
9.2. Visualizations from models of.....................................................................258
9.3. Standardized automated drawings of bookbinding structures..................259
9.3.1. Summary of design principles..................................................................262
9.4. Uncertainty visualization in bookbinding studies......................................263
9.4.1. Summary of uncertainty visualization application...................................264
9.5. Added value & benefits of visualization automation.................................265

viii



Chapter 10. Future work.....................................................................................268

10.1. Integration of automated transformations within the data gathering
framework...........................................................................................................268
10.2. Verbal to visual and reverse: a mixed approach.......................................269
10.3. Related work: Manuscript Collation Project.............................................271
10.4. Final remarks.............................................................................................274

Bibliography........................................................................................................275
Illustration credits...............................................................................................314

Volume 2

Appendix A. Human memory system...............................................................318

Appendix B. Shape library.................................................................................321

B.1. Page markers...............................................................................................322
B.2. Board markers.............................................................................................235
B.3. Bookmarks...................................................................................................328
B.4. Endleaves.....................................................................................................335
B.5. Sewing..........................................................................................................340
B.6. Boards..........................................................................................................347
B.7. Spine shape..................................................................................................355
B.8. Spine lining..................................................................................................359
B.9. Endbands....................................................................................................368
B.10. Coverings....................................................................................................383
B.11. Furniture - fastenings................................................................................408

Appendix C. Coding example: endleaves...........................................................413

C.1. Endleaf XSLT..............................................................................................414
C.2. Endleaf master SVG....................................................................................473
C.3. Endleaf CSS..................................................................................................476

ix



Table of figures

Figure 1. Exograms.................................................................................................5
Figure 2. Communication cycle..............................................................................9
Figure 3. Example of Text-to-Picture transformation.........................................21
Figure 4. Example of classical architecture column parts...................................26
Figure 5. Basic structure of signs.........................................................................33
Figure 6. Mexican on a bicycle............................................................................35
Figure 7. Mexican on a bicycle — as seen from above.......................................37
Figure 8. Units in visual language are context dependent..................................39
Figure 9. An angle growing from 0° to 180°........................................................46
Figure 10. Dimensions of a category systems......................................................47
Figure 11. Stirrup ring shape and its prototypified diagram..............................57
Figure 12. A sewing pattern for a Coptic binding...............................................72
Figure 13. Knot-tack sewing................................................................................73
Figure 14. Knot-tack sewing: alternative solution...............................................74
Figure 15. Unsupported sewing sample..............................................................75
Figure 16. Sewing photograph and drawing example.........................................78
Figure 17. Archaeological drawings.....................................................................83
Figure 18. Naturalistic and three-dimensional representations of gatherings....84
Figure 19. Schematic representations of gatherings............................................85
Figure 20. Categorical dimensions of drawings...................................................88
Figure 21. Archaeological-style drawings............................................................89
Figure 22. Naturalistic drawings..........................................................................91
Figure 23. Line drawings, the graphic basic level of abstraction........................93
Figure 24. Generic-shape drawings.....................................................................94
Figure 25. Bookshop scene..................................................................................96
Figure 26. Well-shaped and free-hand thread paths...........................................98
Figure 27. Drawing of an endband sewn at the frame........................................98
Figure 28. Sketch-like illustrations on how to work a Greek-style endband.....99
Figure 29. Step 11.................................................................................................99
Figure 30. Line drawings of different styles of endbands.................................100
Figure 31. Diagram of the construction of the Nag Hammadi codices............127
Figure 32. Dashed lines indicating both invisible parts and folds....................131
Figure 33. Paper survey form.............................................................................144
Figure 34. Drawing survey paper form..............................................................146

x



Figure 35. Visualization of the Ligatus schema.................................................148
Figure 36. Example graph schema showing the graphical conventions...........152
Figure 37. Board-marker graph schema............................................................154
Figure 38. Endleaf graph schema.......................................................................155
Figure 39. Board visualization example.............................................................159
Figure 40. Example of mechanical drawing with three projected views..........160
Figure 41. Fastening visualization example.......................................................160
Figure 42. Visualization of the Ligatus schema.................................................162
Figure 43. Gothic binding bosses......................................................................163
Figure 44. Schematic endleaf drawings.............................................................164
Figure 45. Example of a complex endleaf structure.........................................165
Figure 46. Examples of cross-section visualizations for endbands I................167
Figure 47. Examples of cross-section visualizations for endbands II...............168
Figure 48. Conjoined double hook detail...........................................................170
Figure 49. Cross-section of endband with no-front-bead sewing.......................171
Figure 50. Fading off of board groove towards the spine.................................171
Figure 51. Front view of a double-core greek-style endband...........................173
Figure 52. Cross-section and front views of a stuck-on endband......................174
Figure 53. Above view of a roller-round-bar catchplate...................................175
Figure 54. Sketchy lines and blurring to indicate uncertainties........................176
Figure 55. Example of endleaf diagram.............................................................179
Figure 56. Complete set of endleaf diagram......................................................179
Figure 57. Graph schema for endleaf diagrams................................................180
Figure 58. Example of integral endleaves..........................................................182
Figure 59. The five basic components of separate endleaves...........................182
Figure 60. Examples of schematic endleaf diagrams.........................................183
Figure 61. Example of endleaf diagram with sewing thread convention.........184
Figure 62. Example of glued component..........................................................185
Figure 63. Example of glued components.........................................................185
Figure 64. Endleaf structure with internal parchment guard...........................186
Figure 65. Example of a complex endleaf structure.........................................186
Figure 66. Example of integral endleaves..........................................................187
Figure 67. Example of single leaf unit...............................................................187
Figure 68. Example of separate pastedown.......................................................187
Figure 69. Example of fold components...........................................................188
Figure 70. Example of endleaf-hook components............................................189
Figure 71. Example of text-hook component...................................................189
Figure 72. Example of outside-hook component with text hooks...................190
Figure 73. Example of guard..............................................................................190
Figure 74. Hand drawing for the right endleaves.............................................191
Figure 75. Example of endleaf diagram.............................................................192
Figure 76. Endleaf hand drawing example........................................................192
Figure 77. Example of automated endleaf diagram..........................................193
Figure 78. Example of endleaf drawing.............................................................193
Figure 79. Example of endleaf diagram.............................................................194

xi



Figure 80. Detail of endleaf diagram.................................................................194
Figure 81. Example of endleaf hand drawing...................................................195
Figure 82. Example of wrong endleaf diagram.................................................195
Figure 83. Example of endleaf diagram.............................................................195
Figure 84. Example of wrong endleaf diagram.................................................195
Figure 85. Example of endleaf hand drawing...................................................196
Figure 86. Example of wrong endleaf diagram.................................................196
Figure 87. Example of endleaf hand drawing...................................................197
Figure 88. Example of endleaf diagram.............................................................197
Figure 89. Example of endleaf diagram.............................................................198
Figure 90. Endleaf diagram generated from corrected XML...........................198
Figure 91. Example of endleaf hand drawing...................................................199
Figure 92. Example of visual output of an incorrect description.....................199
Figure 93. Endleaf diagram generated from corrected XML...........................199
Figure 94. Example of pagemarker diagram.....................................................203
Figure 95. Example of boardmarker diagram...................................................203
Figure 96. Graph schema for pagemarker diagrams.........................................204
Figure 97. Graph schema for boardmarker diagrams.......................................205
Figure 98. Example of bookmarker diagram....................................................206
Figure 99. Graph schema for bookmark diagrams...........................................207
Figure 100. Examples of sewing diagrams.........................................................209
Figure 101. Example of sewing measurement diagram.....................................210
Figure 102. Graph schema for sewing measurement diagrams........................211
Figure 103. Examples of change-over stations..................................................213
Figure 104. Example of sewing measurement diagram.....................................214
Figure 105. Graph schema for sewing measurement diagrams........................215
Figure 106. Example of board diagram.............................................................217
Figure 107. Graph schema for board diagram..................................................218
Figure 108. Graph schema for board diagram..................................................219
Figure 109. Graph schema for spine shape diagrams.......................................219
Figure 110. Example of spine lining diagram....................................................220
Figure 111. Graph schema for spine lining diagrams.......................................221
Figure 112. Example of sewn endband diagram...............................................224
Figure 113. Graph schema for sewn endband diagrams...................................225
Figure 114. Example of stuck-on endband diagram.........................................226
Figure 115. Graph schema for sewn endband diagrams...................................227
Figure 116. Example of guard covering diagram..............................................229
Figure 117. Graph schema for guard or drawn-on covering diagrams............229
Figure 118. Example of over-in-board covering diagram.................................230
Figure 119. Graph schema for over-in-board covering diagrams.....................231
Figure 120. Example of case covering diagram.................................................232
Figure 121. Graph schema for case covering diagrams....................................233
Figure 122. Example of furniture diagram........................................................234
Figure 123. Graph schema for furniture diagrams............................................235
Figure 124. Example of fastening photographs................................................237

xii



Figure 125. Example of spine shape hand drawing..........................................238
Figure 126. Example of spine shape diagram....................................................238
Figure 127. Error due to schema convention misinterpretation.......................240
Figure 128. Error due to description convention misinterpretation................240
Figure 129. Error due to inaccurate encoding..................................................240
Figure 130. Slip error during encoding.............................................................240
Figure 131. Project communication cycle..........................................................243
Figure 132. Errors in the hierarchical arrangement of elements......................249
Figure 133. Small multiples................................................................................270
Figure 134. Screenshot from the Manuscript Collation Project webpage........273
Figure 135. A model of the human memory system..........................................319

xiii



Table of tables

Table 1. Samples of ambiguous visual stimuli.....................................................42
Table 2. Table showing the cause for error examples........................................242

xiv



Table of listings

Listings 1. Snippet of XML description of a sewing station...............................79
Listings 2. Snippet of endleaf XML description...............................................200
Listings 3. Corrected endleaf XML snippet description...................................201

xv



Acknowledgements

Qual è ‘l geomètra che tutto s'affige / per misurar lo cerchio, e non

ritrova, / pensando, quel principio ond’ elli indige, // tal era io a

quella vista nova: / veder voleva come si convenne / l’imago al cer-

chio e come vi s’indova

As the geometrician, who endeavours / To square the circle, and

discovers not, / By taking thought, the principle he wants, // Even

such was I at that new apparition; / I wished to see how the image

to the circle / Conformed itself, and how it there finds place*

Dante Alighieri, The Divine Comedy: Paradise, XXXIII, 133-138.

*Henry Wadsworth Longfellow translation (1867)

I would like to express my sincere gratitude to Dr Athanasios Velios and Professor

Nicholas Pickwoad, my research supervisors, for their guidance, and enthusiastic encour-

agement throughout this research work. My grateful thanks are also extended to the Arts

and Humanities Research Council for funding this research and for giving me the oppor-

tunity to pursue it.

My most heartfelt thanks also go to Dr Elena Pierazzo, Dr Peter A. Stokes, Dr John

Bradley, and Dr Raffaele Viglianti at the Department of Digital Humanities, King’s College

London for their support and for allowing me to attend, as auditor, their courses in

‘Advanced Text Technologies: XML, TEI, XSLT’ and ‘Python Programming’. I am also

indebted to Syd Bauman and Martin Holmes for their helpful course (and subsequent

pointers) in XSLT programming ‘Introduction to XSLT for Digital Humanities’ at the

Digital Humanities Summer Institute 2011, University of Victoria, BC, Canada. I would

also like to thank the Social Sciences and Humanities Research Council of Canada and

the Association of Computers and the Humanities for granting me a scholarship and

xvii



travel bursary to attend the workshop at the Digital Humanities Summer Institute 2011.

My thanks also go to the Alliance of Digital Humanities Organizations for granting me

a bursary to attend and present my paper at the Digital Humanities Conference 2013,

University of Nebraska-Lincoln.

I would like to express my very great appreciation to Dr Alejandro Giacometti for his

valuable suggestions on the management of my digital data and my research in general,

and for his help in the development of an SVG filter for the rendition of sketchy lines. I

would also like to extend my thanks to Dot Porter and Doug Emery at the University of

Pennsylvania for allowing me to collaborate with them in their Manuscript Collation

Project, and thus to test the methodology developed for this project on a different dataset.

My gratitude also goes to Dr Arianna Ciula, my dissertation supervisor for the Master of

Arts in Digital Culture and Technology at the Department of Digital Humanities (formerly

Centre for Computing in the Humanities), King’s College London for her interest, support,

and stimulating questions and discussions during the course of this research.

I am grateful for the interest shown in my research by Professor Carlo Federici, Melania

Zanetti, Director of the Associazione Italiana dei Conservatori e Restauratori degli Archivi

e delle Biblioteche (Italian Association of Archive and Library Conservator-restorers),

and Andrea Galeazzi, Director of Kermes — La rivista del restauro (Kermes — The

Magazine of Conservation).

I am deeply thankful to Hilary Canavan for her language support and to her partner

James Scott for accommodating me in their house. I am also grateful to Dr Caroline De

Stefani for her support and for offering me a place to stay during my visits to London

this last year.

Finally I would like to thank all my friends and colleagues at the Vatican Library. In

particular I would like to thank Msgr. Cesare Pasini, Prefect of the Vatican Library, Dr

Ambrogio M. Piazzoni, Vice Prefect of the Vatican Library, and Ángela Nuñez Gaitán,

Head of the Conservation Workshop at the Vatican Library for allowing me to work

flexibly and to take the necessary time off to finish this research. Special thanks also go

to Elena Antonazzo, Sister Gabriella Pettirossi, Maria Rosaria Castelletti, Marta Grimaccia,

Maruska Di Remigio, Pierre Sentagne, Silvia Foschetti, and Simona De Crescenzo for

their support and patience. I would also like to thank Dr Luca Milasi for his support and

for offering me a laptop on which to work in times of need.

Finally, I wish to thank my friends and family for their support and encouragement

throughout my studies.

xviii



Quidquid agis, prudenter agas, et respice finem

Whatever you do, do it carefully, and with an eye to the end

(Unsere Probleme sind nicht abstrakt, sondern vielleicht die

konkretesten, die es gibt.)

(Our problems are not abstract, but perhaps the most concrete

that there are.)*

Ludwig Wittgenstein, Tractatus logico-philosophhicus, 1922.

*translation by Pears & McGuinnes (1961)

xix



Chapter 1. Introduction

[La pittura] col suo principio, cioè il disegno, [...] insegna ai

prospettivi ed astrologhi ed ai macchinatori ed ingegneri.

[Painting] with its principle, i.e. drawing, [...] teaches to per-

spectivists and astrologists and machiners and engineers.*

Leonardo da Vinci, Libro di pittura, Biblioteca Apostolica

Vaticana MS Urb.lat.1270, [ca. 1540], 12v. *Author’s transla-

tion

In the cultural heritage field, the documentation of artefacts is both iconographic

and textual, i.e. both pictures and drawings on the one hand, and text and words

on the other are used for documentation purposes.1 This research project aims to

advance a methodology to transform automatically verbal descriptions of material

objects, such as bookbinding structures, into standardized and scholarly-sound

visual representations.

Graphic-representation methods range from drawings, sketches and watercol-

ours, to direct acquisition methods such as photographs, 3D scans, etc., and are

an intrinsic part the surveying process.2 During the last few decades, the recording

and management of documentation data has switched from paper-based archives

to databases and digital management systems.3 Similarly, documentation entities

that used to be analogue-based have increasingly been replaced by digital coun-

1. Szczepanowska 2013.
2. Drap et al. 2012; Kioussi et al. 2012.
3. Ravenberg 2012.

1



terparts. For example, photographs on glass-plate-negatives and slides have been

replaced by digital photography, or surveying maps drawn over layers of translu-

cent paper have been replaced by vector-based maps and geographic information

system (GIS) technologies. Most surveying activities, in fact, have been influenced

or replaced in one way or the other by digital technologies. Drawings are an im-

portant part of the survey process as their execution helps in understanding the

object.4 Computer-Aided Design (CAD) modelling or similar technologies are not

a substitute for quick sketching, as these take time and effort, and cannot be

quickly generated while surveying an artefact. For example, in archaeological

surveys, drawing is the last form of documentation still carried out by hand, even

if some projects have implemented the use of tablet computer technologies to

substitute paper as a medium.5 When not executed on a digital medium, the

drawings are scanned and linked as images within the databases. There are also

projects that have implemented retrospective conversion digitisation, which uses

digital drawing software to capture lines from the scanned drawings and transforms

these into vector graphics to allow for easy rescaling. 6

The drawing process is then an important surveying tool which has remained

virtually unchanged by the introduction of digital technologies. It is in essence an

interpretative process that is visually encoded, through which the essence of an

artefact, or its parts, is conveyed. These sketches are in general essentially diagram-

matic line drawings.

Diagrams, and images in general, are important cognitive and representational

tools for the human mind as they can be used not just for representation, but also

as reasoning and cognitive tools. Our brain, in fact, devotes more than half of its

processing power to vision.7 We have limited information processing capabilities,

but we are extremely adept at recognizing and analysing patterns, and visual pat-

terns in particular.8 When presented with visual information, this is immediately

4. Pickwoad 2004.
5. Warwick et al. 2009; Wright 2011.
6. Wright 2011. Vector images represent graphics by storing a series of mathematical equations

describing a graphic object; these can be modified through parameters, retaining a high level of
quality when transformed and scaled.

7. Lu & Dosher 2013.
8. Pizlo 2008; Terras 2008; Shin et al. 2013.

2



and integrally available to us, allowing us to perceive patterns and to compare

features between different percepts quickly.9

Diagrammatic representations can be defined as two-dimensional entities

composed of representing parts in mutual spatial and graphic relationships, which

are directly interpreted as relationships in the entity described.10 In other words,

diagrams are composed of elements that refer to real-world entities. These elements

are in relationship one with the other, and this network of elements mirrors the

structure of the real-world entities that are being represented. Hand-drawn dia-

grams hold a great deal of information for interpreting and categorizing the

structures there described, but they cannot be used as they are by a computer.11

Some of this information is unique, but, coupled with the information recorded

verbally, one finds a certain degree of information redundancy between the two

modes of communication. The same kind of information about an artefact is, in

fact, recorded within databases through verbal means as these guide the user to

careful and conscientious attention to the evidence. Verbal information stored

within databases has the added value of allowing for efficient and flexible retrieval

of the information.12 If the same essential information about an artefact is contained

in both visual and verbal representations, it might be possible to use computing

technologies to harness verbal information stored within a database and automat-

ically generate visual representations. By linking the two representations it would

be possible to present the information in graphic form to the user directly during

the survey, and diagrams could be generated on the fly whenever needed. In this

way, the information recorded verbally would also be available in graphic form.

This would provide a standardized visual access to the verbal information and

eliminate the need for redundant verbal and visual information, thus shortening

the description process but still providing the added values of ‘visual intuition’,13

and quick retrieval of verbal information.

9. Ware 2013.
10. Stenning & Lemon 1999.
11. Velios & Pickwoad 2005a.
12. Eiteljorg 2004; Ramsay 2004.
13. Greaves 2002; Sandywell 2012.

3



This thesis outlines a methodology to transform verbal descriptions of material

objects into visual representations automatically. In particular, the methodology

will be applied to the field of bookbinding studies. Bookbinding structures are

an ideal case study, as they range from simple to extremely complex and are,

therefore, a good testing ground for the proposed methodology. Moreover, the

archaeology of the book, the discipline devoted to the study of bookbinding

structures (amongst other aspects of the materiality of books), is rather young,

lacking in standards and rigorous methodologies, and it thus poses interesting

additional issues and challenges.

1.1. The cognitive importance of diagrams

Diagrams are important cognitive devices. Donald14 describes their importance

as memory storage devices, which he calls exograms as they reside outside our

minds. Their power resides in the fact that, by grouping information in a coherent

way, they allow the human mind to exceed its limited working memory capabilit-

ies.15 Working memory plays an essential role in our cognition, as it is in there that

we maintain all relevant information during the performance of a cognitive task.16

Data is naturally synchronous and simultaneous within diagrams, and this allows

them to present information already arranged and grouped, which makes it effort-

less to ‘chunk up’17 and to be given verbal labels that are easy to recall.18

Thanks to their ability to present information synchronously, diagrammatic

representations have been used as heuristic problem-solving tools for a long time

in domains that span from mathematics and geometry to architecture and engin-

eering, to mention but a few.19 Diagrams, in fact, exhibit expressive advantages

over sequential representations, as they can show effortlessly spatial and whole-

14. Donald 1991; Donald 2001.
15. Miller 1956; Cowan 2001; Saaty & Ozdemir 2003; Gobet & Clarkson 2004; Ericsson &

Moxley 2012. See also Appendix A.
16. Baddeley & Hitch 1974.
17. Baddeley 2013.
18. Brandimonte et al. 1992a; Brandimonte et al. 1992b; Brandimonte & Gerbino 1996; Mc-

Collough & Vogel 2007.
19. Shin et al. 2013.

4



Figure 1. Diagram of the interplay between working memory and the external memory
field embodied in exograms. The conscious mind sits in the middle of the two systems of
representation, one inside and one outside (after Donald 2001, fig. 8.3, p. 311).

part relationships.20 For example, maps are more helpful than verbal descriptions

of a landscape when it comes to navigation,21 or construction plans are the only

efficacious way of showing workers all the information needed to construct an

object designed by engineers.22

1.1.1. Diagrams as visual proofs

Diagrams have also exhibited uses as verification tools and visual proofs. Geo-

metry is but one example of a domain that historically has made ample use of

visual languages as proofs and verification systems; these are deductive techniques

that exploit the topological and spatial features of diagrams and human spatial

intuition. Intuition here means a non-deductive method by which we can infer

20. Allwein & Barwise 1996; Barker-Plummer 2002; Shin et al. 2013.
21. However, there are kinds of geographical information that can be encoded in texts, but

which are impossible to express faithfully in maps (Eide 2012a).
22. Ferguson 1992.

5



classes of truths about the world.23 Diagrammatic proofs, in fact, are linked to our

ability to manipulate intuitively and interpret spatial relationships in diagrams,

and in associating these with truths about the world.24 Chapter 7 shows a possible

use of the diagrams generated for this projects as data verification tools, linking

the diagrams that have been automatically generated with the object represented.

1.2. Problems with the adoption of diagrams

Notwithstanding their undisputed usefulness, diagrams create a series of

problems. One criticism that is often brought forward is that diagrams are cum-

bersome to reproduce or communicate.25

As mentioned above, hand-drawings are also not directly usable by a computer,

and they have to be interpreted and provided with relevant metadata during data

inputting by a human user.26

Also, they often hold redundant information, already secured through verbal

means within the databases. Redundancy of information is not a problem in itself,

as it is delivered through different means of communication; however, this does

impact on surveying time as the same information has to be recorded twice during

data input: once verbally and once graphically.

The proposed methodology aims to solve the problems of the generation, digit-

ization, and reproduction of diagrams of bookbinding structures. This will allow

automated diagrams to be promptly available during the surveying process, and

for them to be effortlessly adopted as heuristic and communication tools, exploiting

their intuitive spatiality and immediacy.

23. Greaves 2002.
24. Greaves 2002; Shin et al. 2013.
25. Hammer 1995; Greaves 2002.
26. Velios & Pickwoad 2005a; Rains et al. 2014.

6



1.3. Verbal to visual: intermediation & intersemiotic translations

Going from verbal descriptions to visual graphic depictions involves two rather

different information media. The relationship between information from different

media is referred to as intermediality.27 The passage of meaning between media

is only one of the many aspects of intermediality, as this also express different

experiences and narrations — i.e. how we experience, read, and interact with

each medium.28 This project is however focussed on the possibility of the passage

of information from verbal descriptions to graphic representations, and it will

only touch on the considerations of the different experiences that the two different

media bring with them for the end-user.

The main concern here is the way in which a medium whose ‘semantic modal-

ity’29 is characterized by symbolism — i.e. relying mostly on arbitrary signs — can

be transformed into another medium whose ‘semantic modality’ is instead ex-

pressed mostly via iconism — i.e. relying on sign that resemble their object in

some respect.30 This is a problem of sign interpretation, or more accurately, a

translation problem. Jakobson31 distinguishes between three ways of interpreting

a verbal sign: (i) intralingual translation or rewording, (ii) interlingual translation

or translation proper, and (iii) intersemiotic translation or transmutation. The

first case takes place when a verbal sign is translated into other signs of the same

language; the second, when the verbal sign is translated into another verbal sign

belonging to a different language; the last case refers to the verbal sign being

translated into another sign belonging to a non-verbal system of symbols.

Intersemiotic translations deal with two different semiotic codes, turning

meaning from one expression code into an entirely different one and present,

therefore, complex issues. The automated transformations proposed by this re-

search are a clear case of intersemiotic translations. This research will concentrate

27. Elleström 2010.
28. Elleström 2010; Ryan 2014.
29. Elleström 2010.
30. See chapter 3.
31. Jakobson 1959.

7



on the mechanics and procedures of intersemiotic translations of verbal descrip-

tions of bookbinding structures into graphic representations.

1.3.1. The essential elements of the communication cycle

Looking at the project at hand it is possible to delineate a cycle which the rel-

evant information should follow for a transformation from verbal to visual to be

successful. In analysing the communication cycle32 in more detail it can be seen

how the project deals with a series of entities, processes, and agents. Specifically,

there are three entities at play: the object being described — i.e. the source of the

cycle — the words that are use to describe it in a verbal description, and the image

used to represent it — i.e. the different message channels of the cycle. There are

also two main processes: the perception and the encoding and communication of

the relevant information. In addition, looking specifically at this project, there is

a third process that needs to be taken into account: that of the transformation of

information from a verbal encoding into its visual counterpart. Finally, there are

two types of agent: a human agent that is both perceiving and communicating the

information, on the one hand, and on the other receiving and interpreting the

perceived information; and, looking specifically at this project, a computer agent

set to transform a verbally encoded message into a diagrammatic representation.

Figure 2[a] shows the diagram of a generic cycle going from verbal to visual, and

Figure 2[b], the cycle specific to this project.

1.4. Bookbinding descriptions and the generation of diagrams

As it will be seen in chapter 4, bookbinding studies lack a systematic vocabulary,

terminological clarity, and a precise recording system; in a similar way, they also

lack graphic representation standards. Description of bookbinding structures are

32. Rothwell 2012.

8



Figure 2. Communication cycle from the object being described — the source — to the
human receiver/interpreter showing the three entities (object, words, images), the three
processes (perception, communication, transformation), and the two types of agents in-
volved (human being, computer).
(a) A cycle comprising of a verbal description and/or a visual depiction as message
channels. The line from the verbal description to the perception process leading to the
human receiver/interpreter has been dotted because, although strictly speaking there is
still perception involved in the recognition and processing of written code (Dehaene
2009), its impact on the decoding and understanding of the message is limited by the
highly symbolic nature of the medium.
(b) The particular communication cycle of a process whereby a verbal description is
transformed by a computer into a visual description, which has then to be interpreted
and understood as a proper description of the original object by a human receiver.
Note the feedback loop checking for the effectiveness of the description of the source
material.

9



therefore often ambiguous, and this hinders comparison, sharing of information,

and, inevitably, progress in the field.

Diagrams and sketches are often used as cognitive devices when describing

bookbinding structures because of their innate capacity of showing effortlessly

spatial and whole-part relationships. They facilitate communication, to the point

that purely verbal descriptions of bookbinding structures are often completely

inadequate as communication devices.33 However, to this author knowledge,

bookbinding diagrams have never adopted methods of uncertainty visualization,

leading to the impossibility of scholarly transparency in graphic representation

of bookbinding structures.

As in the case of the documentation of artefacts in other fields within the cul-

tural heritage sector, also databases devised to describe bookbinding structures

often contain both iconographic and textual information, following the usual

practice of accompanying words with visual representations for easier communic-

ation, thus creating a redundancy or overlapping of information.

This project will aim at exploiting that redundancy, investigating the possibility

of automatically generating bookbinding structure diagrams from verbal descrip-

tion within databases. As it will be seen, it is possible to generate such diagrams,

and these, being standardized, allow easier comparison amongst similar structures

in different books. The fact that these are automated diagrams, also permits their

use as visual proofs, thus leading to better data being recorded within databases.

In addition, thinking in terms of intersemiotic translations, permits to identify

efficient description modalities that can inform any bookbinding description

methodology, whether automated diagrams are included or not.

The contributions of this research to the field of bookbinding studies offer a

way to improve communicability of information and scholarly transparency, thus

fostering the progress of the discipline.

33. See §4.2.1-2.

10



1.5. Terminological issues

It is important to clarify three similar, but separate concepts: shape, form, and

structure. Shape refers to ‘those geometrical characteristics of a specific three-di-

mensional object that makes it possible to perceive the object veridically from

many different viewing directions, that is, to perceive it as it actually is in the world

out there.’34Form refers to the structural organization of material objects, i.e. the

configuration or arrangement of the components of a material object; in chapter

3, a further specification of form will also be introduced, referred to as logical

form. Both these terms are clearly synonyms of structure, however, the term

structure is used only in reference to bookbinding structures — e.g. a particular

kind of endband, a sewing pattern, etc. — to thus avoid having to speak of ‘the

structure of a binding structure’ creating unnecessary confusion; such a concept

will thus be expressed as ‘the form of a binding structure.’

In this project, the terms diagram, diagrammatic drawing, or diagrammatic rep-

resentation will be taken to signify an outline drawing of a bookbinding structure

showing parts and their relationships.

Finally, the terms verbal and visual are intended throughout this thesis as cog-

nitive psychological coding modalities. The former refers to any word-based

communication or cognition process, often in contrast to things, realities, visual

information, or to the fact that some information pertains to, or is manifested in

words. The latter, instead, is related to sight and vision, what can be obtained

through vision, what is carried out by vision, or what is produced or occurs as a

picture in the mind.35

34. Pizlo 2008, p. 1.
35. Neisser 1967/2014; Paivio 1971; Paivio 1986; Oxford English Dictionary 2015a; Oxford

English Dictionary 2015b.

11



1.6. Aims and objectives

The overall aim of the project is the generation of an automated diagrammatic

visualization of bookbinding structures based upon structured verbal descriptions

contained in a database.

Because these transformations are based on a model of the structures, some of

the problems encountered by the examples in the last section will need to be

considered in relation to the fact that, as it will be seen, bookbinding descriptions

do not hold spatial data or detailed measurements. Amongst these problems are

the placements of the components in space, and the specificity of the objects

(measurements and shape).

In order to generalize a verbal-to-visual approach for material objects, one

needs to be able to:

identify the minimum amount of information that is needed for a

visualization of bookbinding structures, or other material objects. A

successful visualization is one that is able to convey the minimum

information required for the object represented to be recognizable.

assert whether a model of bookbinding structures can indeed lead to

successful visualizations.

In addition to these, as noted in chapter 1, diagrams are often used as visual

proofs. Considering that the visualization approach proposed in this project

gathers data directly from a database and is therefore stricly connencted with the

data recorded there, one can:

assess whether automated diagrams of bookbinding structures can be

used as data verification tools.

It also useful to set a series of practical objectives for the generation of successful

automated visualizations of bookbinding structures:

12



Selection of bookbinding structures to be visualized from the information

available within the Ligatus schema.

Selection of the characteristics of the appearance of the visualizations,

taking into consideration perception, cognitive psychology,

communication theories, and graphic conventions.

Identification of shapes for each transformation, their visual and formal

characteristics.

Identification and selection of the most useful and appropriate

technologies for the generation of the automated visualizations.

Identification of suitable strategies to visually communicate the various

degrees and typologies of uncertainty and, where necessary, imprecision

inevitably contained in the data to be visualized.

Compilation of coding to transform the verbal information on

bookbinding structures contained in the XML data into visuospatial

information integrated with befitting verbal labels.

1.7. Chapter overview

This introduction has introduced the fundamental concepts behind the starting

point of this project. Diagrams are an important source of information. They can

convey spatial and parthood36 information better, and in more immediate ways

than sequential communication systems. Also, during surveys, some information

is collected redundantly both verbally and visually, but it is more easily absorbed

through visual communication systems. It is, therefore, valuable to investigate the

possibility of automatically transforming verbal information into graphic repres-

entations automatically, saving time and augmenting its possible uses. This kind

of automated intersemiotic translation system requires the understanding of both

modes of communication, their perception, and the mechanics of the transfer of

36. In philosophy, the relational quality of being a part. (Varzi 2014).

13



meaning from one sign-language to the other. This work will look at the entities

and communication forms involved, the translation process, and the application

of automated intersemiotic translations to the field of the study of bookbinding

structures.

Chapter 2 considers the process of modelling reality, including material objects,

and also considers previous projects that aimed at transforming information

automatically from verbal to visual. It then sets a series of primary objectives for

this investigation.

Chapter 3 looks into the nature of the various languages and entities involved

in the project — words/verbal and images/visual — and their nature as signs

bearing meaning. It also covers the fundamental processes of perception and

prototypification,37 while considering what is the minimum amount of essential

information for the communication of a material object.

Chapter 4 provides an overview of the various kinds of verbal and graphic de-

pictions of bookbinding structures found in the literature. These are categorized

and analysed in reference to their efficacy in delivering their message.

Chapter 5 discusses verbal and visual representations of material objects, and

the translation between languages. It also investigates the nature of visual commu-

nication devices and the expression of the uncertainty of information.

Chapter 6 introduces the technological aspects of the automated transformations,

including the database description schema, and the dataset to which the transform-

ations have been applied. Visualizations issues and considerations are explored

in relation to the automated diagrammatic representations of bookbinding struc-

tures.

Chapter 7 examines in detail the transformation of endleaf structures and intro-

duce generally all other transformations of bookbinding structures. It then con-

siders the human factor in the data input process and the errors in the dataset.

37. In cognitive psychology, prototypification is a general feature of conceptualization, a process
through which conceptual and perceptual prototypes are formed and elaborated in our mind.
(Rosch 1978; Leyton 1987; Bossche et al. 1992).

14



Chapter 8 presents a set of recommendations for the practical task of tranforming

information from a verbally encoded dataset to series of meaningful and scholarly

sound diagrams.

Chapter 9 draws conclusions from the project and declares what contributions

were made in the course of this research, whilst Chapter 10, explores related and

future work.

The Appendices cover an overview of the human memory system, a detailed

examination of all the shapes and elements developed in the course of the project

for the generation of the automated diagrams of bookbinding structures, and an

example of the coding involved in one transformation.

15



Chapter 2. Modelling & related work

Each language speaks the world in its own ways. Each edifies

worlds and counter-worlds in its own mode.

George Steiner, Real presences. Is there anything in what we

say?, 1989/2010, p. 63.

When we want to study a material object we initiate an abstraction, categoriza-

tion, and modelling process.38 This process allows us to select only a restricted set

of its characteristics, and to work with these models of reality. Reality, in fact,

varies from instance to instance and has an unwieldy amount of detail.

When we want to communicate a description of such an object, we use a similar

strategy: we can use representations or models of reality that convey a restricted

set of characteristics of the object for which they stand.39 These models of reality

can take various forms, each of which can be used for different purposes.

This chapter looks at the different kinds of modelling and communication sys-

tems that are involved in this project. It also defines where this project stands in

relation to other projects and methodologies that go from verbal to visual.

2.1. Modelling

In order to study and communicate research results, a common and essential

academic activity is that of modelling the reality that needs to be described.

38. Cohen & Lefebvre 2005.
39. Sebeok 1994.

16



Through a process of abstraction, the researcher, considering which research

questions need answering, determines what aspects of the real world need to be

included, and the granularity and level of detail the final model will need to have.

Conceptual modelling requires decisions and assumptions in regard to the scope

of the model, its level of detail, the nature of the reality to be described, and the

simplifications to be made.40

The modelling activities take different forms. Amongst these forms McCarty41

lists analogy, representation, diagrams, and maps. Maps are defined here as any

schematic spatial representation.42 Each of these forms can be either analogical or

digital, and their boundaries can be blurry, one model being possibly associated

with more than one form.

When dealing with material objects, the modelling activities can take any of the

forms listed. (i) Analogies and representations, as there needs to be a relationship

between the model and the artefact, so that by studying the former one can infer

facts about the latter. (ii) Diagrams and maps, for the fact that models represent

the structure, spatiality, and relationships between the essential parts of the arte-

facts.

Typically, analogy and representation models of material objects take the form

of descriptive or database schemas. Databases are often used to record information

on bookbindings, especially in the field of conservation. 43 Also, in archaeology,

classics, and libraries, databases have been used for data management since the

1970s.44

Diagrams and maps are used to represent material objects. Because of the im-

mediate way of conveying spatial and complex structural information, diagrams

are considered as an integral part of record keeping.45

40. Kotiadis & Robinson 2008.
41. McCarty 2005.
42. McCarty 2005.
43. Ravenberg 2012.
44. Eiteljorg 2004; Terras 2008.
45. Eiteljorg 2004.

17



2.1.1. Modelling of and modelling for

As mentioned above, a model is the result of decisions and assumptions in re-

gards to its scope.46 Considering the final scope of the modelling process, scholars47

distinguish between two different activities: the modelling of reality, and the

modelling for something. McCarty48 defines the former as ‘a representation of

something for the purposes of study’, and the latter as ‘a design for realizing

something new’. In the first case, the model, denotative and descriptive in nature,

renders reality and its physical relationships apprehensible and serves the scope

of developing a theory of such a reality; it tends towards the general. One can

think of models of as a sort of recipe that describes how an object can be composed

— e.g. a table composed of a surface and a set of legs. Different instances of tables

may have different number of legs, but by referring to the same model/recipe,

one can compare each instance with the rest. In the second case, the theory behind

the modelling process is exemplary in nature and guides the manipulation and

organization of entities and their relationships for the creation of something new;

it tends towards the particular.49 By nature, both kinds are simplified representa-

tions. Modelling of is in reference to an idealized reality and helps to describe and

understand it. Modelling for, instead, aims at the creation of new entities as

heuristic and pragmatic instruments of investigation (often through manipulation).50

The two types of modelling activities can then be regarded as distinct, each

with specific conceptual characteristics. This, however, does not mean that one

cannot turn into the other, often in a cyclic system, that aims at gaining better and

more detailed knowledge of the reality being modelled.51

Database design is a typical example of a modelling of process. Diagrams and

maps can be either models of or models for. An architectural plan, for example,

is a diagram/map that is used to generate something new, and is, therefore, an

example of modelling for. However, if the same plan was, in fact, the result of the

46. Kotiadis & Robinson 2008.
47. Goodman 1976; Geertz 1973; McCarty 2005.
48. McCarty 2005, p. 24.
49. Goodman 1976; Geertz 1973; McCarty 2005.
50. McCarty 2005.
51. Geertz 1973; McCarty 2005.

18



conceptual reconstruction of the layout of a historical building, based on archae-

ological evidence, then that plan, in the first instance, would be a model of. The

project at hand, as it will be seen, deals with models of.

2.2. From verbal to visual: related projects and approaches

This project aims to transform automatically verbal descriptions of bookbinding

structures to visual representations. Also, as there is not a highly structured and

symbolical way to represent bookbinding structures in diagrams, the visual rep-

resentations necessarily need to resemble the object that they represent. Therefore,

this research needs to consider the transformation of information from verbal to

visual, the implications of its automatism, and the iconicity of its output.

In the literature, one encounters a range of projects, standards and established

practices of modelling and design based on verbal input to generate parametrically

visualizations of objects by algorithmic means; although none of these projects

considers bookbinding structures.

2.2.1. Hardware description languages

One example is that of the Hardware Description Languages (HDL) used to

describe the structure of Circuit Boards (CB) to then automatically program their

design and function. HDLs are textual modelling languages based on standard

high-level text-based computer-interpretable expressions that describe the structure

of an electronic circuit board. This modelling is both physical, as it relate the

various components in space, and functional, as it can mimic the interactions

between the components and the final functions of the circuit. The circuit is then

virtually tested and eventually physically printed.52 Before the design is programmed

through HDLs, the engineer has worked a higher-level model of the circuit, and

it is this model that is then translated into the CB. This makes the HDL modelling

52. Mermet 1993; Ullman 2010.

19



stage clearly a modelling for activity, whose purpose is not that of study of the

principles of circuits in general, but the manufacturing of a specific functional

circuit. Also, in this example, the placement of items on the plane is guided by

logic: the aim is to put in the desired circuit components using the least amount

of space and in the most efficient way. The reference model is the idea of the circuit

devised by the engineer, not an object.

2.2.2. Natural language processing

Since the late 1980s, a number of Artificial Intelligence (AI) studies have focused

on the interpretation of verbal descriptions in natural language for the generation

of visual scenes. In 1984, Adorni and colleagues53 developed a strategy for depicting

a static scene described by means of a sequence of predefined phrase-forms. Bijl54

advances a methodology that looks at integrating AI studies with Computer-Aided

Design (CAD) for the structuring of information and recovery of construction

semantics from natural language. A concept then revisited by Coyne and col-

leagues55 and extended in order to express the importance of functional knowledge

in design from natural language.

Coyne and Sproat56 presented a system to convert text to representative scenes

automatically. Their system was aimed at the general public to allow scene mod-

elling without having to deal with programming languages. Their software is re-

stricted to its own database for semantics and objects.

Johansson and colleagues57 advanced a method to convert (Swedish) narratives

describing road accidents into graphic scenes. Texts are analysed by a natural

language processor and converted into a symbolic representation of the scene that

highlights nodes (objects), relationships, quantities, events, and the environment.

Symbolic representation works in such a way that on the one hand it captures

53. Adorni et al. 1984.
54. Bijl 1986.
55. Coyne et al. 1990.
56. Coyne & Sproat 2001.
57. Johansson et al. 2004; Johansson et al. 2005.

20



enough information for visual modelling and, on the other hand, it is close to the

way in which a human would read and describe a scene. Environmental informa-

tion aids spatial framing of the objects and their movements.

Zhu and colleagues58modelled natural language processing in a similar way to

text-to-speech synthesis to convey the gist of texts. Their system basically recog-

nizes words denoting objects and actions, then looks for pictures that can offer

the desired meaning. The images are then presented in a loose structure to the

viewer (see Figure 3 for an example).

Figure 3. Example of Text-to-Picture transformation (from Zhu et al. 2007, p. 1590).

More recently, Claude-Lachenaud and colleagues59 have presented an algorithm

designed to visualize the movement of objects expressed in text sequences through

verbs. By means of natural language processing, action words are detected, and

movement identified in verb-to-motion tables. Trajectories that could be defined

as random towards a general direction — e.g. to walk: onward generic direction;

to climb: random vertical motion — were interpreted correctly by the system.

A common problem for all of these projects is the ambiguity of spatial inform-

ation encoded within natural language. Motion verbs may only indicate general

direction unless some kind of reference frame is provided as in the case of car

58. Zhu et al. 2007.
59. Claude-Lachenaud et al. 2014.

21



accident descriptions.60 Spatial prepositions are often ambiguous too, as each can

convey a series of different meanings.61 Consider, for example, the different spatial

meaning of the preposition on in the following sentences: (i) the book is on the

table; (ii) the painting is on the wall. In both cases, the preposition indicates contact

with some surface, but the surface is horizontal in (i) and vertical in (ii).

Similar problems are highlighted in a recent project that has investigated the

passage of information from verbal description of landscapes to maps. Eide62

considers what kind of information can be transformed and classifies textual

spatial expressions in ‘fully specified’, and ‘underspecified’ ones. A fully specified

text is one from which only one map can be drawn: every entity is fully specified

geometrically. Examples of these are texts in formal languages, such as GML. On

the other hand, more than one map can be drawn from an underspecified text,

and some of these maps can be significantly different. The under-specification

derives from the fact that natural language does not indicate direction and spati-

ality in absolute terms. For example, East of some place does not mean at 90°

from the point of observation, but rather a range of possible locations roughly in

an eastern direction.

These projects make use of models for the generation of their visual output,

but in the first step, they generally model the reality they are dealing with as

models of in order to be able to extract relevant information from texts. Examples

of these model of are the semantic modelling at the basis of Eide’s investigation,

or the symbolic scene representation found in Johansson and colleagues.

2.2.3. Cultural heritage modelling

There have been applications of semantic modelling — i.e. verbal-based mod-

elling — applied to the cultural heritage field. The aim of these projects is to allow

for modelling without the designer having to learn how to use complicated mod-

60. Johansson et al. 2004; Johansson et al. 2005.
61. Boggess 1978; Boggess 1979; Herskovits 1980; Adorni et al. 1984; Ferrier 1996; Garrod

et al. 1999; Tyler & Evans 2003; Lockwood et al. 2006.
62. Eide 2012a; Eide 2012b; Eide 2013; Eide 2014.

22



elling suits suck as Computer-Aided Design (CAD), three-dimensional (3D)

modelling software, or similar.

Stilla and Michaelsen63 resort to knowledge representation through associative

networks64 to allow experts to model man-made objects found in aerial photo-

graphs. Complex shapes are described as a series of primitive elements combined

together through logical operations. The modelling activity is focussed on the

production of two-dimensional (2D) shape renderings of the buildings found on

the photographs.

De Luca and colleagues65 describe a methodological approach to the modelling

of classical buildings based on verbal inputs and the mapping of photographs and

3D scanning data onto the semantically produced model. A similar project, fo-

cussed on the semantic modelling of vernacular buildings of South-east China, is

presented by Yong Liu and colleagues.66 These projects aim to decrease the com-

plexity and workload of the modelling task for the designers, by utilizing67 or de-

veloping68 a formalized description language. Formal language allows one to in-

crease the modelling level from the usual basic graphic units (points, lines, triangles,

curves, etc.) to a higher level of semantic elements (walls, windows, roofs, houses,

streets, etc.) which are intelligible by humans, and interpretable by a computer.

It is interesting to note the fact that De Luca and colleagues base their semantic

interface on the highly formalized language of classical architecture: in particular

the works by Palladio and Vitruvius.69 The technical language of classical architec-

ture is well equipped to describe components and their spatiality.70 In chapter 5,

this will be analysed in more detail.

During modelling, the user inputs data verbally into the system that in turn

transforms it immediately into a visual representation. Thus, the parameters

defined through verbal input are modified until the desired output has been

63. Stilla & Michaelsen 1997.
64. Findler 1979.
65. De Luca et al. 2005; De Luca 2013.
66. Yong et al. 2006.
67. De Luca et al. 2005; De Luca 2013.
68. Yong et al. 2006.
69. De Luca et al. 2005; De Luca 2013.
70. Goulette 1999; Borillo & Goulette 2006.

23



produced. Measurements are provided by direct input in Yong and colleagues,

whereas in De Luca and colleagues the model is adjusted and registered to fit the

data from the 3D scans and the digital images of the building. The modelling

activities in these projects are for the generation of specific models.

Another approach to produce large scale visualization efforts utilized in the

cultural heritage sector is that of procedural or parametric modelling. This ap-

proach allows for rapid prototyping, i.e. the generation of series of models all as-

sembled at random on the basis of specific prototypical models and parametric

ranges of values. Müller and colleagues,71 expanding on the procedural shape

grammar advanced by Stiny,72 have developed a system to build large architectural

models with significant geometric detail. Dylla and colleagues,73 have applied

procedural modelling to rebuild virtually ancient Rome at the height of its urban

development in 320 AD. The placement of the buildings is derived from Geo-

graphic Information System (GIS) data. Saldana74 investigates the value of the

modelling process behind the outcomes of the subsequent rapid prototyping as

an important scholarly activity and as an information-rich source for each randomly

designed model. This modelling process is prototypical in nature as it describes

types of buildings rather than specific buildings, and it is implemented for the

generation of a visualization.

In all these examples, the models on which the automated transformation from

verbal to visual are based are models for the generation of the desired graphic

representation. These models represent an extra layer of modelling, separated

from survey database data, that has been conceptualized specifically for the sole

scope of generating those graphic outputs from verbal or high-level inputs, and

they have done so successfully.

71. Müller et al. 2006.
72. Stiny & Gips 1972; Stiny 1980; Stiny 2006.
73. Dylla et al. 2010.
74. Saldana 2014.

24



2.3. From verbal to visual through models of

The novelty of this project does not lie strictly in the fact that it advances a

methodology for going algorithmically from verbal to visual - although this does

lead to implications and considerations that are specific to this methodology - but

rather in the fact that, unlike other cultural heritage projects, it uses a model of

and not a model for as its starting point. In other words, instead of relying on ad-

ditional information specifically introduced, according to a model for, for the

generation of the graphic output, this research project will take the information

directly from a survey database. The approach is not just technically, but also

epistemologically different: it relates to the selection and processing of information,

and to the nature of the outcome. As will be seen, in fact, the diagrams generated

through this methodology are necessarily prototypical in nature; this renders them

also graphic models of the object depicted, thus facilitating comparison, analysis,

classification based on visual clues, and, to an extent, memorization. In addition

to this, no other project has tried to produce automatically drawings of bookbind-

ing structures.

A typical problem of words-to-image transformations is that of the spatial rela-

tionships between the various components. In the cultural heritage examples ex-

posed above, the focus of the modeller is the production of two- or three-dimen-

sional models utilizing semantic parameters and descriptions in order to visualize

a specific object. The typical visualization problems of the placement of elements

in space and of their dimensions are solved by turning to additional quantitative

data (GIS data, 3D scan data).

Another problem is that of the shape of the objects to be represented and their

specificity. In the Rome 2.0 project,75 specific parts that are known because they

still exist are manually designed using parameters such as exact measurements

and texturing with photographs, whilst the unknown parts of the buildings are

reconstructed according to a generic model of that type of building, using the

existing parts as a frame of reference for placement, measurements, and shape.

75. Dylla et al. 2010.

25



De Luca and colleagues,76 as mentioned, guide the shapes of each architectural

component model on the basis of its definition. They can do so because the

technical vocabulary developed for classical architecture is a highly structured

and granular verbal way of representing atomic visual features. For example,

column outlines are subdivided into the smallest unit necessary to reproduce

them: torus,77 listel,78 scotia,79 etc. (see Figure 4). The generic model thus obtained

is subsequently adjusted and registered by the modeller to fit the quantitative data

gathered through direct acquisition methods (3D scanning, photography).

Figure 4. Example of classical architecture column parts (from De Luca et al. 2005, p.
3).

Bookbinding studies lack a highly formalized technical terminology that aims

at describing each atomic visual feature of a structure.80 Models of describe proto-

typical instances of structures, and are, therefore, not concerned with specific

measurements of an item. Similarly, spatial positioning of components is only

76. De Luca et al. 2005; De Luca 2013.
77. Torus: convex, semi-circular moulding often at the base of columns (Encyclopædia Britan-

nica 1911).
78. Listel: A small list or fillet (Oxford English Dictionary 2014b).
79. Scotia: concave moulding with a lower edge projecting beyond the top and so used at the

base of columns as a transition between two torus mouldings with different diameters (Lewis &
Darley 1986).

80. See chapter 4.

26



suggested by the explication of the relationships between items — e.g. A is part

of B. All of these problems, it would seem, impede the successful generation of

diagrammatic visualization of bookbinding structures, or other material objects

given the same procedural constraints, when compared with the formal application

of semantic modelling in the examples given above.

Notwithstanding this, models of are specifically developed to describe reality

as an object of study, and by generalizing and categorizing it allow to compare

examples. If this information can be harnessed and transformed into diagrams,

these too can be used as visual models of.

There is a lot of experience that can be gathered from the examples above, even

if the means are different. All of the above-mentioned projects have been successful

in their endeavours. However, none has analysed what it is exactly that makes the

project a success, what are the minimum requirements for a successful intersemi-

otic translation from verbal to visual of a material object.

This project aims to generate diagrammatic visualizations, directly from the

data included in databases, and not through a second layer of modelling that

guides a modeller in the production of diagrams. Also, for bookbindings, there

is not a highly structured and granular terminology that can be used in the same

way as classical architecture vocabulary. Because of all of these reasons, the exper-

ience accrued in the semantic modelling project presented above, would need to

be adapted and generalized.

In addition to this, the visualizations for this project are not based on an extra

layer of modelling with information specifically selected for the generation of the

graphic outputs, and this means that there is a direct link between the data recor-

ded verbally in databases, and the automated diagrams. This way, the diagrams

are associated with the object that they represent indirectly and through the de-

scription of it in the database. As mentioned in chapter 1, visualizations have been

used in a number of fields as verification tools. It is, therefore, interesting to in-

vestigate whether this relationship between data and visualizations can be harnessed

and the automated diagrams used as data verification tools.

In order to be able to adapt and apply a similar transformation from verbal to

visual to another field, such as bookbinding studies, one needs to understand

27



what kind of information is essential. To generalize and understand the issues

linked with an intersemiotic translation from a verbal model of a material object

such as a binding structure, into a diagrammatic representation, one needs to

consider the nature of the two languages involved and of the items to be repres-

ented, and also issues linked with human perception and cognition, since the in-

tended user of the diagrams are human beings. Through this kind of analysis, one

can advance a general methodology applicable to any material object.

Summary

Two main categories of modelling activities can be distinguished: modelling of

and modelling for. The former is focussed on discerning the essence of a material

object, whilst the latter aims at the generation of a particular entity. The former

looks at the general, the latter at the particular.

Verbal to visual automated transformations in the cultural heritage field have

generally been based on modelling for methods, making use of information on

the spatial configuration of the components, their measurements, and their shape.

None of these has considered bookbindings and none of these relies on information

already recorded within databases. Instead, they develop an extra layer of model-

ling, and typically gather more data, specifically for the purpose of generating a

graphic output. In these projects, qualitative data (verbal input) is often mixed

with quantitative data (measurements, scans, or other direct acquisition methods)

to solve the problems of indeterminateness found in verbal descriptions. Also,

none of these projects has attempted to developed a general methodology to as-

certain the minimum amount of information that is required for successful verbal-

to-visual transformations for material objects in other fields. It has not been ad-

dressed whether it is possible to generate successful transformations based on less

specific, but more prototypical information contained in a model of, using data

already gathered whithin databases, and whether these visualization could be used

as data verification tools.

28



Transformations based on models of need to address the problems of the spatial

configuration of the components and their specificity in different ways. To under-

stand how this is possible, the next chapters will investigate the nature of the

various entities and processes involved in the project, then to apply these trans-

formations to the field of the study of bookbinding structures.

29



Chapter 3. Signs & material objects

We should, however, recall that our mind can be stimulated by

many things other than images — by signs and words, for ex-

ample, which in no way resemble the things they signify. [...] It

is enough that the image resembles its object in a few respects.

Indeed the perfection of an image often depends on its not resem-

bling its object as much as it might. You can see this in the case

of engravings: consisting simply of a little ink placed here and

there on a piece of paper, they represent to us forests, towns,

people, and even battles and storms; and although they make us

think of countless different qualities in these objects, it is only

in respect of shape that there is any real resemblance. And even

this resemblance is very imperfect, since engravings represent to

us bodies of varying relief and depth on a surface which is entirely

flat. Moreover, in accordance with the rules of perspective they

often represent circles by ovals better than by other circles,

squares by rhombuses better than by other squares, and similarly

for other shapes. Thus it often happens that in order to be more

perfect as an image and to represent an object better, an engrav-

ing ought not to resemble it.

René Descartes, Discourse IV of The Dioptrics, 1637.

At the end of the introductory chapter, reference to the communication cycle

for the project was made, highlighting the various entities and processes that come

into play. Communication can be defined as the exchange of any kind of message

30



and of the system of signs of which it is made.81 Signs then are the basic units of

communication. To understand what a sign is, one can begin from a general

definition: anything that stands for something else in certain respects.82 It is through

the process of standing for that meaning is created.

This chapter looks at the communication process through signs, and how

meaning is conceptualized in the human brain. It also investigates the nature of

material objects and what is the mininum amount of information needed to

communicate their essence.

3.1. Signs, signification, and communication

Semiotics, the science of signs, studies signs in processes such as communication,

cognition, and linguistics.83 There are two main traditions of semiotics. One,

started by the linguist Ferdinand de Saussure (1857–1913), proposes a dyadic

notion of signs, whereby every sign consists of two parts: the signifier, a sound,

word, or image, and the signified, the mental concept to which the signifier is re-

lated. Whilst Saussure admits that signs can be other than words, his theories are

mostly concerned with how meaning is created through words and conventions,

arbitrary signs. Another tradition, coming from the investigations of the logician

Charles Sanders Peirce (1839–1914), proposes instead a triadic theory of sign, with

each sign consisting of three parts: a sign-vehicle, an object, and an interpretant.

Sausurre’s tradition, usually followed by linguistic studies, has been criticized

by scholars interested in a wider notion of sign — asides written or oral language

— for offering a rather static notion of signs and focussing only on systems of ar-

bitrary signs.84 Therefore, many semiologists, while acknowledging the fundamental

understanding of the structure of signs in Saussure’s theories, prefer to turn to

Peirce’s richer typology of signs, thus allowing for different modes of signification,

81. Sebeok 1991.
82. Hoopes 1991.
83. Moriarty 1994.
84. Iversen 1986; Bryson 1991; Alpers et al. 1996.

31



not just arbitrary signs.85 Given the issues of the project at hand, the theories set

by Peirce are of interest. However, Saussure’s dyadic (signified-signifier) relation-

ships are included in Peirce’s categories.

3.1.1. Signs

Let us consider briefly the basic concepts of Peirce’s triadic theory of signs. A

sign consists of three interrelated parts: a sign-vehicle, an object, and an inter-

pretant.86 Figure 5 shows the relation between the three elements, as customarily

schematized as a triangle. The sign-vehicle is the signifying element, or signifier,

e.g. a word; the object indicates what is being signified, e.g. the object to which

a word attaches; the interpretant, the most significant and distinctive feature of

Peirce’s theory, is best described as the understanding that one has of the

sign/object relation, thus, since the signification is not simply a dyadic sign-object

relationship, a sign has a signification only for the fact that is being interpreted:

the meaning of a sign is made manifest through the interpretation that it generates

in the sign users. Further, a sign is determined by the object it signifies, as the

object imposes certain parameters that the sign must follow in order to represent

that object, but not every characteristic possessed by an object is relevant to signi-

fication and the process of sign determination.87

In other words, the object determines the sign by imposing some constraints

that the sign ought to meet in order to signify it, and the sign signifies the object

only by virtue of certain features possessed by the object; that is to say that the

nature of the object itself limits its sign’s nature in terms of what is required for

successful signification in a selective abstraction88 process. The sign also determines

85. Rose 2012.
86. Peirce, and many scholars after him, call the three components of a sign: sign, object, and

interpretant. This particular nomenclature has the potential of creating confusion for the fact that
it appears to be that of the three elements of a sign, one is the sign itself (leading to the misconcep-
tion that a whole is in fact a part of itself). For this reason, Atkin (Atkin 2010) refers to that element
of the sign responsible for signification as the sign-vehicle; the same convention is followed here.
Other terms used by Peirce for the signifying element are: representamen, representation, and
ground.

87. Short 2007; Atkin 2010; Burch 2010.
88. Greenlee 1973.

32



Sign-vehicle Object

Interpretant

Three Types of Signs

Icon Index Symbol

Figure 5. The basic structure of signs and its trichotomous nature according to Peirce.
Note that a sign may display a combination of iconic, indexical, and symbolic character-
istics.

the interpretant by focusing the attention on some features of the sign/object re-

lationship.89

According to Peirce, the object then determines its sign by dictating some

constraints for a successful signification. The nature of the requirements of these

constraints determines the typology of the sign, and this allows a distinction to

be made between three types of sign. When the sign is required to reflect, exhibit,

or exemplify qualitative features of the object, i.e. to resemble the object in some

respect, Peirce calls this sign an icon, e.g. a diagram. If some kind of connection

between the sign and the object, be this connection physical or existential, is re-

quired, then the sign is an index e.g. a footprint, or a rubbing. Finally, if it is re-

quired that the sign is connected to its object by means of conventions, or social

rules, the sign is a symbol, e.g. a word. It should be noted that the three typologies

of signs are not and need not be clear-cut and mutually exclusive categories, but

rather any sign may display a combination of iconic, indexical, and symbolic

characteristics.90 In addition to these three basic types, semioticians have distin-

89. Burks 1949; Short 2007; Atkin 2010; Burch 2010.
90. Short 2007; Atkin 2010; Burch 2010.

33



guished between other kinds of signs. Sebeok91 lists the species of signs that are

found most frequently in the literature; amongst these are zero signs, icons, indexes,

symbols, and names (or verbal labels).

Sebeok92 also signals how the receiver of a message inevitably interprets both

signs and their context to interpret it. Context, as it will be seen, is of particular

importance for iconic signs.

There are two kinds of signs involved in the project at hand: symbols, or even

better names, as verbal descriptions use words, names to convey their meaning,

and diagrams, which fall under the category of iconic signs. To be more precise,

the project’s visual sign should be referred to as hypoicons, since there are a

number of conventional rules in use in drawings of bookbinding structures.93

Peirce, in fact, defines hypoicons as a particular case of iconic signs in which the

likeness to the object is ‘aided by conventional rules’.94

3.1.2. Reading and remembering iconic signs

Iconic signs have sparked a lot of discussion in the field because of the problems

that they pose as entities bearing meaning: to read an icon, in fact, unlike written

language with letters and words, is not a simple matter of knowing its components.95

This has significant implications for this project, as the automated transformations

will need to generate diagrams, whose meaning is definite and unambiguous.

3.1.2.1. Reading modalities of icons

Eco96 has identified two reading modalities97 of icons — i.e. two different ways

in which icons can be read. The border between the two is for the most part fuzzy.

91. Sebeok 1994.
92. Sebeok 1994.
93. See chapter 4.
94. Peirce 1931-1935/1958: CP 2.276; Peirce 1931-1935/1958: CP 2.279; Peirce 1992-1998:

EP 2:273, 1903.
95. See Eco 1997 for a summary of the debate on iconicity.
96. Eco 1997.
97. Eco 1997, p. 336.

34



Figure 6. Mexican on a bicycle.

Except for a few specific cases, they are not clear-cut modalities, but rather the

reading modality can gradually switch from one to the other in the course of an

interpretation. The selection of the appropriate reading modality is linked both

to the nature of the sign and to its context.

First there is an alpha modality.98 Some signs are characterized by the immediacy

of perception. They are interpreted and given some meaning strictly on the basis

of their perception. When one can approach a sign and interpret it solely through

immediate perception there is an alpha reading modality. For example, in Figure

6, by simple perception and based on our knowledge, one can easily interpret the

image as showing a man on a bicycle, and in doing so one reads the image through

the alpha modality.

Then there is the beta modality.99 According to this modality one has to assume

that what is being perceived is in fact a sign whose function was intentionally that

of the communication of a specific meaning: one sees that there is a sign and that

such a sign stands for something else. Returning to the man on a bicycle one could

notice that the man is wearing a specific hat. The hat could be recognized as being

a sombrero, a typical Mexican hat. One could then infer that the image is showing

a Mexican on a bicycle. If one then realizes that the image is on the webpage100 of

98. Eco 1997, p. 336.
99. Eco 1997, p. 336.
100. http://www.amigos.co.uk/Our-Philosophy.aspx (accessed April 2015).

35



a chain of Mexican food specialized in home delivery, one could happily conclude

that the interpretation was indeed correct.

The same image can then be read according to two modalities in order to get

to its intended meaning. However, in the case of the Mexican on a bicycle the

beta modality was weak and the reading process oscillated between the two

reading modalities. In a scale that goes from strong alpha modality to strong beta

modality there are signs that maintaining a high level of detail definition, and are

mostly read according to the alpha modality. Then there are other signs, which

are highly abstract and whose interpretation is necessarily linked to a beta reading

modality. In fact, while still visual signs, their interpretation through the alpha

modality would not go much beyond the basic shape information that the percep-

tion provides. Let us consider the image in Figure 7. Through simple perceptive

interpretation and alpha modality the droodle101 would have very little meaning.

The title provides the reading key: ‘A Mexican on a bicycle’. Only when one has

given or has guessed the correct interpretative key can one decide that what the

figure is showing is a certain object or scene, thus adapting what one is perceiving

with what is known.

In generating diagrammatic visualizations for bookbinding descriptions, one

needs to be aware of these reading modalities, and efforts should be made in order

to provide the necessary information to infer from an icon the desired meaning.

3.1.2.2. Formalization of hypoicons

One aspect of the difference between symbolic signs and iconic signs is that of

the reproducibility. One could recreate an exact copy of a sign — e.g. every letter

within this text is an exact copy of itself, every letter ‘a’ belonging to the same

font is an exact replica of itself — but, for the most part, in everyday life, one

deals with replicas — e.g. a letter ‘a’ within one font is understood as signifying

the same letter type a as any other font, for example this ‘a’, even though they

101. Droodles are nonsensical pictures containing a few abstract pictorial elements that are very
difficult to understand without being given a caption or thematic clue (Nishimoto et al. 2010).
Roger Price, a famous author of droodle books, defines a droodle as ‘borkley looking sort of
drawing that does not make any sense until you know the correct title’ (Price 1976, p. 2).

36



Figure 7. Mexican on a bicycle — as seen from above (after Eco 1997, p.344).

are not exact copies of each other — whereby only the essential characteristics

are preserved, the a-ness of the letter type a, disregarding every other characteristic

not perceived as fundamental for the message to be conveyed.102

Within signs, one can identify a type/token ratio between a particular occurrence

of a sign and the class of all occurrences of the sign, between the ‘a’ of a particular

font, and the class of the letter a in whichever form.103 Eco104 further distinguishes

between two kinds of these ratios: a ratio facilis and a ratio difficilis. In a ratio fa-

cilis each occurrence of a type of information is generated and presents itself in

accordance to the code of an expression system that has formalized its appearance,

and it is thus easily remembered and reproducible. This is the case of words and

highly stylized — or symbolic — graphics, like any letter a. In the case of a ratio

difficilis, instead, each occurrence is generated and presents itself in accordance

with its own content, either because a pre-formed type does not exist that is

available as a code, or because the expressive type (the sign) and the content type

(the information to be communicated) coincide. These signs, because of the lack

of graphic conventions, are more difficult to reproduce while maintaining them

their true meaning. These are typically iconic signs.

There can be two cases of ratio difficilis: on the one hand the sign is a precise

unit that corresponds to a precise content, it is not too difficult to reproduce and

in the long run can be seen as being ruled both by ratio facilis and by ratio difficilis.

On the other hand the sign presents itself as a whole composed of undefined

portions of imprecise content.

102. Eco 1975/2008.
103. Sebeok 1994.
104. Eco 1975/2008.

37



What this means is that, even in absence of an established convention, some

iconic signs, with time, can become recognisable as discrete units of precise

meaning. In chapter 6 it will be shown how endband cross-section diagrams can

be read in this manner.

3.1.3. Visual texts

There is a fundamental issue that has troubled semioticians: while in language

one can identify discrete units at every level of investigation — i.e. phonemes,105

morphemes,106 textual chains — in visual expressions instead there are units that

are dramatically mutable and undefined, with an infinite number of acceptable

variants. Even when iconic signs present themselves as discreet units — e.g. two

circles can signify a pair of eyes (see Figure 8) — these depend on context, often

provided by a set of shapes in a specific range of spatial relations, or by some

verbal anchors, like a title for an image. If taken away from their context they lose

their meaning: they are forced to be read only through the alpha modality and all

that can be inferred is their shape.107

Images then, it would seem, need to be regarded as macroscopic blocks of in-

formation, within which it is possible to identify pertinent units, but these units

are not independent. One can see, understand, and label the legs in the drawing

of a table, thus perceiving them as separable discreet units, but if these are indeed

separated and presented on their own they would lose their meaning of ‘legs of

table’ and would instead be perceived as ‘parallelograms’, a label still correct, but

too general to convey the meaning they possessed in context. These units do not

possess a fixed distinct value that opposes itself to that of the other units within

the system, but rather, their oppositional value and meaning depends on the

context.108

105. In linguistics, units of sound in a language that cannot be analysed into smaller linear units
and that can distinguish one word from another. (Oxford English Dictionary 2014e).

106. In linguistics, minimal and indivisible morphological units that cannot be analysed into
smaller units. (Oxford English Dictionary 2014c).

107. Eco 1975/2008; Barthes 1978; Polidoro 2008.
108. Eco 1975/2008.

38



[a] [b]

Figure 8. Units in visual language are context dependent: the same elements that are
perceived as parts of a smiley face in [a] are not recognized as eyes and mouth in [b].

Saint-Martin109 introduces the idea of special kind of visual sememe110 that she

calls coloreme to indicate the non-discreet semantic units within icons. These en-

tities are ‘continuous and spatialized topological entities, endowed with somewhat

fuzzy boundaries’.111 The continuous and spatialized nature of visual sememes has

important consequences for their generation, as it will be seen in chapter 5.

In verbal communication, there are discreet sign units that can be decomposed

into smaller pertinent and discreet units taken from within a pre-determinable

and finite set dependent on a specific code. In visual communication instead, the

sign units are not further analysable into discreet and separable sememes. Images

can then be seen as visual or iconic texts that, instead of depending on a predeter-

mined code, institute themselves an inferential code.112 Texts are in fact not defined

by the presence of linguistic elements, they can rather be defined as portions of

reality composed by a number of elements that are combined to produce a coherent

meaning.113

Images need to be analysed and read as texts, whose meaning, while depending

on the information conveyed by the units that compose them, becomes evident

only when they are considered as wholes. Unlike word-based texts though, the

receiver does not scan the text in a sequential manner, but rather engages in an

active search activity, rapidly scanning the entire sign in a synchronic manner,

jumping from part to part, looking to recognize some of these parts, thus finally

109. Saint-Martin 1990.
110. In linguistics, a sememe is the smallest unit of meaning. (Oxford English Dictionary 2014g).
111. Saint-Martin 1990 pp. 5-7.
112. Eco 1975/2008; Polidoro 2008.
113. McKenzie 1999.

39



obtaining a global understanding of the whole sign. Each new part is understood

in relation to the others until the whole sense — or indeed a whole sense — can

be inferred. Thus, meaning can be extracted through a process of active vision

and interpretation of what is being perceived from major or minor details.114

This synchronicity of the information read in visual texts allows to gather in

one scan a picture of the whole without having to first understand all of its con-

stituent parts. Diagrams are a more immediate form of information delivery systems

than sequential verbal information as they facilitate to grasp in a short time a

picture of the whole. In the literature, in fact, images are often described as whole,

integrated entities, whose elements are perceived in a simultaneous and synchron-

ous way.115 As it will be seen in chapter 7, a consequence of the synchronicity and

of the immediacy of visual texts is that problems and errors in the data can be

immediately perceived. The interdependence of the signs within visual texts is

also an important factor that needs to be kept in mind in the generation of auto-

mated diagrams, as the appearance of one bookbinding component will depend

on that of the other components that come together in a specific diagram.

3.2. Multimodal communication

As seen above, visual information may need to be anchored through verbal labels

in order to initiate a reading in beta modality. Therefore, introducing verbal in-

formation within bookbinding diagrams might be beneficial.

114. Saint-Martin 1990; Findlay & Gilchrist 2003; Ware 2013.
115. Ivins 1953/1969; McLuhan 1964/2001; Goodman 1976; Arnheim 1969; Dondis 1973; Ivins

1973; Mitchell 1980; Snyder 1980; Mitchell 1986; Eco 1993.

40



3.2.1. Dual-coding representation

Experiments by cognitive psychologists have investigated the effects of verbal-

visual mixed communication. Paivio116 has brought forward a dual-coding theory

according to which the human mind stores information in the brain through two

modes of representation: verbally, through what he refers to as logogens, and non-

verbally, through imagens. Imagens denote mental representations of visual in-

formation. Logogens denote mental representation of language information.

Learning and behaviour are thus mediated by two independent and interconnected

systems: a verbal system that is more logical and abstract that through logogens

stores information about language (but not of the sounds of the words), and a

visual one that through imagens stores information on objects, grouping, whole-

parts relationships, and spatial information and arrangement. The two systems

are said to be independent, as they can work in isolation, and interconnected, as

information can be transferred from one system to the other and they can influence

each other. Pavio117 also predicted that dual-coding might facilitate performance.

Experimental evidence suggests the correctness of his assumptions: subjects recall

twice as much when they are presented information both visually and verbally,

than just visually and even more drastically than just verbally.118 Interestingly, re-

search on memory for complex visual material also demonstrated better recalling

for visual input that is immediately followed by verbal descriptions; however, this

beneficial effect has only been demonstrated for cases in which the verbal inform-

ation integrates visual information or when it makes data more salient and distinct-

ive by simplifying and fixating it.119

Information presented duo-modally facilitates cognition, but only when the

equilibrium between the two modes is maintained. Verbal labels can be used to

facilitate memorization of the information conveyed through diagrams, and this

in turn facilitates comparison between different exemplars.

116. Paivio 1971; Paivio 1975; Paivio 1986; Paivio 1991; Paivio 2007. Also Brandimonte &
Gerbino 1996; Ware 2013.

117. Paivio 1975; Paivio 1986; Paivio 1991.
118. Paivio & Csapo 1973; Paivio 1991; Paivio 2007.
119. Klatzky et al. 1982; Wiseman et al. 1985; Paivio 2007; Brandimonte & Gerbino 1996;

Gauthier et al. 2003; Brown et al. 2010.

41



3.2.2. Verbal recoding of ambiguous visual stimuli

A famous experiment by Carmichael, Hogan, and Walter120 adds an important

aspect to this discussion. As we have seen, according to the dual-coding theory,

the two systems are interconnected and they can influence each other. This can

have a profound effect for ambiguous visual percepts. In their experiments, in

fact, they showed how a verbal label associated with an ambiguous visual stimulus

determined how the stimulus was interpreted and subsequently remembered. The

label affects drastically the perceptual interpretation.

Table 1 shows three of the visual stimuli used by Carmichael and colleagues in

their experiment (visual stimulus), the verbal labels (label 1 and 2), and the sub-

sequent interpretation (reproduced image 1 and 2). The subjects were presented

an ambiguous stimulus, for example two circles connected by a line, and they

were told that the stimulus represented one or another object (label 1 and 2), for

example eyeglasses or a dumbbell. Subsequently, the interpretation and recall of

the information were completely different, depending on and strongly influenced

by the different verbal label provided at time of perception (reproduced image 1

or 2). Language can affect and anchor the interpretation of a visual stimulus. The

same effect can also be seen when ambiguous visual stimuli are not given in isola-

tion but in context: as in Figure 8, the two circles, when in the right context are

read as eyes, otherwise are simply read as circles, but if a label ‘face’ were provided

with the picture, with a little effort one could see the circles as the eyes of a

somewhat deformed upside-down face.

Rep. image 2Label 2Visual stimulusLabel 1Rep. image 1
dumbbelleyeglasses

sunship’s wheel

tablehour glass

Table 1. Samples of ambiguous visual stimuli and their subsequent interpretation as re-
ported by Carmichael et al. 1932 (after Cornoldi & Logie 1996, fig. 1.3, p. 15).

120. Carmichael et al. 1932.

42



To summarize, there is a beneficial effect in the combination of visual and verbal

information. This affects both interpretation, memorization, and subsequent re-

calling of the information. The ambiguity of visual stimuli can be anchored through

verbal labels and/or visuo-spatial context, thus allowing for beta reading modality

and subsequent correct interpretation and recalling of the information.

As it will be shown, in the diagrams generated for this project, verbal labels are

provided as titles and schematic description texts — i.e. label: value, e.g. attach-

ment: sewn.

3.3. Categorization and prototypification

At one time it was commonly thought that the grouping of entities into categories

was the effect of arbitrary, cultural-based conventions. However, researchers in

linguistic, philosophy, and psychology121 have exposed the existence of symmetries

in classification models across different cultures and languages, linking the phe-

nomenon of categorization to common structures in the perception of the world.

This model of common structures is also appropriate for the categorization of

material objects and their verbal labels.

It is on the basis of this mental model that we can understand the world around

us and to communicate. For material objects, this model is linked to their logical

form. The next few sections will expose the formulation of this model.

3.3.1. Prototypes

Rosch122 argues that categories are not simply logical entities whose members

are equally defined by possession of a simple set of features. Rather, many natural

categories are structured towards best examples — i.e. prototypes — with non-

prototypical members ranging from better to poorer examples. Therefore not all

121. Rosch 1973; Rosch & Mervis 1975; Rosch et al. 1976; Hampton 1981; Lakoff 1987.
122. Rosch 1973; Rosch 1975.

43



members of a category are equivalent, and the best examples can serve as reference

points in relation to which the other members of the same category are evaluated.

In other words, these best example entities serve as prototypes of the category.

The more attributes a member of a category has in common with other members

of the same category — and conversely, the less attributes a member of a category

has in common with members of another category — the more prototypical this

member appears to be. The concept of the prototype functions as a sort of con-

ceptual referential schema against which percepts123 are compared, and in turn

classified.124

Psychological studies on human perception have often argued that we define

arbitrary forms in terms of a restricted vocabulary of more regular shapes or shape

concepts referred to as prototypes. These are represented by clear-cut entities

forming a referential base. These perceptual punctuations of a phenomenon are

usually associated with linguistic labels. Any shape that perceptually falls between

two good examples is seen as deviating or approaching a clear-cut label.125For

example, Rausch126 analysed the perception and categorization of an angle growing

from 0° to 180°. In his experiments he found that the continuous perception of

the growing angle was punctuated by a series of clear-cut percepts to which all

other percepts tended. He categorized these as (i) straight line, (ii) arrow, (iii)

acute angle tending towards a typical obliqueness of 45°, (iv) right angle, (v) obtuse

angle tending towards a typical obliqueness, (vi) bent straight line, no longer

perceived as an angle. All other angles are seen as tending towards these clear-cut

exemplar percepts, with the range of possible angles tending to the same percept

varying: very few will be seen as tending towards the arrow or the bent line, some

deviations from the normal will be perceived as almost or bad right angles,127 with

the great majority of angles tending towards the idea of typical obliqueness.

123. In philosophy and psychology, a percept is an object of perception, or the mental product
or result of perceiving something. Oxford English Dictionary 2014d

124. Eco 1997.
125. Mach 1897/1959; Rausch 1966.
126. Rausch 1966.
127. Arnheim 1969

44



This highlights the tendency of the human mind to classify shape percepts as

instances within a categorical system in which some features are perceptually

fundamental while others are secondary and do not need to be perceptually realized

for the percept to be recognized and labelled according to a clear-cut entity refer-

ence framework. Because of this tendency, verbal descriptions lack the ability to

deliver the specificity, the precise appearance and shape, of objects. However, it

is thanks to this labelling system that we can communicate successfully general

information about the visual aspects of reality. In practical terms, therefore, when

generating diagrammatic visualizations of bookbinding structures from verbal

descriptions, the final output will only represent generic information about the

appearance of the bookbinding components, but this does not impede the com-

munication of the essence of the elements represented.

3.3.2. Basic level of abstraction

Rosch128 describes category systems as having both a vertical and a horizontal

dimension. Along the vertical dimension one would find the level of abstraction

(or inclusiveness) of the category — the greater the inclusiveness of a category

within a category system (e.g. a taxonomy) is, the higher the level of abstraction

— while the horizontal dimension concerns the segmentation of categories at the

same level of abstraction. For example, the entities kitchen table, table, furniture,

thing vary along the vertical dimension, while the horizontal dimension would

include the variations between entities like table and chair, or car and bus. Rosch

and colleagues129 have pointed out that along the vertical dimension not all levels

are equally useful, but there is a particular level of abstraction – the basic level –

at which people tend to use a concept. This level is the most appropriate for using,

thinking about or naming an object in most situations in which the object occurs.

It is also the level at which people tend to functionally interact with an object, as

the focus is on the functionality of the category. This level usually happens to be

128. Rosch & Mervis 1975; Rosch 1978; Garner 1974.
129. Rosch et al. 1976.

45



Figure 9. An angle growing from 0° to 180° is seen as (i) a straight line, (ii) an arrow, (iii)
an acute angle tending towards a typical obliqueness of 45°, (iv) a right angle, (v) an obtuse
angle tending towards a typical obliqueness, and (vi) a bent straight line, no longer per-
ceived as an angle (after Rausch 1966).

46



Vertical and Horizontal
dimensions of a category system

L
ev
el

of
ab

st
ra
ct
io
n

Segmentation of category

thing

furniture vehicle...

table chair...

kitchen table coffee table...

wooden kitchen table

Figure 10. Dimensions of a category systems (after Rosch et al. 1976; Rosch 1978; Bobick
1987). In bold the basic level of abstraction.

the middle level of abstraction. The basic level is that of table, car, hammer, as

opposed to furniture, vehicle, tool. It is at this level that material objects are most

naturally divided into categories. Consider for example the series

furniture>table>kitchen table

furniture being the superordinate level, table being the basic level, and kitchen

table being the subordinate level. If one ponders which would be the most useful

term in the great majority of situations, the answer would have to be table, i.e. the

middle level of abstraction. Rosch’s experiments also suggested that for objects

at the basic level it is possible to form a mental image that will resemble the ap-

pearance of members of the class as a whole.130

Psychological studies on mental imagery131 have highlighted how mental repres-

entations of visual information are seen as possessing structural and spatial char-

acteristics. Every part of an object and its spatial relations are encoded in the

mental representation in such a way that they are readily available for thought

processing. In addition, most importantly, such spatial and structural information

130. Rosch 1978.
131. Pylyshyn 1976; Kosslyn 1980; Jolicoeur et al. 1984; Cave & Kosslyn 1993; Biederman 1995;

Cornoldi & Logie 1996; Kosslyn & Ganis 2010; Pylyshyn 2011; Pitt 2012; Thomas 2013.

47



gathered through visual perception is not available to language-base thinking.132

Interestingly, research in neuroscience133 has confirmed that when we perform

tasks that involve mental imaging, the same neural pathways that are activated

during perception become active: both actions of seeing a table, and picturing it

mentally activate, for the most part, the same visual processing centres in the brain.

In turn, the same neural pathways are activated by surrogate stimuli in drawings

leading to the possibility of the generation of graphic prototypes.134

Jolicoeur and colleagues135 have introduced notion of an entry point level for

atypical objects. Very distinctive and atypical exemplars have a semantic entry

point into the category at the subordinate level. The entry point is therefore an

attribute of specific objects rather than, like the basic level of abstraction, of the

category as a whole. It can be argued that the notion of the entry point is applicable

also to specific uses of objects and their category. Prototypification at the basic

level of a category is typically too inclusive to be useful in the classification of

particular kinds of objects, or their structures. Thinking of bookbinding structures

for example, a description at the basic level of abstraction would refer to an end-

band, a sewing structure, a corner, but would not allow further specification. For

the purpose of this project, a more useful level of abstraction and entry point

would necessarily be represented by a more subordinate category, thus allowing

for a more specific identification, and in turn, as it will be seen, a more useful

visualization. If one were to describe the ‘table’ items of a furniture shop, the term

table would be as generic as furniture in usual circumstances, while the terms

kitchen table or coffee table would provide the necessary level of abstraction to

allow for successful communication. In the same way, the basic level of abstraction

for this project would necessarily need to be less inclusive than usual, as it is at

that level that communication on binding structures can be valuable. For simplicity,

the basic level of abstraction and the entry point level can be considered as equi-

valent.

132. Ware 2013.
133. Kosslyn & Thompson 2003.
134. Kosslyn et al. 1993; Kosslyn & Thompson 2003.
135. Jolicoeur et al. 1984.

48



Tversky and Hemenway136 point out that parts and their configuration are a

predominant feature at the basic level. A chair includes a seat, a back, and legs; a

knife is seen in terms of a handle and a blade. These parts, and their names, refer

both to perceptually identifiable segments of an object, and to specialized func-

tions. The prevalence of parthood could be what grants its special status to the

basic level. Specifically, knowledge about parts could explain the superior inform-

ativeness of entities at the basic level. In addition, different objects at the basic

level seem to differ in respect to their parts, while sharing other attributes;

whereas different objects belonging to subordinate levels of the same basic level

category share parts, but differ in respect to other attributes.

Entities at the basic level are the most general and they have similar and recog-

nizable shapes. They are also the most abstract entities for which a single image

can be formed for such a category. The basic level of abstraction is the highest

level for which a generalized outline can be identified and which best reflects the

redundancies of the category. Thus, a single line drawing can be used to represent

the entire category, if it represents the objects at the correct level of abstraction.137

Prototypical entities represent their category and convey information on the

whole/part relationship and structural configuration of objects. When information

about material objects is not presented at the correct level of abstraction, the

communication process fails to convey the necessary information. In particular,

using descriptive terms at a superordinate level created indeterminateness. This

will be explored in more detail in the course of this thesis as it highlights a funda-

mental issue that needs to be taken into careful consideration when selecting de-

scriptive terms.

136. Tversky & Hemenway 1984.
137. Rosch & Mervis 1975; Rosch et al. 1976; Tversky & Hemenway 1984.

49



3.4. Objects: their form and their shape

Signs signify their object in some respect, and iconic signs resemble their object

in some respect. Of all the characteristics that material objects possess, some are

more important than others for their identification and signification.

In order to be able to produce successful diagrams of bookbinding structures,

there are two features of material objects that need particular attention and that

should be communicated: form and shape. These are in fact, as it will be seen in

the following sections, amongst the most important and fundamental characteristics

that define an object both perceptually and conceptually.

3.4.1. The form of objects

Material objects are spatial entities that occupy a single specific region of space-

time as autonomous forms of organization which can consist of multiple parts.138

Each one of these parts is in spatial relation to the others and to the whole object.

What needs to be ascertained is whether a system of communication claiming to

inform about the specificity of an object should deliver information regarding this

structural organization, or whether, the collection of the parts is in fact enough

to describe the whole.

Mereology (from the Greek μέρος, ‘part’) is a branch of philosophy and

mathematical logic dedicated to the study of the relations between parts and the

wholes they form. Its root as a field of investigation can be traced back to the early

days of philosophy (the Presocratics, and then Plato, Aristotle, Boethius), then to

Leibniz, Kant, Husserl, and the modern ontologists and metaphysicians.139 Increas-

ingly, mereological studies have been gaining interest in the fields of knowledge

representation and database construction.140

138. Saint-Martin 1990; Koslicki 2008.
139. Varzi 2014.
140. Doerr et al. 2001.

50



3.4.1.1. Material components and form

In the typical mereological analysis of parts and wholes that came to dominate

the scholarship in the field during the 20th century, a well-structured whole and

a simple heap of its parts arranged in no particular order is the same: under-estim-

ating the role of wholes and concentrating on the parts, an object is simply the

sum of its material parts, arranged in any way. This came to be for a number of

historical reasons and to answer a number of metaphysical problems,141 nonetheless,

as pointed out by Koslicki,142 there must be a way to distinguish between a motor-

cycle in running condition and the sum of its disassembled parts. Some contem-

porary philosophers143 propose to re-accept in the discourse on parthood and

wholes, the Aristotelian notion of form, and argue that material objects can be

regarded as structured wholes, to which identity is fundamental, not only that all

parts are accounted for, but also that these parts exhibit a certain configuration,

or arrangement. Therefore, enumerating the parts a complex object is not enough

to describe its essence.

Any communication about material objects should be able to communicate

both the range and type of their material components as well as their configuration.

In the next chapter, examples of descriptions of bookbinding structures as mater-

ial objects will be presented, and it will be seen how not being able to communicate

the form of objects poses important problems.

3.4.1.2. Logical form and expression

In addition to form, in order to describe an object, one also needs to be able

to refer to its essence. In other words, to communicate an object table, in such a

way that the concept is applicable to any table, there needs to be a way to define

an idea of table-ness capable of prescribing what components could come together

to make a table, and in which spatial arrangements, i.e. a conceptual categorization

of the prototypical instance of table. This abstract idea of the objects would

141. Simons 1987; Albertazzi 1999.
142. Koslicki 2008.
143. Fine 1999; Harte 2002; Johnson 1994; Koslicki 2007; Koslicki 2008.

51



function like a model of it, by prescribing which components are necessary or

possible for a given object, and in which relationships these can stand. In philo-

sophical terms, this abstract entity is referred to as logical form.144

Material objects, therefore, posses an abstract form that determines all possible

configurations, a sort of recipe for how to build objects of a particular kind K;

while their material components can be thought of as the ingredients that are

called for in the recipe. Each kind K has associated with it a set of requirements

(the recipe) specifying the range, configuration — and sometimes the number —

of material components eligible to compose an object of that particular kind K.

The logical form makes available positions for the material components to occupy,

provided that the occupants satisfy the type restrictions imposed on the positions

in question. As a consequence, objects exhibit a particular configuration imposed

on them by the logical form. The logical form contributes to two sorts of constraints

on its content: (i) constraints concerning the types of components; (ii) constraints

concerning the topological or geometrical configuration.145

To communicate descriptions of different instances of the same kind of mater-

ial object — e.g. different endleaf structures — one must account for the various

material components and for the different ways in which these can stand in rela-

tions to one another. Also, in order to be able to describe any instance of the same

reality and still make sense, the descriptions must all have something in common,

something that corresponds to the logical form of the kind of material objects.

Wittgenstein146 calls expression what they have in common and defines all propos-

itions147 having the same expression in common as belonging to the same class of

propositions: this expression is what is constant within the class of propositions,

144. Pietroski 2014. Of particular interest is the reference to the logical form by Wittgenstein
in his his Tractatus Logico-Philosophicus (Wittgenstein 1922/2012). In accordance with a more
recent reading of Wittgenstein’s work that replaces the traditional logistic interpretation with one
more metaphysical, or physicalist, (Nordmann 2002) and considering the engineering background
of the philosopher,(Hamilton 2001b; Hamilton 2002; Hamilton 2012) Wittgenstein’s investigations
on the relationship between language and the world can be scaled down from the general to the
practical issues of the verbal communication of material objects.

145. Wittgenstein 1922/2012T: 2.0232; T: 2.0233; T: 2.02331; T: 2.024; T: 2.025; Koslicki 2008;
Pietroski 2014.

146. Wittgenstein 1922/2012T: 3.31; T: 3.311; T: 3.312; T: 3.313.
147. Proposition: in logic, statement which is capable of truth or falsity (Oxford English Dic-

tionary 2014f).

52



whilst everything else can vary. The expression acts as a recipe of the material

object through which all different propositions can be generated. These important

concepts will be discussed more in detail in chapter 5.

Both the logical form and the expression then act as a general prescriptive

schema for the material object. Logical form is inherent in the conception of ma-

terial objects, or in the model of the object in our mind. The expression instead

is the reference to the logical form that can be found in the descriptions. For ex-

ample, the expression is embodied in the descriptive schema of databases as it

prescribes what variables are possible for a certain object.

In other words, the expression is the model of the class of objects that need to

be described, and the logical form is that essence of the object that the model of

strives to capture.

As it will be covered in the following chapters, the lack of a reference to the

logical form has critical consequences for the efficacy of a bookbinding description.

3.4.2. The shape of objects

Another concept and percept that is essential for the specificity of material

objects, and thus for their communication, for the fact that it allows us to distin-

guish between different instances, is their shape. If compared with other charac-

teristics of objects — e.g. depth, motion, speed, colour, etc. — shape is exceptional.

Many objects can share the same colour, or measurements, whereas, the shape of

an object is only shared by objects belonging to the same category: an apple can

be red, yellow, or green, or any other colour one can think of, just as much as a

car can be characterized by the same colours; no two apples are exactly identical,

and there are many different kinds of cars, but still the human brain has no trouble

identifying an apple as an apple and a car as a car. The only constant, the only

perceptual property sufficiently complex and unique to allow for an object to be

identified unambiguously, seems to be the general shape of the object, and in

particular its prototypical nature.148

148. Pizlo 2008.

53



3.4.2.1. The nature of shape

Defining what shape is can be difficult. Pizlo149 proposes a working defintion:

Those geometrical characteristics of a specific three-dimensional object that

makes it possible to perceive the object veridically from many different

viewing directions, that is, to perceive it as it actually is in the world ‘out

there’.

Shape is both structured and complex, and this prevents shape ambiguity and

having to rely on the context of the percept for its recognition.150 This last point

does not contradict what exposed on page 38 when it was stated that the percep-

tion of the two circles as a pair of eyes depends on the context. In that case what

depends on the context is the perception of what the two circles symbolize, not

the perception of their shape. In fact, in Figure 8[b] the two circles are still clearly

perceived as two dots or circles, what changes and depends on the context is the

label of the meaning attached to them. In other words, the right context is what

allows to read the image in beta modality, but it is not necessary for it to be read

in alpha modality.

Shape is structured in the sense that there can be some invariants and a priori

constraints, such as symmetry, planarity, compactness, that when applied to the

two dimensional projection on the retina help in reconstructing the three-dimen-

sional shape of an object: shape recognition does not depend solely on the phys-

ical perception as much as, for example, colour perception does. Shape is complex

because it can be described along a large number of dimensions, and its complexity

is almost never eliminated by the perspective projection onto the retina. To change

one shape into another an infinite number of points would have to be moved.151

The general shape of an object is an essential component for its identification,

and therefore any communication of a material object should convey shape inform-

ation, at least in its prototypical form.

149. Pizlo 2008, p. 1.
150. Pizlo 2008.
151. Pizlo 2008.

54



3.4.2.2. Regularities for the prototypification of shape

Research in Gestalt psychology152 has shown that arbitrary shapes are seen in

terms of more symmetrical versions of themselves. Some forms are visually per-

ceived as tending towards other more regular shapes to which they are seen as

similar or equivalent. Some features were found to be more important for the

general impression than others as even slight changes of these result in noticeable

effects on similarity, while others can be substantially changed without affecting

the general impression. This special characteristic is usually referred to with the

German words Prägnanz — i.e. singularity, or meaningfulness, pithiness — and

these particular features are called prägnant — i.e. a singular, meaningful, special

value of a trait or parameter.153

Goldmeier154 reports on the perceived similarity amongst shapes. Amongst the

many various values that a geometrical variable — e.g. parameters, measurements,

relationships — can assume in a figure, the few prägnant ones are psychologically

singled out and phenomenally realized. For non-prägnant characteristics instead,

particular values are not realized as such, and ranges of different values will result

in the same phenomenal experience.155 From the results of his experiments

Goldmeier delineates the general principle that the basis of similarity is the con-

servation of prägnant values of geometrical parameters, the conservation of sym-

metry being of particular importance. Symmetry is counted amongst the a priori

constraints for shape recognition by Pizlo.156

In addition, symmetry and other regularities in a perceived object may allow

to reduce the amount of information that needs to be collected and stored about

that perception. Regularities, allowing for simplification, in fact, are capable of

representing more of an image than arbitrary or irregular features.157

152. Koffka 1935.
153. Palmer 1999; Sternberg & Mio 2009.
154. Goldmeier 1972.
155. Arnheim 1969.
156. Pizlo 2008.
157. Attneave 1954; Barlow & Reeves 1979; Feldman 1997; Feldman 2000.

55



Symmetry allows a figure to be perceived and coded abstractly and economically.

Simplification by regularization, and in particular symmetrization, is at the core

of shape perception and conception.

The diagrams for this project require that the shape drawn be recognizable,

but also prototypical as they derive from verbal labels. This calls for highly sym-

metrical shapes. Let us consider for example the diagram of a stirrup ring, a Greek-

Byzantine clasp in the form of an ornate ring with a flat slot below it to take the

strap,158 in Figure 11. The diagram does not exhibit any of the irregularities of the

example in the photograph, but it is still recognizable as belonging to the same

class of objects.

3.4.2.3. Line drawings

Prototypical shapes can be generated using two processes: (i) one of regulariz-

ation that preserves all prägnant features; and (ii) one of simplification that only

keeps the essential shape features, without affecting their perception.

Gestalt psychologists159 have studied the perceptual phenomenon by which an

object is distinguishable from the background and refer to the dominant shape

as figure and to the background as ground; the phenomenon is referred to as figure-

ground organization. The figure is seen as having a closed contour and a shape,

while the ground seems shapeless and as extending behind the figure. Figure-

ground organization is an essential step in our perception of surfaces, shapes and

objects.160

Figure-ground organization and the perception of shapes based on their contour

against a background is rooted into the human perception system and we perceive,

reproduce, and communicate percepts solely based on their contour. Lines, in

line drawings, act as surrogate stimuli for the visual contours of the depicted ob-

jects. There is strong evidence161 suggesting that the identification of lines with

158. Ligatus 2013.
159. Rubin 1915.
160. Pizlo 2008.
161. Dehaene 2009; Pinker 1990; Sayim & Cavanagh 2011; Pettitt & Pike 2007; Cook 2013;

Halverson 1992; McKie 2012; Sayim & Cavanagh 2011; Cook 2013; Walther et al. 2011; Hochberg
& Brooks 1962; Kennedy 1974; Kennedy & Ross 1975; Kennedy 1977; Halverson 1992.

56



Figure 11. Comparison between a stirrup ring shape and its prototypified diagram. The
diagram is generated as part of this project; the photograph is taken from the images
collected during the survey of the printed books in the Library of the Monastery of St
Catherine on Mount Sinai, Vol. 4077.2841α (see §4.2.4, §6.1.4).

objects’ contours is an innate function of normal human vision, not linked to any

social or representational convention, that occurs both cross-culturally and irre-

spectively of the level of development of the subjects. Line drawings seem to

harvest a particular feature of visual perception that is hard-wired into the human

perceptual system and can be therefore exploited as an international feature of

visual communication.

To depict something strictly by its outline is to simplify, reduce its visual inform-

ation to the bare necessities: complex optical arrays are transformed to simple

lines, and the linear depiction is limited to just the salient features that make the

object recognizable. The figure is reduced and simplified, stripped of all non-

strictly essential information, through an abstraction process that selects only

those features deemed as distinguishing. Outline drawings can be considered as

prototypical instances of the classes that they represent.162

Biederman and Ju163 have demonstrated that an object depicted as a simple line

drawing can be recognized as quickly and accurately as high-resolution photo-

graphs that preserve the object’s details, surface texture, and colour. This phe-

nomenon has been more recently researched by psychologists and computer sci-

entists at Stanford University, Ohio State University and the University of Illinois

162. Halverson 1992.
163. Biederman & Ju 1988.

57



at Urbana–Champaign.164 The researchers concluded that the features preserved

in line drawings are sufficient for the categorization of natural scenes, and the

same features are most likely used in categorizing both line drawings and photo-

graphs.

Line drawings exhibit, therefore, the minimum amount of information necessary

for shape recognition, and are useful communication tools. In the next chapter,

it will be seen how line drawings of bookbinding structures can successfully

communicate information by visual means.

3.4.2.4. Perceptual biases

Shape prototypification by process of simplification, regularization, symmetriz-

ation, and preservation of other prägnant features is sufficient for the generation

of functional graphical prototypes. However, one should also make sure that such

graphic prototypes preserve and communicate in a psychologically sound manner

the composition of the object represented along with its form, and the spatial re-

lations amongst its material components. For this one might need to rely on exag-

geration of some features and accurate but imprecise spatial relations for immediacy

sake.

There is psychological evidence of systematic differences between perceptual

displays and their mental representations.165 Locations and precise measurements

are not remembered absolutely by humans as there is not a coordinate system

hard-wired in our eyes or brain; rather they are indexed schematically and approx-

imately. Memory for locations and spatial relations is relative. Relative to other

elements at the same level of analysis, or to a frame of reference on a higher level

of analysis.166 For these reasons, Tversky167 suggests that the design of effective

graphic representations should ensure that (i) content and form of the visualization

correspond to those of the desired mental representation and (ii) are readily and

164. Walther et al. 2011; Shen & Walther 2012.
165. Tversky 1991; Schiano & Tversky 1992; Tversky 1997; Tversky 2006; Tversky 2007.
166. Tversky 1997; Tversky 2005a.
167. Tversky 2006.

58



correctly perceived and understood, complying with what she refers to as the (i)

principles of congruity and (ii) apprehension.

Naturalism in the diagrams might then actually not be advisable for a successful

communication.

Summary

This project deals with three different entities — objects, words, and images.

Information about objects has to be able to be communicated to a human receiver

through three processes: perception, communication, and automated transforma-

tion. In particular, information has to be able to reach its final destination, while

undergoing the transformation process from a verbal communication system to

a visual one. Verbal and visual communication systems are quite different, but

can work in tandem, complementing each other.

Diagrammatic visualizations can undergo a process of simplification and ab-

straction from reality, being reduced to the essential shape and form, and still be

read and recognized by a human mind as the object that they represent. In fact,

it is through the regularities and generalities of shape that an object’s appearance

is understood and conceptualized, and thus not seen merely as a particular and

specific entity.

A way to ensure memory, reproducibility, and usability of graphic prototypes,

is to reduce the amount of information to the minimum essential for shape com-

munication, while guaranteeing redundancy of information, so that it can be

‘compressed’ in our brain’s memory, and in turn easily ‘decompressed’ at will.

Line drawings are sufficiently simple and, at the same time, complex enough to

preserve shape constancy and object identification, to be successfully used in the

generation of graphic prototypes. Entities at the basic level of abstraction have a

recognizable generalized outline shape that best reflects the redundancies of the

category they come to represent. Further, object recognition from outlines is

universal and innate amongst humans. Therefore, simplified line drawings are

59



sufficient for the communication of material objects, bookbinding structures in-

cluded.

Vision is not a mechanical and passive system, but rather percepts are psycho-

logically realized and referentially utilized in different degrees of accuracy, with

some features being strongly more psychologically salient than others.

Verbal labels can anchor the ambiguity of visual information, and can serve as

referential punctuations in the fluidity of visual percepts, thus providing a clear-

cut structure to an otherwise essentially analogical system.

This chapter has considered a number of considerations on the nature of the

entities and processes involved in this project and has outlined a theoretical

framework onto which a method could develop. The next chapter will review the

ways in which bookbinding structures have been described, both verbally and

visually. Different kinds of verbal descriptions of bookbinding structures will be

covered, and these will be then analysed to see which kind could be best used for

appropriate communication and even more so for a successful intersemiotic

translation.

60



Chapter 4. Bookbinding descriptions

Like men, books have a soul and body. With the soul, or literary

portion, we have nothing to do at present; the body, which is

the outer frame or covering, and without which the inner would

be unusable, is the special work of the binder. He, so to speak,

begets it; he determines its form and adornment, he doctors it

in disease and decay, and, not unseldom, dissects it after death.

William Blades, The enemies of books, 1902, p. 96.

Traditionally, books are not considered as susceptible to archaeological treat-

ment as most are still living tools kept in libraries to be used, and they are not se-

cluded objects kept in museum collections as examples of earlier times. It is true

that some books are indeed treated as museum objects, but the prime function

of the great majority of them is their service as tools, as useful objects.168

4.1. Books as artefacts

The book, studied as an object, holds a wealth of information beyond its textual

content. Bindings are a prominent part of the artefact book. Books were often

bound in accordance to the choices made by their owner, and bookbindings can

therefore reflect financial status or the intended use of a book. From the study of

bookbindings it is often possible to deduce information about where and when

168. Adams & Barker 1993.

61



a book was first used, even when such information is not offered by the content

itself.169

One of the main problems faced by scholars of the history of bookbinding is

that the discipline is young, lacking a well-established vocabulary and a stable

description system. This is especially true for research on historical bookbinding

structures.170 In 1945, Goldschmidt referred to the discipline as a ‘humble auxiliary

[one], rather childish to some, attractive to others, not entirely useless and un-

doubtedly innocuous.’171 Despite such disparagement, a great number of fine and

valuable scholarly studies were pursued in the 20th century, spanning the mainly

aesthetic interests of the Arts and Crafts Movement towards the more scholarly-

approach of bibliography.172

4.1.1. Binding structures and decoration

The first role of a binding was — and still is — that of protecting the text, but

because of the primacy of its impact on the eye, it soon was instilled with and re-

flected an additional function, that of indicator of social status through decoration

and preciousness of covering materials. For the historian of bookbindings there

are two levels of research, one more internal to the object, looking at the structure

and the materials used to construct a binding, and another focused on externally

visible elements such as the cover and overall appearance of the book, i.e. the

material employed in covering the book and its decoration. The study of structure

and decoration can uncover the reception, and function of books, as well as help

establishing the date, provenance and status of a book.173 Interesting is, for example,

the identification of the Turbutt Shakespeare as the original Bodleian copy of the

First Folio solely by its (rather plain) binding,174 or, more recently, the use of the

169. Pickwoad 2012; Velios & Pickwoad 2012.
170. Szirmai 1999.
171. Goldschmidt 1945, p. 175.
172. Foot 2004.
173. Adams & Barker 1993; Pickwoad 2011a.
174. Madan et al. 1905.

62



evidence from the bookbinding of the New York copy of Galileo’s Sidereus

Nuncius (SNML) to confirm its forgery.175 Bookbinding structures and decoration

should therefore not concern exclusively the bookbinding historian, but also his-

torians involved in the history of the book, who should take a holistic approach

to the object of their research.176

Goldschmidt,177 in 1928, referred to the considerable amount of literature on

bookbinding already in existence by that time, and to the fact that the state of

knowledge on bookbinding, in spite of the great amount of literature, was still so

imperfect, that it seemed impossible to give a coherent history of the binding of

books. He pointed out that ‘far fewer people [could] give a reasoned opinion on

the country of origin and the approximate date of an old bookbinding, than a

piece of pottery or furniture.’ More than eighty years later, the situation has not

changed much. With time, the discipline has evolved and accrued knowledge on

bookbinding decoration and historical bookbinding structures.

4.1.1.1. The evolution of bookbinding studies

There are two levels of research for books as artefacts: one more internal to the

object, looking at the structure and the materials used to construct a binding, and

another focused on externally visible elements such as the cover and overall ap-

pearance of the book, i.e. the material employed in covering the book and its

decoration. As well noted by Adams and Barker178 ‘structure and decoration are

basic factors in understanding reception, function, influence and survival of

books’.179

When scholars first became interested in the bindings of books and started

studying them as artefacts, they addressed their interests towards an art history

approach, focussing their attention on exterior, aesthetic elements, such as binding

decoration, and the idea of ‘styles’, to the extent that for the first half of the 20th

175. Pickwoad 2014b.
176. McKitterick 2007.
177. Goldschmidt 1928, p. 112.
178. Adams & Barker 1993, p.21.
179. See alsoMcKitterick 2007.

63



century, the history of bookbinding was virtually synonymous with the history of

bookbinding decoration.180 Decorative elements, and the specific tools used to

create them, have been analysed, compared, and identified to hypothesize the

links between specific tools and specific workshops, and subsequently, of distinct,

individual binders. The consideration of decorating tools from a purely utilitarian

point of view as clues to reconstruct parts of the history of an artefact, have proved

fruitful and have been central to the work of the majority of scholars interested

in the field. Just in the same way as the study of art history can either be based

upon the notion of styles, or that of historical sequence and continuous change

over time, sporadically in the early twentieth century and, more frequently during

the last thirty to forty years, a new interest in a holistic study of the whole bound

book flourished, encompassing the materials employed and its structural features,

extending the research beyond the study of decoration.181

The artefact book is a complex object constituted of multiple components (e.g.

gatherings, textblock, binding structure, covering, decoration, metal furniture,

and so forth) that, when considered in its totality, proves of invaluable historical

interest. If subdivided into individual elements, the study of the book loses much

of its meaning. So, while the study of decoration is useful, it should not be con-

sidered merely on its own.182

Furthermore, the great majority of books have little or no decoration and their

study has to be based upon that of their structures. It has been estimated that the

extensively expensive bindings that are often the sole objects of study of book-

binding history constitute only a very small minority — perhaps no more than

one per cent — of the entire corpus of bookbindings ever produced.183

The artefacts are of course the principal objects of the above-mentioned studies,

but especially for the very early periods of the history of the codex there are but

a scarce number of exemplars. It has been estimated that for Coptic, Anglo-Saxon,

and early medieval bindings we are left with no more than 0.01% of the probably

180. Foot 1993.
181. Pollard 1956; Pollard & Potter 1984; Foot 1993; Foot 2004.
182. Clarkson 1978.
183. Pickwoad 2011b.

64



total number of items produced, having lost any evidence for the rest.184 For this

reason, scholars have to integrate into their research evidence other than what

can be directly observed today. For instance, there are a small number of written

accounts put together by binders and other observers, varying from simple sets

of instructions and aides-mémoires, to accurate trade literature in the form of

bookbinding manuals describing the processes involved. Although occasionally

rather accurate and explicit, these written accounts more often than not are lacking

in details and sometimes almost incomprehensible, as for the most describe the

techniques just by means of words; nevertheless, they are noteworthy and useful

to the historian. Pollard and Potter185 have traced written evidence of bookbinding

as a craft back to the sixth-century CE186 and the first known detailed description

of a bookbinding process comes from a text written in the tenth century by an

Islamic binder.187 In the West, apart from the drawings contained in the frontispiece

of a twelfth-century manuscript (Bamberg, Staatsbibliothek, Msc. Patr. 5) showing

some bookbinding operations, there are not records of detailed bookbinding in-

structions earlier than the beginning of the seventeenth century. Although these

are comparatively recent, they still provide insight into the techniques adopted

by earlier bookbinders as, like all trades, bookbinding is conservative, and binding

techniques did not change much up to the nineteenth century.188

184. Pollard & Potter 1984; Szirmai 1999; Foot 2004.
185. Pickwoad 2011b.
186. His etiam addidimus in codicibus cooperiendis doctos artifices, ut litterarum sacrarum

pulchritudinem facies desuper decora vestiret, exemplum illud Dominicae figurationis ex aliqua parte
forsitan imitantes, qui eos quos ad cenam aestimat imitandos in gloria caelestis convivii stolis nup-
tialibus operuit. quibus multiplices species facturarum in uno codice depictas, ni fallor, decenter ex-
pressimus, ut qualem maluerit studiosus tegumenti formam ipse sibi possit elegere. ‘Further, we have
added to the scribes skilful craftsmen, so that a beautiful external form may cover the beauty of
the Sacred Scriptures, somehow following maybe the famous example of the Lord’s parable, who
covered with nuptial gowns those who he wanted to invite to the banquet in the glory of the
heavens. For the bookbinders, if I’m not mistaken, we have conveniently exhibited in one book
many styles of binding descriptions, so that one can choose for himself the sort of binding he
prefers.’ In Cassiodorus Senator, De institutione divinarum litterarum, Liber I, XXX, 3, (c.550
CE). Author’s translation. (Cassiodorus 1961).

187. Abū Ja‘far al-Nahhās’s treatise Craft of the Scribes providing a number of instructions
for bookbinders was included in the eleventh-century work by Tamin ibn Al Mu‘izz ibn Bādīs,
Book of the Staff of the Scribes and Implements of the Discerning with description of the line, the
pens, soot inks, l q, gall inks, dyeing, and details of bookbinding, which describes in its final chapter,
On the art of binding books in leather and the use of all its tools until it is finished by the bookbinder,
tools and techniques. (Levey 1962; Bosch et al. 1981; Pollard & Potter 1984; Breslauer 1986; Foot
1993).

188. Pollard & Potter 1984; Breslauer 1986.

65



Quite a few references have been discovered both in manuscripts and in printed

books, but it should be noted that workshop manuals – perhaps the most useful

of the written information on bookbinding one could hope to find – only survive

by accident, as they are generally used until they literally fall to pieces and are

then thrown away. Important information can luckily, sometimes be gathered

from written accounts as information is transmitted from master to apprentice.

As with any other craft, bookbinding is normally learned and passed on from

master to apprentice by means of trials and practical demonstrations rather than

from books or other documents. In fact, a careful examination of the literature

on bookbinding gathered throughout the centuries shows how only about ten per

cent of it actually covers binding techniques and structures in enough detail to

be regarded as truly useful to the bookbinding historian.189

Other significant sources of information on bookbindings – on their appearance,

internal structure or the way they function - can be derived from artistic sources

found in paintings and sculptures, when these are examined with a critical and

proficient eye that discerns reality from artistic interpretation. There are examples

of artists who would merge characteristics of different structures in one item, thus

inventing fictitious books. In most cases, however, one has to turn to extant

books.190

As noted above, the study and description of the structures and of the materials

employed is a new approach to the study of the history of the book that has ap-

peared consistently only during the last few decades. Scholars have begun to focus

their attention on the materials that constitute a binding, and the techniques em-

ployed to create it, observing the elements that make up a binding structure, or

lacking the original binding, the traces that remain. Roger Powell (1896-1990)

and Sydney Cockerell (1906-1987) deserve special consideration. Drawing from

the experience of their master, Douglas Bennett Cockerell (1870-1945), they created

a new approach to book conservation that they then passed on to their apprentices,

thus not only fundamentally influencing the evolution of the modern discipline

of book conservation, but also subsequently that of the archaeology of the book.

189. Pollard & Potter 1984; Breslauer 1986; Schmidt-Künsemüller 1987; Szirmai 1999.
190. Pickwoad 2008.

66



Their legacy came to be commonly known as the Powell-Cockerell School.191 Their

interest in the techniques and workmanship of binders of the past, the materials

used, the morphological and decorative characteristics of books, the circumstances

in which they were commissioned, and the uses of these books, were all passed

on to new generations of apprentices and book conservators.192

As a result, considering books not just as supports for text, but rather as complex

objects, whose study can contribute to research efforts in cultural history, a new

approach was established. They developed a prototype for a descriptive method-

ology of binding structures that is now shared by those currently interested in the

archaeology of the book; this involves careful description of materials, techniques

and structures, recording of the physical characteristics of the artefacts examined

and statistical analyses of such information as well as illustrations, diagrams and

photographs.

In analysing the various structures that have been employed in bookbindings

over time, it becomes clear that, although there is indeed a limited number of

basic structures, the variations of the details of the different components seem

endless. There are three main reasons accountable for this variety: firstly, the fact

that the processes evolved and were adjusted over time; secondly, different solu-

tions were devised for similar problems and different materials were employed in

different geographic locations; thirdly, various workshops had their own special

techniques taught as trade secrets.193 From these, in turn, one can deduce, for ex-

ample, information about where and when a book was first used, its financial

status, or its intended use.194

4.1.2. Terminology

With the ever growing interest in a holistic approach to the study of the books

as artefacts, scholars have focused their attention on the description of binding

191. Petherbridge 1987, p. 5.
192. Sharpe 1996.
193. Pickwoad 1995; Foot 2004.
194. Pickwoad 2011a; Pickwoad 2012.

67



structures, materials and techniques. But given the young age of the discipline,

and the complexity of some structures, an agreed and generally understood

vocabulary is lacking. Despite the publication of a number of technical glossaries,195

bookbinding has suffered historically from a lack of explicit and widely accepted

and understood terminology for materials and structures. Many terms have been

used to describe similar, though nevertheless significantly varied structures of

different provenance and age. The existing terms are usually derived from trade

terminology and dealers in antiquarian books or bibliographers. In addition, many

scholars have developed their own specific lexicon of terms to describe bindings.196

There are terms that are too vague and open to different interpretations, with

terms referring to multiple structures, or multiple entities being referred to by the

same name. In addition, binding traditions in different countries have developed

many different practices, to the point that often there is not an equivalent or precise

term for certain structural details across different languages. Also, many binding

components have never been verbally labelled at all.

Another problem is that many structures need to be described at the component

level, as one structure that could be generally labelled with one term — e.g. ‘integ-

rated hook endpapers’ — can indeed present a great number of different possible

configurations. For this reason, one needs a precise terminology that could label

each component, and a way to communicate the specific form of the structure.

Usually the latter is attained through ad hoc graphic representations.

The lack of a uniform and established terminology in English — or in any other

language for that matter — renders descriptions imprecise and liable to misinter-

pretation.197

To complicate matters further, there is not as yet a standard method for describ-

ing bindings and offering guidelines on what to observe, which elements should

be recorded, to what degree of detail, and in what way. Different projects would

need different levels of detail, but such guidelines, if flexible enough, could pro-

mote best practice, and contribute towards developing a standard of description,

195. Cloonan 1984; Tanselle 2002.
196. Cloonan 1984; Ravenberg 2012.
197. Szirmai 1999.

68



in a similar manner to what the Text Encoding Initiative (TEI)198 has done for the

description of texts and their supports. Incidentally, it should be noted that the

TEI P5 Guidelines for the physical description of manuscripts do provide an

element for the description of bindings; however, in line with the traditional de-

scription of bindings that can be found in catalogues of manuscripts,199 this element

is far from adequate for the purposes of a bookbinding historian, as the unstruc-

tured information there recorded tends to be generalized, superficial, and incon-

sistent.200

The Ligatus Research Centre of the University of the Arts London201 has been

developing a descriptive schema for bookbinding structures utilizing eXtensible

Markup Language (XML) technologies, which, to tackle the terminology problem,

has fed into the development of a comprehensive thesaurus of bookbinding terms202

that goes down to the component level of complex structures.203 An earlier version

of the schema was used for the survey of the printed books of the library of St

Catherine’s Monastery on Mount Sinai, Egypt in 2007.

4.2. Verbal and visual descriptions of bindings

In the literature, one can find different types of descriptions of bookbinding

structures. There are some that only make use of natural language without any

visual integration, whilst others accompany text with photographs or drawings.

More recent examples make use of controlled vocabularies and structure the in-

formation within databases.

The next few sections cover examples from the main description methodologies

found in the literature. These are pure textual descriptions without any graphic

198. Sperberg-McQueen & Burnard 2005.
199. See for example Jemolo & Morelli 1990, or Petrucci 2001.
200. Pickwoad 2012.
201. Ligatus from now onward. http://www.ligatus.org.uk/
202. http://www.ligatus.org.uk/glossary/
203. Pickwoad 2004; Pickwoad & Gullick 2004; Velios & Pickwoad 2004; Velios & Pickwoad

2005a; Velios & Pickwoad 2005b; Velios & Pickwoad 2008; Velios & Pickwoad 2009.

69



representation, textual descriptions with photographs, textual descriptions with

diagrams, controlled vocabularies, and structured descriptions.

4.2.1. A description of Coptic bindings

A first example is taken from Henri Hyvernat’s files in the Corpus Scriptorum

Christianorum Orientalium in the Catholic University of America in Washington,

DC, and transcribed by Theodore C. Petersen. This is a handwritten account in

Italian from the librarian Enrico Castellani on the binding of Coptic Codices written

in Rome in 1920. Since the text is rather short and unpublished, the whole page

is given here (the translation from the original Italian is made by this author):204

Coptic BindingLa Legatura dei Codici Copti
On the spine of the quires (ternions, qua-
ternions, quinions, etc.), regularly placed one
on top of the others, make four sewing stations.
Each sewing is executed with two threads of the
length of 60 cm each.
On top of the sewn quires, thus already forming
a bookblock, place the external boards, in such

Sul dorso dei fascicoli (ternioni, quaderni,
quinterni, etc.) sovrapposti regolarmente gli uni
sugli altri, si fanno quattro cuciture.
Ogni cucitura è fatta con due spaghi della
lunghezza di 60 Cent. l’uno.
Sui fascicoli cuciti e formanti già un volume, si
accomodano i cartoni esterni, avvertendo che

a way that on the three free sides, less the spine
edge, these project by about 1 cm.
On the board, at about four centimetres from
the spine edge, make four holes with an awl at

questi abbiano ai tre lati liberi, meno il lato del
dorso, una sporgenza uguale di un centimetro
scarso.
Sul piano del cartone, a circa quattro centimetri
dal dorso, si praticano in corrispondenza delle the four sewing stations; [i.β] tie then the two
quattro cuciture, altrettanti fori, con un threads of each sewing station together in an
punteruolo, [i.α] si prendono poi i due spaghi open circular knot; pass then the end of the
di ciascuna cucitura, e se ne fa un nodo aperto bottom thread through the hole in a downward
a forma di anello, si prende poi il capo dello motion; join then the two threads together at

the spine.
Once the two threads are thus joined together,
pass them through the open circular knot, and

spago di sotto, si fa passare attraverso il foro, e
si fa riuscire al di sotto; si riuniscono quindi i
due spaghi al dorso.
I due spaghi così riuniti si fanno poi passare
corrispondentemente nell’anello, e si stringe il

tighten the knot at the spine. The sewing pat-
tern that joins the other four on top of the board
is simpler.
Take a thread length 60 cm long, and, with two
needles, following the same sewing pattern as

nodo sul dorso. La cucitura poi che deve unire
le altre quattro sul piano del cartone, è più
semplice.
Si prende un pezzo di spago lunga [sic] 60 Cent.;
s’infilano due aghi grossi ai rispettivi capi, e col

before, that is crossing the needles, pass them
through all the holes; having once reached the
last hole, form a knot on the spine.

Rome, 3rd February 1920 / (received 4th Febru-
ary from Enrico Castellani) / Henri Hyvernat.

medesimo sistema della prima cucitura, ossia
incrociando gli aghi, si passano da un foro

204. Copied from the papers in Henri Hyvernat files: MC VIII.A.3.

70



all’altro per giungere fino all’ultimo dove si
formerà il nodo di chiusura.
Roma, 3 febbraio 1920 / (remis le 4 fevrier par
Enrico Castellani) / H[enri].H[yvernat].

Let us consider one passage as example:

(i.β) Tie the two threads of each sewing station together in an open
circular knot; pass then the end of the bottom thread through
the hole in a downward motion; join then the two threads
together at the spine.

What follows is an almost verbatim translation that retains as much as possible

the flow of the original Italian description:

(i.γ) The two threads of each sewing station are taken and an open
knot is formed with them in the shape of a ring, the end of
the thread below is then passed through the hole, and is driven
out below; the two threads are then re-joined together on the
spine.

What this exactly means is not really clear, in either translation, as well as in

the original Italian text. There are not real points of reference and the motion of

the thread, as it develops from the bookblock sewing onto the board lacing,

however detailed the description might have seemed at the time of writing, is

rather confusing. A possible explanation of this particular kind of sewing might

be found in the drawings of Theodore C. Petersen of the Coptic bindings at the

Morgan Library in New York. See for example Figure 12.

This example shows how difficult it is to follow and understand structure from

a purely textual description.

4.2.2. A particular sewing pattern

Let us take a look at another verbal description. This time the original text is

in German and describes a particular sewing pattern on double or split bands

with the creation of a small knot between the two parts of the sewing support.

71



Figure 12. A sewing pattern for a Coptic binding at the Morgan Library, New York.
(Petersen 1948).

Klee first described this sewing in 1978.205 In the paper, Klee described the

pattern by verbal means, and accompanied her description by a photograph of

the spine of an actual book sewn in this manner. In 1989, Klee’s article was

translated and republished in English by Dorsey in the pages of the Binders’ Guild

Newsletter.206 Dorsey had the original article translated into English, and he

published both (ii.γ) a verbatim translation of the sewing pattern description and

(ii.β) the same text translated in a more fluent manner.

All three texts are reported here: (ii.α) the original German description of the

sewing pattern, and (ii.β; ii.γ) the two translations published by Dorsey.

(ii.α) Der Faden führt von der Lagenmitte durch den Doppelbund
nach vorn, die erste Bundhälfte wird umschlungen, dabei der
Faden hinter dem Bund zur zweiten Bundhälfte geführt, und
ehe der Heftfaden wieder zurück zur Bundmitte gestochen
wird, wird die erste Bundumschlingung noch einmal erfaßt
und der Faden gestrafft, wodurch der Knoten entsteht.

(ii.β) The thread goes from the inside of the signature through the
center of the split band, from back to front. It then goes tightly
around the center of the split band, from back to front. It then
goes tightly around one half-band and behind both half-bands.
Before the thread goes back between the half-bands into the
signature as is usually done, it first goes around the first half-
band again, then forward (from back to front) through the
slit, and back over itself, then into the signature. When
tightened, a knot is formed.

205. Klee 1978.
206. Dorsey 1989.

72



[a]

[b]

Figure 13. Two interpretations of Klee’s description of the so-called knot-tack sewing.
([a] Dorsey 1989, p. 12; [b] Wurfel 1989, p. 14).

(ii.γ) The thread guides from the signature middle through the
double band forward from back to front, the first band half
will become tight around while the thread behind the band
to the second half is guided, and before the thread is stitched
back to the band middle, the first band tight around is grabbed
again a the thread will be tightened, resulting in making the
knot.

Dorsey then adds a graphic representation of the sewing pattern to make

clearer his interpretation and to test the verbal description (see Figure 13[a]).

However, either Klee’s interpretation of the historical sewing pattern was flawed,

or the description did not deliver the message. In fact, Dorsey’s explanation, fol-

lowing the German text, is, by his own admission, not producing the same pattern

depicted in the photographs in Klee’s article.

73



A month later Wurfel,207 in answer to Dorsey’s article, proposed yet another

interpretation of Klee’s text based on Dorsey’s translation (see figure 13[b]).

In 2012, Benvestito208 proposed a new explanation of the sewing technique after

the discovery and examination of two volumes presenting the characteristic knot

in the sewing at the centre of the split band in the collections of the Marciana

National Library, in Venice, Italy. Through direct examination of the volumes

and ignoring the verbal descriptions that had accumulated over the years, she

reached a different conclusion (iii), but one that would seem consistent with the

photographs in the original article by Klee.

(iii.α) La cucitura knot-tack [...] viene eseguita in due tempi: l’ago
esce dal fascicolo, al centro del nervo tagliato, avvolge la
porzione destra del supporto e, passando dietro al doppio nervo,
compare sulla sinistra per dirigersi di nuovo verso il centro;
qui aggancia dall’alto verso il basso il filo già presente, lo serra
e, portandosi verso l’alto, rientra nel fascicolo [...] è evidente
che nella cucitura knot-tack non fosse prevista la com-
pensazione.

(iii.β) The knot-tack sewing is executed in two phases: the needle
exits the gathering in the middle of the split band, it wraps
the right section of the sewing support and, passing behind
the double band, it appears on the left and then goes back to
the centre. Here it locks with a downward motion the thread
already present, and it then re-enters in the gathering. [...]
Evidence shows that this was not a packed-sewing.209

Figure 14. Knot-tack sewing alternative solution (Benvestito 2012, p. 297).

These examples show how photographs alone are not able to convey enough

information about complex structures, and confirm the inadequacy of strict verbal

descriptions in communicating spatial information. The diagrammatic visualizations

207. Wurfel 1989.
208. Benvestito 2012.
209. Author’s translation.

74



that accompany the last three examples are however capable of integrating the

texts, communicating structure more successfully.

4.2.3. A controlled vocabulary description

The authors of the two verbal descriptions in the examples above made use of

natural language and described the binding and sewing techniques in a logical

but free manner. In the next example instead, Spitzmueller and Frost210 have de-

veloped a controlled vocabulary for the description of sewing patterns through

the fold of the gatherings. The result is a precise description of the path of the

thread with a rather ‘mechanical’ feeling to it. Examples (i), (ii), and (iii) are easy

to read, but, as seen, can be difficult to interpret, whereas, example (iv), with its

cumbersome asyndetic list of actions, is not easy to read, but, with some effort, it

can be precisely interpreted. Figure 15 shows a graphic representation of the

sewing described in example (iv).

Figure 15. Unsupported sewing sample (Spitzmueller 1982, 46).

(iv) Unsupported structure: chainpattern across spine; in-line,
periodicfold pattern. There are 4 sewing stations — A is at
the head and D is at the tail. 2 needles are used, each sewing
between 2 stations — A & B or C & D. They sew independ-
ently but identically. Enter at A, continue-on to B, exit, dropto-
the-outside, link, climb, enter at B, reverse, exit at A, dropto-
the-outside, link, climb, change-over and enter at A of section
2.

Natural language and controlled vocabulary are mixed in the description. In

the text, the terms prescribed by the proposed vocabulary have been underlined.

The terminology develops on three levels: (1) from the general sewing structure,

210. Spitzmueller & Frost 1982; Spitzmueller 1982.

75



(2) to the resulting sewing pattern, as seen both from inside the fold of the gath-

ering and from the spine, and finally (3) the sewing stitch. This third level describes

in detail the motion and path of the thread from a sewing station exit to its entry

in the next sewing station for books in codex-form sewn through the fold. Many

of the terms are traditional bookbinding terms, but where traditional terminology

lacked precision, the authors selected new terms or phrases to pinpoint verbally

the detail in focus. The sewing structure is described here as being unsupported

(unsupported structure), as it has only the sewing thread as a way to secure the

sections to each other. On the spine the sewing pattern takes the form of a chain

sewing (chain patterns across spine), while inside the fold it presents a periodic

pattern (periodic fold pattern), with intervals within some stations. All sections

have exactly the same internal pattern and the thread lengths appear between the

same stations (in-line). The thread path is described as entering at the sewing

station more towards the head of the bookblock (A), it then enters a station (B)

in the same gathering (continue-on), it exits on to the spine, it moves downward

(drop) towards the tail of the bookblock (to-the-outside), it passes under another

thread (link), it moves upward (climb) and re-enters at the station B, it continues

in the opposite direction of progression (reverse), it exits at station A, it moves

downward (drop) towards the head of the bookblock (to-the-outside), it passes

under another thread (link), it moves upward (climb) and enters station A in a

gathering that is different from the one exited (change-over).

While this verbal description does take some effort to read, it is nonetheless

able to convey more precisely than the previous examples the motion of the thread.

This is achieved through the use of the proposed controlled vocabulary, and it is

achievable because the nature of the structure described is essentially sequential

in nature. Following the motion of the thread step by step, in fact describing how

the sewing was carried out, this description dictates a series of actions to be taken

in order to reproduce the sewing.

This methodology and controlled vocabulary has been extended by Palmer

Elbridge,211 and used by Szirmai,212 integrating it with the descriptive approaches

211. Palmer Elbridge 1993; Palmer Elbridge & Prior 2008.
212. Szirmai 1999.

76



taken from the Dutch systematic vocabulary introduced by Gnirrep and col-

leagues,213 to describe a link-stitch process.214

This approach resembles the use by De Luca and colleagues215 of classical archi-

tecture structured and granular technical vocabulary included amongst the cultural

heritage visualization examples in chapter 2. Where one terminology uses actions,

the other uses precise figural concepts, but in both cases, the descriptions can

proceed strictly in a sequential manner.

4.2.4. Structured and controlled vocabulary descriptions

Other scholars have developed a more rigorous approach to the recording of

bindings and their structures. Abandoning the idea of natural language descriptions

altogether, they have turned to record the necessary information within databases

thus allowing for the implementation of a systematic method of examining a book.

This way, every book is described in the same manner, and this usually means

following how the object was constructed, from the inside out, from the formation

of the gatherings, through the codex assembly structure, to the covers and their

decoration.216 The normative nature of these descriptions guides both the compiler

and the reader through a commonly understood hierarchy of information, thus

solving, up to a point, some of the problems found in prose-based descriptions.

An example of this methodology of description is the XML schema developed

by Ligatus. This is a good example of the use of a hierarchical database for storing

data on bookbinding structures and their state of conservation.217 The hierarchical

structure, running from the more general to the more specific, allows book com-

213. Gnirrep et al. 1992.
214. ‘The thread, proceeding in the centrefold, exits through a sewing hole at a given sewing

station where it drops in order to make a link under the sewing thread of the previous quire [...]
it then climbs and re-enters through the same hole and makes a long stitch to the next sewing
station’ (Szirmai 1999, p. 16). Underlined words are from the Spitzmueller and Frost (1982; 1982)
controlled vocabulary, italic words are from Gnirrep and colleagues (1992).

215. De Luca et al. 2005; De Luca 2013.
216. Federici & Houlis 1988; Federici et al. 1988; Houlis & Pescalicchio 1988; British Census

Project 1990; Federici & Pascalicchio 1993; Grosdidier de Matons et al. 1993; Sharpe 2000;
Pickwoad 2004.

217. Ravenberg 2012.

77



ponents and their characteristics to be arranged in groups, and then an unlimited

number of groups within these groups, down to the desired level of detail.218 The

records are combined with photographs of the volumes and freehand drawings

of the binding structures.

Listing 1 and Figure 16 show an example of an XML description for a sewing

station of a printed book volume from St Catherine’s Library along with detail of

the photograph of the spine showing an exposed sewing support and drawing of

the sewing type:

Figure 16. Photograph and drawing of sewing for Vol. 126.60β, St Catherine’s Library.

218. Ravenberg 2012.

78



<sewing>
    <numberOfStations>7</numberOfStations>
    <status>
        <firstSewing/>
    </status>
    <type>
        <allAlong/>
    </type>
    <stations>
        <station> ... </station>
        <station>
            <measurement>36</measurement>
            <group>
                <current/>
            </group>
            <preparation>
                <NK/>
            </preparation>
            <numberOfHoles>1</numberOfHoles>
            <type>
                <supported>
                    <type>
                        <double>
                            <route>
                                <drawingDone>
                                    <no>1</no>
                                </drawingDone>
                            </route>
                            <linking>
                                <linked/>
                            </linking>
                        </double>
                    </type>
                    ...
                </supported>
            </type>
        </station>
        ...
    </stations>
    ...
</sewing>

Listing 1. Snippet of XML description of a sewing station from codex 126.60β, St Cath-
erine’s Library.

79



The XML schema functions as a grammar for the description of the binding,

it is a model of the binding of books. Elements described by verbal means within

the hierarchical structure of the XML records are readily available for machine

searching, making it easy for specific pieces of information to be found and com-

pared. In these descriptions, information is organized into hierarchies, and within

these hierarchies, tokens of information that together serve to describe a particular

feature or part are grouped and presented conjointly. This allows significant bits

of information to be brought forward that carry with them all the semantic inform-

ation contained in the structure.219 By grouping the information in this manner,

it is possible to provide elementary details of the configuration of the material

components being presented within the description. This will be further analysed

in chapter 7 which presents the empirical data for the transformation of structured

descriptions of bookbinding structures into diagrammatic visualizations.

The fragmentation of the information within the fields of the XML-based

database description is an obstacle for a human reader attempting to analyse and

compare the data. A reader would have to muster in his mind — and more precisely

in his working memory buffer — a number of fragmented pieces of information;

the more components come together in a given structure, the more pieces of in-

formation would have to be kept in mind, analysed, and subsequently joined into

a coherent entity. As noted in the introduction, our working memory can hold

only a very limited amount of information220 — some scholars221 mention three to

five chunks of information, others222 up to seven or nine — and the fragmented

information contained in databases can consequently be difficult to reintegrate

into a mental conceptualization and visualization utilizing the structured verbal

description alone. The working memory buffer can be freed by recurring to ex-

ternal memory devices, or exograms,223 thus allowing memory space for analysis

219. Brachman & Levesque 2004; Gnoli 2008.
220. Baddeley & Hitch 1974; Baddeley 1992; Baddeley 2004; Alloway & Alloway 2012; Bor

2012; Baddeley 2013.
221. Cowan 2001; Gobet & Clarkson 2004.
222. Miller 1956; Saaty & Ozdemir 2003.
223. Donald 1991; Donald 2001; Alloway & Alloway 2012.

80



and comparison. A system to gather the fragmented data and present it in a coher-

ent unified manner to the reader could easily work as a suitable exogram.224

From these examples it is clear that natural language alone is not capable of

communicating specific and detailed descriptions to allow a reader to derive one

and only one possible interpretation of what was described.

Also, there is another crucial problem with natural language descriptions: that

of the translation into other idioms, e.g. English to Italian. As each language has

its own grammatical rules and terminologies, the passage of information can create

indeterminateness. The problem of the translation and verbal imprecision is

lessened by the use of controlled vocabularies and multilingual thesauri as they

fix the concept and not the labels for it.

4.3. Prototypical visualization of bookbinding elements

In 1928, Goldschmidt225 pointed out that the literature on bookbinding can be

divided into three main kinds: (i) books of plates, (ii) bookbinding manuals and

handbooks, (iii) and highly focussed articles in a number of journals that need a

good corpus of reliable published material for reference and comparison. The

current situation is still virtually the same.

Books of plates are plentiful, though for the most part concentrated on decorated

bindings and the reproduction of just the exterior of bindings. Manuals and many

articles make often ample use of schematic drawings as well as photographs to

help interpret and categorize bookbinding structures. Not all drawings, however,

are informative enough and suitable for communication, and different styles are

useful for delivering different kinds of information. The lack of a standard way

of depicting bookbinding structures can be linked to the fact that the bookbinding

trade is based on craftsmanship. Artisans, unlike engineers for example, do not

need to show a vision of a structure for others to then construct it, but rather they

224. Kirsh 2002.
225. Goldschmidt 1928.

81



visualize it in the their mind and then will construct it themselves. Therefore,

there was never the need to develop a graphic language to communicate book-

binding structures efficiently to others.226 In addition to this, unlike archaeology,

the young age of the discipline has not allowed for any standardized drawing

methodologies to be established.

There is, therefore, no standard practice to illustrate bookbindings and their

structures, and different authors employ different styles and conventions; in fact

it is not uncommon to find different conventions used by the same author within

one publication. Authors of conservation and bookbinding manuals have made

use of professional illustrators for their publications,227 but otherwise, for the most

part, the scholars are responsible for both the text and the illustrations. As men-

tioned, archaeology has developed standardized graphic conventions (see Figure

17 for examples of archaeological drawings), but these are seldom applied to

bookbinding studies.

4.3.1. Gathering assembly diagrams

One simple structure that is common to any book in codex form, and that often

needs to be graphically represented by scholars dealing with the range of subjects

that have the book as their object of study — e.g. codicology, palaeography, bib-

liography, or bookbinding — is the gathering or section assembly.

Different conventions are used by different scholars to represent the pages of

a book within a gathering. Different uses and kinds of information do call for

different conventions and spatial layout, however, it can be argued that there is

scope for a more standardized approach to the representations of such basic book

structures, one that takes into account perception and prototypification of shape

on the one hand, while allowing for special cases, if needed, being general and

prototypical enough on the other.

226. Ferguson 1992.
227. Cockerell 1953: with drawings by Noel Rooke and other illustrations; Cockerell 1958: with

illustrations by Joan Rix Tebbutt.

82



Figure 17. Archaeological drawings. [a] Examples of standard practice to illustrate pottery
with elevation and surface details on the right-hand side and a regularized section in
solid black on the left. Examples of texture rendering: [b] plain and patterned bronze
piece; [c] cone of dark glass with stipple and dot technique to show texture and shading
together. Surface texture is used to enhance appearance and value of a drawing, but care
is taken not to confuse structural lines and texturing. The drawings here reproduced are
not in scale ([a] Webster 1964, fig.1, p.35; [b] and [c] Brodribb 1970, fig.24–25, pp.42-
43).

Figure 18 and Figure 19 show some graphic representations of gatherings taken

from various books and articles published by scholars ranging from bibliographers

and codicologists, to conservators and bookbinding historians. In the first group,

the author of each drawing has decided to use a naturalistic three-dimensional

approach to represent a gathering, while in the second group a more symbolical

approach has been taken.

4.3.1.1. Naturalistic representations of gatherings

There is no doubt that a naturalistic and three-dimensional representation is

more impressive and, maybe even pleasing to the eye when well-executed. How-

ever, the scope of such representations, it can be argued, is not that of showing

83



Figure 18. Naturalistic and three-dimensional representations of gatherings. ([a] Gaskell
1972, fig. 45, p.83; [b] Federici & Houlis 1988, fig. 11, p.22; [c.1-2] Sheppard 1995, p.
193; [d] Clarkson 1996, fig.4, p.218; [e] Regemorter 1992, p. 141).

how a gathering might look like in general, rather, the aim is to present the reader

with some information regarding the specific gathering represented, and to make

it comparable with other gatherings.

4.3.1.2. Schematic representations of gatherings

The collection of drawings in Figure 19 shows a series of schematic represent-

ations of gatherings. In these drawings though, the balance between the iconic

nature of such representations and their inherent symbolism generally leans greatly

towards symbolism, often disregarding almost completely the natural shape of

the object represented. Thus the pages of a gathering are represented by V-like

signs on the one hand (Figure 19[a]), and square or rounded horizontal U-like

signs on the other (Figure 19[a-d]). The problem with the V-like representation

is that, although easy to draw, the overall shape of the gathering tends to either

fan out (like in Figure 19[a]) or to distort the shape of the inner pages (that become

84



Figure 19. Schematic representations of gatherings. ([a] Muzerelle 1985, fig. 56; [b] Noel
1995, fig. 11, p. 8; [c] Szirmai 1999, fig. 9.2[s], p. 179; [d] Szirmai 1999, fig. 8.4[e], p.
147).

smaller and smaller), or else these are unnaturally projected forward as each sheet

representation maintains the same sharp V-shaped fold and size. Either way, apart

from being particularly inelegant and cumbersome, these representations, unnat-

ural as they are, make it particularly difficult to compare representations of differ-

ent gatherings and do not take into consideration the objects they are trying to

represent. This is because, even if a single sheet of paper, when folded, does create

a V-shaped fold, when multiple sheets are folded together, the thickness of the

other sheets forces the V-shaped fold to round up creating a more gentle U-shape.

Another solution often used is that of foregoing altogether with any naturalistic

representation and using square open boxes to symbolize each sheet. This can be

seen in Figure 19[b] and [c]. In this case the length of each sheet is directly pro-

portional to that of the sheet represented, but the shape of the fold is completely

unnatural. Although there is nothing wrong in deciding to use a totally symbolic

shape, the choice does not have to compromise the possibility of representing

particular cases and it should not be possible to confuse such elements with others

85



of the same structure. It is indeed easier to draw straight lines and square angles,

rather than nicely shaped curved lines, but, such a shape should not be confused

with anything else. See for example the diagram in Figure 19[c]. This drawing

comes with a very useful legend explaining how to read and interpret each element,

however, on a closer look, because of the choice of drawing the gatherings with

a square spine, the legend and the drawing result in being confusing: if the symbol

for the board is an open box with square corners, one might read the drawing as

describing a strange book made up of a series of boards. Obviously this is not the

case, but the square gathering spines create a possible problem in the interpretation

of the drawing. See how this is not the case in Figure 19[d], where the pages of

the book are represented by nicely rounded shapes.

As in many structures that form a codex, there is a great number of possible

permutations of how a gathering structure can present itself, depending on the

number of sheets and their configurations. Bibliographers and manuscript scholars

have developed precise and highly condensed collation formulas to communicate

this information,228 however it is often necessary to also visualize these for easier

communication and understanding of more complex structures. Because of this,

and since drawing appropriate shapes by hand can be difficult, it would seem that

a system that allowed an automated generation of such shapes and diagrams would

be welcome. Better shapes could be drawn this way, and the process would not

be as time-consuming as drawing them by hand. Recently, the issue has been

raised in a few occasions in the community. 229

To conclude, an ideal shape should be able to accommodate particular cases,

and maintain consistency at all times. For elements of an object’s structure — as

in the case of bookbindings — keeping a good balance between the iconic and

the symbolic nature of the representation, and preserving the prototypical shape

of its object, would allow to generate a standardized approach that takes into

consideration perception and prototypification of shape, and is symbolic enough

to allow straightforward comparison between distinct representations. For the

228. Bowers 1949/1994; Gaskell 1972; Gruijs 1974; Muzerelle 1985; Zappella 1996.
229. See for example Wragg 2012; Porter 2013.

86



particular case at hand, to represent a gathering, a better choice would be a cross

section view that preserves the natural roundness of the sheets at the fold, like

the example in Figure 19[d].

4.4. A categorization of bookbinding illustrations

In the literature on bookbinding, different styles of drawings are used for dif-

ferent reasons and to deliver a diverse range of information. Each style is appro-

priate for a specific set of uses, and it should be chosen according to the kind of

information that one needs to deliver.

Each different style can be regarded as a step in a categorical scale, going from

very specific and exclusive, to general and inclusive, just as words and concepts

within a categorical system, as presented in the previous chapter. A representation

of the categorical scale is reported in Figure 20. As mentioned in chapter 3, line

drawings represent a basic level of graphic abstraction, as they preserve the min-

imum amount of information that is necessary to precisely represent the category

of an item.

4.4.1. Archaeological-style drawings

Some bindings are so particular and unique that it is necessary and good practice

to present the information gathered on them by both photographic means and

detailed drawing. This is common practice for archaeological finds and it is not

surprising that drawings executed for these purposes make use of the same set of

graphic rules usually implemented by archaeologists to depict their finds for

publication.230 For this reason these drawings will be referred to as archaeological-

style drawings. Some examples of this kind of illustrations are presented in Figure

21. As it can be seen, the unique peculiarities, and the history, the signs left by

230. Brodribb 1970; see also Figure 17 for examples of archaeological drawings.

87



Categorical dimensions
of drawings

L
ev
el

of
ab

st
ra
ct
io
n

Segmentation of category

scene

generic shape

complex

naturalistic drawings

archaeological-style drawings

line drawings
schematic

line drawings

Figure 20. Categorical dimensions of drawings. The basic level of abstraction, in bold, is
represented by line drawings; these can be complex or schematic.

time on the items depicted, are preserved and faithfully reproduced in the illustra-

tions, and no attempt is made to generalize shapes and characteristics.

As noted earlier, there are no standards and guidelines in the literature on how

to draw bookbindings and their structures. Sharpe231 sets some general guidelines

on how to approach and execute archaeological-style drawings, with direct refer-

ences to drawing manuals for archaeologists and scientific illustrators.

Drawings of this style represent a specific object. As seen in the previous chapter,

verbal descriptions are not capable of conveying the specificity of an object. It is

thus not possible to generate automatically archaeological-style drawings from

verbal descriptions.

4.4.2. Naturalistic drawings

Sometimes it is appropriate to describe graphically a specific item of a specific

category with very detailed and naturalistic drawings, either because of its

uniqueness, or as a first step in drawing general conclusions on the category being

231. Sharpe 2000.

88



[a]

[b]

[c] [d]

[e] [f]

[g.1] [g.2] [g.3] [g.4]

Figure 21. Archaeological-style drawings. The history and the peculiarities of the item
depicted are preserved and faithfully reproduced. ([a] Szirmai 1988, fig. 3, p. 24; [b]
Clarkson 1993, fig. 1, p. 184; [c] Szirmai 1988, fig. 8[a], p. 30; [d] Szirmai 1988, fig. 8[c],

89



p. 30; [e] Adler 2010, fig. 3–02, p. 61; [f] Adler 2010, fig. 3–07, p. 62; [g.1-4] Szirmai
1988, fig. 9[a–d]).

represented by the item. These will be referred to as naturalistic drawings. Ex-

amples of illustrations of this kind are presented in Figure 22. These illustrations

tend to be a kind of graphic reconstruction of specific items, that, while preserving

a high level of detail and surface texture, come closer to a generalization of shape,

purposely removing signs of time — e.g. asymmetries are mostly removed and

converted into symmetries — and showing an idealized graphic description. The

drawing approach of these drawing resembles that of highly detailed scientific il-

lustrations.232

Also in this case, these drawings tend towards the specificity of an obejct, and

it is therefore not possible to generate them automatically from verbal descriptions

without additional information.

4.4.3. Line drawings

The next step up the graphic categorical scale is represented by line drawings.

As mentioned above, these are able to convey shape information precisely and

for the most part effortlessly. They also represent a graphic equivalent to the basic

level of abstraction of verbal categories. Line drawings, in fact, are capable of

delivering the minimum amount of information that is necessary to represent

graphically the category of an item.233 They are inclusive enough to be successfully

used to communicate shape, and exclusive enough not to represent too wide a

group of instances of their category. Figure 23 shows examples of line drawings.

These drawings, all highly selective in the amount of information delivered, can

further be subdivided into two groups: on the one hand, one can distinguish multi-

layered and three-dimensional drawings ([a], [b], [c], and [d.2]); on the other

hand, there are schematic and strictly bi-dimensional drawings ([d.1], [d.2], and

[e]). The two kinds of line drawings can be seen as two variations of the same

232. Wood & McDonnell 1994.
233. Rosch & Mervis 1975; Rosch et al. 1976; Tversky & Hemenway 1984.

90



[a]

[b]

[c]

[d]
[e]

Figure 22. Naturalistic drawings. The general shape of the item is reconstructed and
faithfully reproduced with abundance of details and surface textures. ([a] Petersen 1954,
fig. 8, p. 46; [b] Clarkson 1993, fig. 5, p. 196; [c] Petersen 1954, fig. 22, p. 55; [d] Clarkson
1993, fig. 12, p. 190; [e] Boudalis 2007, fig. 58, p. 44).

91



basic level of abstraction within a graphic categorical system, developing its hori-

zontal dimension. The former group will be referred to as complex line drawings,

whilst the latter as schematic line drawings. Both groups are generated by a process

of generalization of shapes that aims at describing graphically a prototypical in-

stance of a category.

If an object is described verbally at the basic level of abstraction, the terms are

capable of delivering enough information for the generation of schematic line

drawings conveying information on the parts of an object and its prototypical

shape.

4.4.4. Generic-shape drawings

A line drawing can also only portray the very generic shape of an item, foregoing,

once again, any naturalistic detail, on the one hand, but also limiting the amount

of information on shape to the bare minimum with very little attention to details,

on the other. These shapes are too general to communicate the necessary inform-

ation to distinguish between specific items, as they fail to preserve the parts-form

relationship of its object and its functions — with the exception of the very basic

elements — and are the graphic equivalent of generic verbal concepts like furniture

as opposed to table. This particular kind of line drawing will be referred to as

generic-shape drawing. Figure 24 displays some examples of such illustrations.

Note for example the books in the timeline [a]: the drawing depicts the evolution

of the book in codex form from the 2nd century CE to the 20th century, but the

shapes are so generic that very little information is given on the actual differences

between the different binding styles.

Drawings of this style could be generated from generic verbal labels, but, as it

will be seen in chapter 6, the shapes thus generated would be too inclusive to be

useful and carry more information than the verbal label alone.

92



[a]

[b] [c]

[d.1] [d.2] [d.3]

[e]

Figure 23. Line drawings, the graphic basic level of abstraction. These drawings can be
multi-layered and three-dimensional — but still highly selective in the amount of inform-
ation delivered — or more schematic and strictly bi-dimensional. ([a] Clarkson 2005, p.
22; [b] Clarkson 2005, fig. 4, p. 8; [c] Pickwoad 2000, fig. 27, p. 158; [d.1-3] Szirmai
1999, fig. 9.22[a–c], p. 207; [e] Szirmai 1999, fig. 7.16[a], p. 116).

93



[a]

[b]

[c]

Figure 24. Generic-shape drawings. The shape of the drawn items is so generic that it
does not show much more information than identifying the object for what it is — e.g.
a book in codex form. Note for example the books in the timeline [a]: the drawing depicts

the codex from the 2nd century CE to the 20th century, but very little information is given
on the actual differences between the various binding styles. ([a] Greenfield 1998, p. 79;
[b] Greenfield 1990, p. 78; [c] Greenfield 1990, p. 79).

94



4.4.5. Scenes

Sometimes bookbindings are depicted within pictorial scenes, like the small

watercolours found in Dirk de Bray bookbinding manual.234 The depiction of

books in these scenes can be very useful to analyse how historical bindings were

bound, or how these bindings worked and behaved, but often, the shapes used

to depict them are generic, deformed by perspective representation, and frequently

misshapen or partially hidden to accommodate other elements in the scene. For

these reasons, they can be put at the very top of the categorical scale of bookbind-

ing drawings. It should be noted, however, that not every pictorial depiction of

books within scenes is so generic, and they need to be evaluated case by case.

These depictions will be referred to as scenes.

Drawings of this style are part of the narrative of a scene and are inherently

visual in their nature. It is doubtful how useful it would be to try to generate them

from verbal inputs.

The different kinds of graphic descriptions of bindings found in the literature

can then be classified according to a categorical scale, going from very specific

and exclusive drawings, to general and inclusive ones, within the vertical and ho-

rizontal dimensions of a categorical system. The various steps within such a scale

have been clearly separated here and examples have been presented for each level

of shape abstraction. Of these, schematic line drawings represent an ideal level

of abstraction for the generation of diagrams from verbal input.

4.5. Clarity of information in bookbinding line drawings

So far, this chapter has analysed and categorized the kinds of bookbinding il-

lustrations found in the literature. Each type can be successfully used to deliver

information for different purposes. Obviously, there are degrees of success in the

use of drawings to illustrate bookbindings, but the primary purpose of graphic

234. Bray 1607-1658; Bray 1658/1977.

95



Figure 25. Bookshop scene: ‘Mensen en twee honden in een boekenwinkel’ (People and
two dogs in a book shop), pen and brown ink, brush in grey on paper, Dirk de Bray,
1607-1658. (Bray 1607-1658).

material — just as writing — should be to convey the right amount of information,

with as much clarity and intelligibility as possible.

There are some general rules that can be learned from illustration in general,

and archaeological drawings in particular:235 (i) edges of boxes and geometrical

shapes should be kept neat and tidy; (ii) the mixing of many different styles togeth-

er should be avoided; (iii) labels should be non-intrusive, clearly legible, and

properly differentiated from the other elements of the illustration; (iv) proportion

and scale should be handled intelligently; (v) important information should be

235. Brodribb 1970; Ford 1993; Tufte 1990; Agrawala et al. 2011; Bertin 1983/2011.

96



emphasized, while irrelevant details should be de-emphasized or eliminated alto-

gether; (vi) whole sets of information and different elements should be kept clearly

distinct and differentiated, and (vii) the visualization should consider the viewer’s

perception and cognition of the information that is meant to convey.

The next couple of sections will consider a few examples from bookbinding

illustrations and analyse whether these practical rules have been followed. Chapter

5 will cover more in details the set of general rules for an effective visual commu-

nication.

4.5.1. Sketchiness and detail views

The use of sketches in published material could be considered an editorial

choice. Nevertheless, a well-executed and regularly shaped illustration, not only

looks more professional, but has also the ability of transmitting information more

efficiently. Look for example at Figure 26 and Figure 27. Within the same public-

ation, Szirmai236 makes use of both well-executed and well-shaped drawings and

sketchy free-hand ones. Note how clearer the sewing thread paths appear in ex-

amples [a] and [b] in Figure 26 as opposed to those in examples [c-h]. This situ-

ation becomes even more evident in the example in Figure 27 from Carvin’s book.237

Here the drawing appears as if it were executed directly on a drawing software

with a mouse or similar device resulting in rather approximate shapes. On top of

this, the author mixes schematic and more naturalistic elements in a confusing

manner. For example, the gatherings are drawn as V-shaped with a single thin

line representing the folds (as it was the case in Muzerelle’s drawing in Figure

19[a]), thus leaving no space for the thickness of the gathering and of the thread.

Because of all these problems, the drawing as a whole is rather difficult to interpret.

The use of sketches is clearly visible in the examples in Figure 28. Here the

author238 makes use of untidy line drawings, but she has also decided to use a scale

236. Szirmai 1999.
237. Carvin 1988.
238. Greenfield 1990.

97



[a]

[b]

[c] [d] [e] [f]

[g] [h]

Figure 26. Well-shaped and free-hand thread paths. ([a-b] Szirmai 1999, fig. 9.8[c-d],
p. 188; [c-h] Szirmai 1999, fig. 9.9[a-f], p. 189).

Figure 27. Drawing of an endband sewn at the frame executed directly on a computer
in 1988. (Carvin 1988, p. 50).

98



i ii iii

iv v vi

vii viii ix

x xi xii

Figure 28. Sketch-like illustrations of the steps required to work a Greek-style endband.
Note the untidy nature of the geometrical shapes, the choice of using a scale that portrays
the whole binding for the majority of the steps, and the difficulty in distinguishing the
different elements of the endband (compare with Figure 30). For this endband the author
has decided not to make use of cross-section schematic illustrations to show the route of
the thread around the cores, although such schematizations have been used for other
endbands in the same book (see Figure 29). (Greenfield 1990, pp. 51-56).

Figure 29. ‘Step 11: go up around B and down behind cores B and C, coming back under-
neath them.’ Here a step in the description of an Armenian-style endband sewing is illus-
trated by use of both a complex line drawing and a schematic line drawing. (Greenfield
1990, p. 61).

99



Figure 30. Line drawings of different styles of endbands. Note the use of regular and tidy
shapes, the choice of using an appropriate scale that shows only a well-selected portion
of the binding, and the ease of distinguishing the different elements of the endband.
(Bibliothèque Nationale (France) 1989, pp. 59, 63, 87).

that portrays the whole binding for the majority of the steps illustrated for the

working of a Greek-style endband. Because of this choice, the actual shapes that

should illustrate the process are too small and confused to be followed by the

reader. The reader of a manual on the sewing of endbands, it can be argued, would

know where on a binding the endband should be sewn and therefore there is no

reason to show the whole book.

Figure 30 shows another example of the sewing of a Greek-style endband taken

from a different manual.239 Here the authors have decided to employ regularly

shaped line drawings using an appropriate scale that shows only a well selected

portion of the binding. Note the ease of distinguishing the different elements of

the endband.

239. Bibliothèque Nationale (France) 1989.

100



In the literature, therefore, one can see different styles of drawings, and these

are sometimes not carefully executed or designed. Considering the importance of

visual information for bookbinding structures, and the problems linked with

natural language descriptions, this can lead to unclear and indeterminateness even

with a multimodal approach that joins verbal and visual depictions.

The shapes to be generated for this project will aim at showing the minimum

amount of information at a useful level of abstraction, just as verbal information

should be provided at the correct semantic entry point to convey a useful level of

information. Line drawings have been identified as the basic graphic level of ab-

straction. The aim is to be able to provide information visually in such a way that

it can be easily compared between instances of structures belonging to the same

category. Considering how three-dimensional representations tend to distort some

details and shapes, making it less immediate to recognize similar patterns,240 the

drawings of choice will be simple line drawings. These will depict items through

the most useful views, be this a general, or a detailed view, a cross-section, etc.,

or, in fact, a combination of views.

Summary

Information on the elements and form of objects is best disseminated and

analysed by means of a multimodal approach that capitalizes on the human ability

to categorize information and to process visual information. Drawing is an import-

ant part of the surveying, recording, and description of bookbinding structures.

Bookbinding structures have been described in different ways in the literature.

Of these, natural language appears to be unable to convey information that can

be interpreted in only one way, creating interpretation problems unless associated

with some kind of visual information that can aid in interpreting the spatial con-

240. Ferguson 1992; Ware 2013.

101



figuration of the components. Controlled vocabularies and structured descriptions

instead seem to be able to provide spatial information in a more reliable manner.

Of the various kinds of illustrations found in the literature on bookbinding,

line drawings, and in particular the schematic kind, appear to be a good choice

as they are able to convey the necessary amount of information, while being simple

enough to be readily interpreted, understood, and remembered.

These diagrams abstract the essential and hold a great deal of information for

interpreting and categorizing bookbinding structures, and, at the same time, are

easily referenced to and compared with other examples.

These diagrams show the elements that make up a certain object, the number

of such elements, and their spatial arrangement, thus easily conveying also the

object’s form. Being schematic, they are straightforwardly enriched with symbols

to deliver other kinds of information, like the material of certain elements, or

functional information (by the addition of arrows).

The next chapter looks more in detail into the descriptions of material objects,

and discerns which verbal descriptions are more suitable for an automated interse-

miotic translation. It then considers the peculiarities of the visual language, and

the relationship between a representation and the reality it refers to.

102



Chapter 5. Communicating & translating
objects

La structure [...] est terrible. On ne peut la supplier, lui dire:

«Voyez comme je suis mieux que H...» Inexorable, elle répond:

«vous êtes à la même place; donc vous êtes H…» Nul ne peut

plaider contre la structure.

The structure [...] is terrible. It can not be implored, saying:

‘Look how I am better than H...’ Inexorable, the structure replies:

‘You are in the same place; therefore you are H...’ No one can

argue against the structure.*

Roland Barthes, Fragments d’un discours amoureux: Identifica-

tions, 1977, pp. 154-155. *Author’s translation

This chapter looks into the languages involved in the project, and considers

what kind of verbal description best conveys the relevant information so that it

can be transformed into a diagrammatic visualization. The term language refers

to any system of representation and communication,241 and in this project the

languages involved are both verbal and visual means of communication.

The first part analyses the examples of bookbinding structures described in the

previous chapter and advances the reasoning for the selection of a structured de-

scription as the most useful model of the structures for this project. The second

part looks at the implications of the transformation of information into a visual

241. Stenning & Lemon 1999.

103



language. The last part covers the issue of uncertainty in the data and in the visu-

alizations.

5.1. Communicating material objects through verbal means

In chapter 3 we have seen how in order to be able to describe a material object,

one has to be able to communicate its material components, its form, and its lo-

gical form. In order to be able to do so, one needs to understand the object com-

pletely,242 i.e. to understand and predict all possible ways in which its components

could be in relationship with one another.243 This leads to important considerations

in regard to the description of material objects.

In the previous chapter, three different kinds of verbal descriptions of book-

binding structures were analysed: (i) natural language descriptions, (ii) controlled

vocabulary descriptions, and (iii) structured descriptions with controlled vocabu-

lary. The first is not capable of successfully conveying the form of the binding

structure. Eide244 describes a similar case for textual maps. He refers to this kind

of description as under-specified, as based on such natural language descriptions,

more than one map can be drawn, and often, these maps can be significantly dif-

ferent. In a similar way, Peirce245 referred to the same concept of under-specifica-

tion as the development of a sign: the meaning of a sign can be translated into

some other sign in which is it more fully (or indeed, less fully) developed. In fact,

for Peirce, an iconic sign does not necessarily have to be an image, but a verbal

text can indeed have iconic characteristics, ‘but the icon is not clearly apprehen-

ded’,246 as the passage of information from visual to verbal is less fully developed

than in proper iconic signs. Natural language descriptions are thus under-specified

and under-developed, leading to the generation of more than one possible inter-

pretation.

242. Wittgenstein 1922/2012T: 2.01231.
243. Wittgenstein 1922/2012T: 2.0124.
244. Eide 2012a.
245. Peirce 1931-1935/1958: CP 5.594; Dewey 1946.
246. Peirce 1931-1935/1958: CP 7.467.

104



The second kind seemed to be a successful mode of communication, though

only for linear structures whose sequential spatiality was mirrored in the sequential

nature of the text. Structured descriptions with a controlled vocabulary, instead,

with their inherent hierarchical structure and grouping of information might have

the ability to describe binding structures successfully, and also for complex non-

linear structures.

Controlled vocabulary descriptions, both alone, and within structured descrip-

tions, try to rationalize what material components are found within a binding

structure. By prescribing what kinds of components can be found within a certain

binding structure through the selection of a precise set of terms, controlled

vocabularies provide both the person describing such a structure, and the reader

of the description, with all of the components that come together to form it. Also,

by precisely defining each term, they try to provide all properties (or better those

properties deemed important and/or essential) for each component. Therefore,

by defining the elements of a material object and its properties, controlled

vocabularies have the potential capacity to describe an object effectively, even for

cases that are not yet encountered. ‘If I know an object I also know all its occur-

rences in states of affairs. (Every one of these possibilities must be part of the

nature of the object.) A new possibility cannot be discovered later’.247

5.1.1. Communicating through controlled vocabularies

Verbal communication is renowned for its difficulty in conveying spatial inform-

ation. Words, because of their sequential nature, are not efficient in communicating

non-linear spatiality.248 Spatiality can be defined as the characteristic of reality to

occupy space as an autonomous form of organization in which multiple distin-

guishable elements can coexist.249 Spatiality and the structural interrelations of

247. Wittgenstein 1922/2012T: 2.0123.
248. Stenning & Lemon 1999.
249. Saint-Martin 1990.

105



elements within an entity are difficult to pin down through linear and sequential

verbal communication.

Words, by working within a categorical labelling system are generally not well-

suited to capture the specificity of an object.250

As seen in the analysis of related works in chapter 2, spatiality and specificity

posed problems for most projects, when such information could not be gathered

through means other than verbal communication. Of the cultural heritage projects,

the one presented by De Luca and colleagues251 seemed particularly interesting,

as it was able to convey, through the use of the formal terminology of classical

architecture, both information relating to the material components of the objects

represented and their form.

The reason for the success of this project is surely to be found precisely in the

fact that it was able to use as verbal input a very particular form of technical

vocabulary. Goulette and Borillo252 have researched architectural composition

and the semantics of spatial expressions within the vocabulary of classical archi-

tecture. They have found that this technical vocabulary is able to convey spatial

and structural information in the form of whole/part relationships, cognitive

spatial relationships sustained by the functional information about the parts and

the various objects, and on regular models of compositions. Together, this inform-

ation can assist a model of spatial reasoning that is able to represent architectural

information conveyed through verbal means. According to this model, the con-

ceptualization of the spatiality of the elements is not limited to geometric aspects,

but it is rather guided by considerations on the functional organization of each

element and the relative disposition of its borders. The fundamental relationships

of this model are: the relation of parthood (x is part of y), and the relation of

‘border of’253 (x is the border of y). Through the expression of these relationships,

the understanding of the elements and their function, and of the rules of compos-

ition that are part of the architectural grammar controlling the relative positions

250. Stenning & Lemon 1999; Saint-Martin 1990.
251. De Luca et al. 2005; De Luca 2013.
252. Goulette 1999; Borillo & Goulette 2006.
253. Borillo & Goulette 2006, p.51.

106



of each class of elements, a verbal description is capable of conveying relative but

adequate spatial and structural information.

Goulette254 has also identified three types of entities in the vocabulary of classical

architecture: the architectural elements, their spatial references and functions,

and their geometric representation. Each term within this vocabulary is capable

of conveying all of this information, as was seen in the case of the elements of the

column profiles presented in chapter 2.

It would then seem that the use of a fully-developed controlled vocabulary can

communicate enough information about a material object: its material components,

its structural organization, its logical form, and its shape. The material components

would be defined as terms, and the definitions would also be providing information

on its function and, therefore, spatial references, and on its geometric representa-

tion; the rules of composition would function as a reference to its logical form.

Bookbinding terminology is not as well-developed, and it generally does not

convey precise geometric representational information; nonetheless, controlled

vocabularies have the potential to be powerful communication devices.

5.1.1.1. Names and prototypical material components

Verbal communication of material objects and their components is based on

names, and names are a type of sign.255 These names can be defined, and when

so, they are necessarily generalized concepts and therefore prototypical signs. As

a consequence, these signs exhibit a certain degree of indeterminateness.256Proto-

types, in fact, as seen in chapter 3, usually coincide with the basic level of abstrac-

tion within a category257 and are, therefore, inclusive and general. It is from their

inclusiveness that the indeterminateness arises in regard of their specificity, e.g.

the specific shape of a material component. In addition, prototypical signs can

only offer a selection of the properties of the object to which they refer.

254. Goulette 1999.
255. Sebeok 1994.
256. Wittgenstein 1922/2012T: 3.24; T: 5.522.
257. Rosch 1978.

107



The very essence of a controlled vocabulary is the definition of its terms.

Therefore, a controlled vocabulary should aim at defining the simplest elements

possible, so to reduce the indeterminateness to a minimum.

5.1.1.2. Sewing structure controlled vocabulary

Let us look again at the earlier controlled vocabulary example:

Unsupported structure: chainpattern across spine: in-line,
periodicfold pattern. There are 4 sewing stations — A is at
the head and D is at the tail. 2 needles are used, each sewing
between 2 stations — A & B or C & D. They sew independ-
ently but identically. Enter at A, continue-on to B, exit, dropto-
the-outside, link, climb, enter at B, reverse, exit at A, dropto-
the-outside, link, climb, change-over and enter at A of section
2.258

As already stated, although this description uses a controlled vocabulary in an

almost natural language setting, it is nonetheless capable of describing the motion

of the thread required for a particular sewing structure. Its controlled vocabulary

is used in an intelligent way. In fact, by developing the terminology in three levels,

it gives an overall structure to the description. (i) It first defines the typology of

sewing, (ii) then the general resulting sewing patterns, and (iii) finally the sewing

layer. It provides spatial coordinates, and it names a series of actions to produce

the sewing being described. More general terms are used to describe the complex

object that is the overall sewing structure (levels i and ii), then other terms name

simpler objects and place them in space within the structure. Of particular interest

is the series of named actions: these in essence subdivide the complex object that

is the sewing thread into a sequential series of sewing thread fragments in specific

spatial configurations. The relations between objects are expressed through the

relation of concatenation259 within the sequential language of the description. What

at first glance looked similar to a natural language sentence with heavy use of

special terms, with closer analysis begins to look different. The very use of those

special terms creates, in fact, a specific framework giving the description a semi-

258. Spitzmueller & Frost 1982; Spitzmueller 1982.
259. Stenning & Lemon 1999, p. 48.

108



structured character. The logical form of the sewing structure is mirrored in the

precise use of the terms and in the natural sequentiality of the information being

recorded, thus making it a useful mode of communication. In practice, however,

what these descriptions lack is the querying capabilities inherent in the information

recorded within databases.

5.1.2. Communicating through structured descriptions

In order to describe a binding structure in any of its possible configurations,

and to make this description fully communicable, one would not only need to

convey information relative to its material components, but also be able to make

direct reference to the logical form it possesses.260 The canon of the rules of com-

position in classical architecture offer such a reference. There is no such canon

for the composition of the elements of bookbinding structures.

The reference to the logical form of the binding structure, the expression, is

what is lacking in non-structured descriptions, besides the fact that in non-struc-

tured descriptions spatiality beyond linearity is not easily conveyed. Two non-

structured descriptions of the same reality do not have the same expression in

common and do not form a class as they do not have a way to refer to exactly the

same logical form. In fact, as seen in the natural language examples in the previous

chapter, even two non-structured descriptions of the same state of affairs do not

necessarily possess the same expression. Because of this, their sense can be, and

usually is, lost, unless the verbal description is coupled with a graphic description,

whose elements correspond to the objects of the state of affairs being described:

it is the graphic description that preserves a reference to the logical form, and can

function as a medium between the logical form of the state of affairs and the form

of the verbal description.

Controlled vocabularies instead are able to contain the building blocks of a

complete description of a material object. Structured descriptions differ from

simple controlled vocabularies. It is this difference that provides the description

260. Wittgenstein 1922/2012T: 3.34; T: 3.341; T: 4.0141.

109



with a more complete representation of the logical form and its inherent possibility

to indicate any form. Structured descriptions, although still relying on controlled

vocabularies for the communication of objects, present information in an orderly

and prescribed manner. The schema, the grammar of the information, ruling

structured descriptions within databases is a manifestation of the expression and,

when properly developed, mirrors the logical form of the material object it de-

scribes.

Within structured descriptions, information is organized into hierarchies that

follow the logical structure of what is being described; within these hierarchies,

information is organized into groups that bind relevant data together, and that

form boundaries between entities.261 The hierarchical organization expresses the

association between parts262 and this preserves a reference to the logical form of

the state of affairs, and the configuration of objects within states of affairs.

The inclusion of controlled vocabularies within structured descriptions

strengthens their communication capabilities. Whole/part relationships and

functional/structural information can be encoded within the structure of the de-

scriptions themselves, and the schema behind them maintains a reference to the

logical form of the object just as much as classical architecture’s rules of compos-

ition.

Controlled vocabularies and structured descriptions offer a well-developed

mode of communication, in relation to the building blocks of material objects.

These descriptions, if based on a model for, are also potentially fully-specified 263

as they can provide precise measurements and spatial relationships, just as a fully-

specified description of a map would provide precise coordinates for every element

to be drawn. This is not the case when they are based on models of. As seen, these

models describe the idea of an object in reality and they strive to capture its logical

form, and only relative information on the positioning of elements and their

measurements are preserved (the logical form of a class of objects is applicable

to all examples of that class and precise measurements are exclusive not inclusive).

261. Salthe 1985; Brachman & Levesque 2004; Gnoli 2008.
262. Pazzi 1999; Doerr et al. 2001.
263. Eide 2012a.

110



This leads to under-specification, but, as seen in the case of the vocabulary of

classical architecture, not to the level that conflicting drawings would be generated.

The outcome would be under-specified in the sense that as it is based on prototyp-

ical information, it would be inclusive and able to represent all items of its category,

and not only one specific case. This is a different level of specificity that was not

encountered by Eide,264 since a prototypical map would be a useless map.

As mentioned in the previous chapter, a good example of a structured descrip-

tion and controlled vocabulary for bookbinding structures published to date is

the description schema developed by Ligatus for the survey of the printed books

in the library of the monastery of St. Catherine on Mount Sinai. The schema will

be presented in more detail in the following chapter. This model of binding

structures, being hierarchical and based on a controlled vocabulary, should be

able to communicate the minimum amount of information necessary to commu-

nicate information on bookbinding structures. In turn, this information should

be able to be transformed into diagrammatic visualizations of the structures.

5.1.3. Diagrams as support for natural language descriptions

In diagrammatic representations of material objects, each element stands in

relation to other elements, and this configuration mirrors that of the material

components of the object that is being represented. These relationships are em-

bedded within the two-dimensional structure of the diagram, without the need

for complex syntax.265 In written and spoken language, because of their innate

sequentiality, the expression of such relationships can prove challenging as they

would need to be based on a complex syntax to compensate for their limited se-

quential nature.266 The relationships are in fact not nameable elements, but manifest

themselves in the configuration of the elements.

264. Eide 2012a.
265. Stenning & Lemon 1999.
266. Shin et al. 2013.

111



This explains why verbal descriptions of bookbinding structures accompanied

by diagrams are capable of expressing spatiality: even when these are unstructured

and make strict use of natural language, their accompanying graphic descriptions

act as medium between the logical form of the state of affairs and the verbal de-

scription, preserving the reference to the logical form of the state of affairs.

5.2. Automated intersemiotic translations

This project is a clear case of intersemiotic translation:267 from verbal to visual

language.

In order to translate a proposition A in one language into a proposition B in

another, one needs to establish rules of definition and correlation between the

two expressions.268

As seen in the communication of material objects, one needs to convey inform-

ation regarding the material components and their form, and to maintain a refer-

ence to their logical form. In addition, for visualization purposes, one also needs

to be able to generate a prototypical shape for the objects and their components.

In a translation process the same information needs to be conveyed and trans-

formed into the target language. For this project, the transformations also have

to be programmable in advance through transformation algorithms capable of

capturing the necessary information and then transforming it into visual repres-

entations.

5.2.1. Translating material components, their form, and shape

Material components are translated by transforming the components of the

proposition A into the components of proposition B in such a way that the signs

that represent a component in language A possess a meaning that is common to

267. Jakobson 1959. See also chapter 1.
268. Wittgenstein 1922/2012T: 3.343; T: 4.0141.

112



all signs that can be substituted to them in a proposition B belonging to any other

language.269

In practice, therefore, an algorithm can be programmed in such a way to estab-

lish a one-to-one relationship between the components of the verbal descriptions

and those of the visual representations. For each descriptive element in proposition

A, something needs to be drawn in proposition B. However, this is not as

straightforward as it might seem. As covered in chapter 3, in visual texts, sememes

are interdependent, and by simply establishing a one-to-one text-to-picture rela-

tionship without considering the nature of the visual language, one risks generating

nonsensical and unstructured agglomerates of visual signs, like those proposed

by Zhu and colleagues.270

5.2.2. Translating material objects’ shapes

In a similar way, for each component in A, one should establish a shape that

its sign will take in proposition B. Through a one-to-one relationship between the

components, an algorithm could be designed to generate a prototypical shape for

each material component. This is possible because, as seen, verbal prototypes at

the basic level of abstraction are associated with a mental image that resembles

the appearance of members of the class as a whole.271 It is therefore feasible to

prescribe a shape that is derived from the mental image for each component.

However, once again, the very nature of visual signs imposes that careful consid-

eration is given to the generation of the shapes, as these are likely to depend on

those of the other signs within the visual text. To accommodate this, shape gener-

ation should be parametrized, i.e. it should use the relationships between compon-

ents as shape forming parameters.272

269. Wittgenstein 1922/2012T: 4.025; T: 3.344.
270. Zhu et al. 2007.
271. Rosch 1978.
272. Monedero 2000. See also Davis 2013, chapter 2, pp. 14-48.

113



5.2.3. Translating an object’s form

Information on an object’s form, as seen, can be conveyed through the use of

controlled vocabularies. The mental images triggered by verbal prototypes possess

structural and spatial characteristics273 that can be rendered graphically to depict

the form of the object. In addition, the hierarchical nature of the information

within structured descriptions is also capable of conveying spatial information.

This will be seen in more detail with practical examples in chapter 7.

In practical terms, the hierarchical organization of the information can be inter-

preted by the algorithm as spatial relationships between components, just as rel-

ative positioning derived from the definition of the controlled vocabulary entities.

This information resembles that highlighted by Goulette and Borillo274 for archi-

tectural vocabulary. Basic information is the relation of parthood (x is contained

in y), the relation of concatenation (x follows y), and the relation of border (x is

a border of y). Combined, these relationships can express the relative spatiality of

the parts. The orientation is derived from the general rules of composition of the

domain.275

5.2.4. The reference to the logical form

A reference to the logical form of the object is an important part of the commu-

nication process about material objects. Without it problems can occur. In auto-

mated intersemiotic translation of the kind described in this project, this reference,

embodied in the schema, is a necessary component of the process. The schema

informs the coding of the transformation algorithms, as it prescribes which com-

ponents are parts of the object described, and in which spatial configurations

these can occur. Without this, the problem would essentially turn into natural

language parsing, attended by all the complexities described in chapter 2.

273. Thomas 2013.
274. Goulette 1999; Borillo & Goulette 2006.
275. Goulette 1999.

114



For this reason, structured descriptions are an ideal verbal description modality.

The following chapters will analyse and assess this intersemiotic translation process

from structured verbal descriptions of bookbinding structures to diagrammatic

visualizations.

The following part of the chapter delineates those general considerations that

are necessary for a visual language attempting to represent material objects.

5.3. Communicating material objects through visual means

As already discussed in chapter 3, diagrammatic representations, unlike most

sentence-based representation systems, i.e. sequential languages,276 function

within systems of representation that are specific to a target domain, i.e. that which

is to be represented.277 That is to say that the signs to be used depend on what is

to be represented, and this is even more so the case in highly iconic representations.

This does not mean that general, meta-domain considerations cannot be drawn.

In fact, despite the arbitrariness of the signs specific to a language, a particular

mode of signifying something is unimportant. What is important is that it is a

possible mode of signification that is able to convey the essential.278

It is, in fact, possible to draw a set of general guidelines or design principles

that, by understanding how perception and cognition of visual information affect

comprehension of the information presented to the viewer, prescribe how to

communicate information effectively through visual means.279

The following sections will cover the most important considerations for an ef-

fective visual language. The following chapters and Appendix B will cover the

domain-specific visual language developed for this project.

276. Stenning & Lemon 1999.
277. Stenning & Lemon 1999.
278. Wittgenstein 1922/2012 T: 3.342; T: 3.3421.
279. Ford 1993; Tufte 1990; Tufte 1997; Tufte 2001; Tufte 2006; Agrawala et al. 2011; Bertin

1983/2011; Ware 2013.

115



5.3.1. Marks on a background

Chapter 3 shows how line drawings are sufficient to understand a shape, and

that these are prototypical enough to be used as graphical prototypes. Also, for

these reasons, line drawings present the maximum amount of data with the least

amount of ink, thus following Tufte’s280 data-ink ratio maximization principle for

graphic excellence.

Line drawings, in essence, are arrangements of linear marks on a white or

neutral background. These marks are the elements of a proposition, and, as such,

they can be sensed281 and sensed specifically through vision.

Visual perception of line drawings and diagrams allows for a total of three di-

mensions that can be used to encode information: (i & ii) the two dimensions of

the plane, and (iii) the possible variation of the marks.282 Bertin283 identifies eight

possible variables in total. Bertin differentiates between spatial and retinal variables.

Spatial variables are the two dimensions of the plane identifying the spatial position

of a mark. Bertin refers to the possible variations of the marks as retinal variables

because they are perceptually salient features that do not require eye movement

to be identified, and they must, therefore, be realized in the retina. Neuropsycho-

logy studies have identified that these features are linked to the firing of specific

neurons within the visual areas 1 (V1) and 2 (V2) of the cortex after preliminary

processing in the retina.284 Bertin distinguishes between six retinal variables: size,

texture, grey value (or lightness), colour (or hue), orientation, and shape. Psycho-

physics, the study of human responses to physically defined stimuli, have identified

more visual features that make items stand out and easy to find.285 For static bi-

dimensional percepts, Ware286 lists: line length and width, curvature, spatial

grouping and numerosity, sharpness/blur, and the addition of marks or halo.

280. Tufte 2001, pp. 123-138.
281. Wittgenstein 1922/2012T: 3.1.
282. Bertin 1983/2011.
283. Bertin 1983/2011.
284. Ware 2008; Ware 2013.
285. Findlay & Gilchrist 2005; Ware 2008; Ware 2013.
286. Ware 2008; Ware 2013.

116



5.3.2. Continuity of the plane

Bertin287 describes the plane on which visualizations are drawn as the richest

of the variables, as any mark has to fall on it. One can distinguish between signi-

fying and non-signifying parts of the plane: the former are intended to convey

meaning, and they exclude the latter, which are formed by the space lying outside

of graphic boundaries — e.g. a geographical border or the frame of a drawing —

and are, therefore, not meant to carry any information. It is possible to have nested

(or layered)288 signifying planes, but the nesting has to implement the Gestalt

principles289 of proximity, closure, and common region carefully to allow for the

correct regions to be regarded as non-signifying.

The signifying plane is continuous as it can be infinitely subdivided by marks

applied on it. As such, it does not admit gaps, informational lacunae, as the absence

of signs, or zero signs,290 is read as absence of phenomena, and not as missing data:

it is difficult to disregard part of the signifying plane.291

Therefore, in visualizations, it is important not to leave gaps in the signifying

parts of the plane, or the risk is to alter the meaning of the visualization, or to

render it meaningless. This also applies to the way in which boxes and edges are

used and designed: the balance between the signifying and the non-signifying

parts of the plane always has to be kept in mind.

5.3.3. Visualization conventions

Graphic languages utilize all of the above-mentioned variables to communicate

information. In the literature, general studies on graphic languages tend to focus

on the visualization of quantitative or geographical data (i.e. cartography), histor-

287. Bertin 1983/2011.
288. Ware 2008; Ware 2013.
289. Koffka 1935.
290. Sebeok 1994.
291. Bertin 1983/2011.

117



ical data, and Geographical Information Systems (GIS) data.292 Bertin293 and

Moretti294 distinguish between three main categories of graphics: diagrams or

graphs for quantitative data, networks or trees for qualitative data that can be ar-

bitrarily arranged on the plane, and maps for geographic or cartographic visualiz-

ations. To these categories, Bertin295 adds symbolic visualizations (i.e. symbols)

that can be used within all of the above categories.

The diagrammatic visualizations needed for this project do not fall precisely

within any of these categories. They are, however, closely related to strictly geo-

graphic and qualitative maps (less so with quantitative cartographic visualizations).

In qualitative geographic maps, the position on the plane and the visual features

of the representations are imposed by the nature of the data to be recorded: a city

has to be placed at the right point on the plane that corresponds with its geograph-

ical coordinates, and a river has to follow the same path on the plane followed by

its counterpart in the real world. In the same way, the position of a sewing support

has to be relative to that of the other elements represented with it — e.g. the fold

of the gathering — and the appearance of the sewing thread path should represent

the path of the sewing structure being visualized. For this reason, it is possible to

draw general concepts of visualizations from these partially pertinent studies.

However, the restrictions imposed by the iconic nature of the visualizations always

have to be kept in mind.

5.3.3.1. Fixing meaning through a reference system

Chapter 3 highlights how visual information needs to be fixed by a referencing

system for its meaning to become apparent. This can be accomplished by

presenting it with a dual modality that combines word labels with the graphics.

Also, the meaning of the elements of visual texts depend on their context. For

these reasons, it is always important to present visual information within a well-

292. For example: Bertin 1983/2011; Tufte 1990; Tufte 1997; Tufte 2001; Tufte 2006; MacEa-
chren 1992.

293. Bertin 1983/2011.
294. Moretti 2005.
295. Bertin 1983/2011.

118



developed referencing system that allows for the identification of the correct

meaning of the overall graphic and of its elements.

5.3.3.1.1. Titling and verbal labelling

Titles and other verbal labels allow fixing the meaning of a visualization. They

permit the reader to identify, the general category of what is being represented

— e.g. an endleaf structure — and to identify, in the drawing, those elements that

can vary from example to example within the same category — e.g. from one

endleaf structure to another.296

5.3.3.1.2. Graphic reference elements

Previously, it was concluded that there has to be a one-to-one relationship

between the components of verbal propositions and those of their visual transla-

tion. However, considering the need for contextualization of visual elements, it

is appropriate to add reference elements within a visualization to help put each

element of the structure in context visually, thus facilitating the signification of

all other elements. For this reason, certain elements taken for granted or simply

not mentioned in a description need to be made explicit within the visual repres-

entation of what is being described. These can be made less visually prominent

to differentiate them from important information. For example, endband cross-

section diagrams, as it will be seen in the following chapter, include the bookblock

on which the endband is attached as a visual reference element.

5.3.3.2. Separation of information

The basic rule for the generation of graphics with a high degree of legibility is

to separate information into discernible visual groups. This is attained by balancing

graphic density by avoiding too many marks per unit area and by keeping mean-

296. Bertin 1983/2011.

119



ingful marks visually separate.297 The same considerations are valid for different

views, sections, and boxes within a visualization.298

A consequence of the separation of information might be that the naturalism

of the representation needs to be reconsidered. Effective graphic representations

should always strive for best legibility, and correct signification, even at the expense

of naturalism, by allowing that content and form of the visualization correspond

to those of the desired mental representation (principles of congruity299) and that

visualizations are readily and correctly perceived and understood (principle of

apprehension300).301

5.3.3.3. Use of colour

Colour is an excellent selective variable that is able to make information imme-

diately distinguishable. However, it does have two main drawbacks: (i) anomalies

in chromatic perception (e.g. colour blindness) are more frequent than it is gener-

ally believed,302 and (ii) inaacurate colour reproduction.

From a neuropsychological point of view, colour perception as it is analysed in

the V1 area, can be subdivided into three information channels: the red-green

channel, the yellow-blue channel, and the black-white (or luminance) channel.

Luminance differences are calculated simultaneously between all adjacent areas

of the retina, and not from the output of a selection of specific cones as is the case

with the other channels. This justifies the greater capacity of the luminance

channel to convey detailed information as opposed to the other channels. Also,

colour blindness does not impede the perception of grey-scale values (and colours

that differ in the yellow-blue direction).303

297. Bertin 1983/2011.
298. Ford 1993.
299. Tversky 2005b, p. 29.
300. Tversky 2005b, p. 29.
301. See Chapter 3. Tversky 2005b; Tversky 2006.
302. About 8% of males and 1% of females are colour blind (Ware 2008).
303. Ware 2013.

120



There are also significant problems involved with colour reproduction from

one medium to another, and the more information depends on a hue-specific

colour coding, the more expensive the reproduction will be.304

Hue variations as carriers of specific meanings should only be used when indis-

pensable. The diagrams generated for this project will therefore make use of black

and white or grey-scale graphic elements.

5.3.3.4. Two-dimensional vs. three-dimensional visualizations

The problem with three-dimensional representations is that they make it harder

to compare the features and forms of similar items. Human pattern perception is

for the most part devoted to planar information as opposed to depth.305 Also,

whilst three-dimensional views seem to deliver information in a more direct way,

in reality, all shapes are in fact distorted — e.g. circles are ellipses, rectangles show

no right angle and no equal sides — and this can lead to ambiguity.306 Two-dimen-

sional views are better suited to facilitate the comparison of features between

different examples.307

5.3.4. Structured visualizations as output

The final output of this project is a series of automatically generated drawings.

Each drawing is the result of a direct translation of the description of a material

object. As mentioned above, in order for the translation to be appropriate, there

has to be a one-to-one relationship between the components of the propositions

in the two languages. This is self-evident, when the two languages have similar

kinds of expressions, as in the case of intralingual and interlingual translations.308

When dealing with intersemiotic translations, however, this might not be evident.

304. Bertin 1983/2011; Ware 2013.
305. Tversky 2001; Tufte 2001; Ware 2008; Ware 2013.
306. Ferguson 1992.
307. Ferguson 1992; Ware 2013.
308. Jakobson 1959.

121



Let us consider again the nature of information shown in images. Images feature

on a plane, which is continuous in nature and capable of infinite subdivisions.309

Images are continuous entities that can be subdivided in smaller elements within

their plane, but these are not always clearly discernible from one another. Images

can then be seen as macroscopic blocks of information, within which it is possible

to identify pertinent units.310 These units are the visual sememes of the visual

message encoded in the image. These units are also the elements of the visual

proposition corresponding to the elements of the object being depicted. They

stand in the same one-to-one relationship with reality as the elements of verbal

descriptions, and this is also the basis for the one-to-one relationship between the

two propositions in the two different languages.

Unlike verbal propositions, however, the units of images are seldom discreet

with clear boundaries. They can thus be generated in accordance to the one-to-

one relationship with the elements of the verbal description — i.e. for each element,

something can be drawn — but their final shape will generally depend on that of

the adjacent units, and this parametrization of the shapes will need to be taken

into account by the intersemiotic translation algorithm. This way, in a combinat-

orial way, the whole image text can be constructed element by element within a

structured visualization system. Structured as the resulting drawing, while appearing

as one (or sets of macro-groups, e.g. different views of the same item), has a clear

internal organization and arrangement of its parts.311

The next chapter will cover this in more detail when looking at the practical

application of these principles.

5.4. The uncertainty of reality

In reading or seeing a verbal or visual representation of an object and when

comparing it to the real object, one can ascertain whether the representation is

309. Bertin 1983/2011.
310. Eco 1975/2008; Saint-Martin 1990.
311. Wittgenstein 1922/2012T: 5.5423.

122



true or false, and the feedback to reality is actually the only way to be certain of

the truthfulness or falsehood of a proposition.312 Certainty is what makes it possible

to distinguish between truthfulness and falsehood, and its essence is in the action

of seeing that something is (or is not) the case. That is to say that a proposition is

only true or false in respect of what it represents. Certainty is an act of judgement

of what a proposition shows about reality, and propositions should make our

judging explicit. A judgement is correct only if the proposition judged is true.313

Reality is usually not so clear-cut: there are degrees of certainty about its truth,

because uncertainty pervades the world in which we live.314

In particular, when dealing with material objects such as historical bookbinding

structures and their visualization, one has to accept that a degree of uncertainty

is inevitable.315 There are many and various definitions of uncertainty,316 but in

general, all definitions imply imperfection of the knowledge about the dataset, or

in the representation of knowledge.317

Brodlie and colleagues,318 distinguish between two problems related to uncer-

tainty in visualization: (i) that of the visualization of uncertainty,319 and (ii) the

problem of the uncertainty of visualization.320 In answer to the former problem,

one has to deal with how to depict the uncertainty specified within the data to be

visualised. The latter problem instead rests in the inaccuracy that occurs in the

process of turning the data into a picture.

Therefore, a system that attempts to describe and represent reality has to ad-

dress, on the one hand, the problem of propositions that are neither completely

true nor completely false321 — i.e. to address the problem of the uncertainty

within the data — and, on the other hand, has to consider that the very process

312. Wittgenstein 1922/2012T: 2.21; T: 2.222; T: 2.223; T: 2.224.
313. Wittgenstein 1969; Stoutland 1998.
314. Gabbay & Smets 1998; Zuk 2008.
315. Campagnolo & Velios 2013.
316. Thomson et al. 2005.
317. Thomson et al. 2005.
318. Brodlie et al. 2012.
319. Brodlie et al. 2012, p. 82.
320. Brodlie et al. 2012, p. 82.
321. Gabbay & Smets 1998.

123



of translating data into a visual representation can produce problems with accur-

acy.322

The inclusion of uncertainty in a system, however, is not an easy task.323 Uncer-

tainty is a complex phenomenon. Thomson and colleagues324 have proposed a ty-

pology about uncertainty of geospatial information subdivided into nine categories

(accuracy/error, precision, completeness, consistency, currency/timing, credibility,

subjectiveness, interrelatedness, lineage), which has been further analysed and

generalized by Zuk and Carpendale.325 Of these categories, the most important

ones for the project at hand are: (i) accuracy — the difference between the obser-

vation and reality, (ii) precision — the exactness of measurement, and (iii) com-

pleteness — the extent to which the information is comprehensive.326

Uncertainty is propagated when we operate with uncertain data. And uncertainty

in the data also needs to be propagated as uncertainty in the graphic output. As

the original data is transformed into a visualization, uncertainty follows, and a

reference to the uncertain data must be maintained. Also, uncertainty adds a

visual dimension to the visualization, and this dimension needs not to have already

been used.327

Most visualization techniques traditionally have been developed, and used,

with the assumption that the visualized data is certain. Also, most people have a

tendency to treat visualized data as facts, and are much less prone to question

visualizations than written words.328 This, in turn, can lead to problems in the

reading of visualized data, as ‘one can draw inferences from a false proposition.’329

Tufte330 brings forward examples of how improper visualizations historically have

misled the interpretation of the original data.

For all these reasons, both uncertainty in the data and the uncertainty inherent

in the visualization medium need to be taken into account and signalled.

322. Brodlie et al. 2012.
323. Brodlie et al. 2012.
324. Thomson et al. 2005.
325. Zuk & Carpendale 2007.
326. Thomson et al. 2005; Zuk & Carpendale 2007.
327. Brodlie et al. 2012.
328. MacEachren 1992; MacEachren et al. 2005; Brodlie et al. 2012.
329. Wittgenstein 1922/2012T: 4.023.
330. Tufte 1997.

124



5.4.1. Uncertainty in structured descriptions

In the first instance, the data to be visualized has to be able to convey informa-

tion in respect of its certainty. In a description, a particular object can be present

or not, or it can correspond to a certain prescribed type or not. This seems simple

enough, however, there are many reasons why the reality of material objects cannot

be easily and completely described by yes or no answers: yes there is such and

such object, no there is not; yes this is such and such object, no this is not.

As in the case of quantitative and geographical data visualizations,331 factors of

uncertainty can be found in inherent problems with the definition of the object

of study — e.g. the limitations of current knowledge regarding bookbinding

structures and their description — or the sources of information and their inter-

pretation — e.g. the books being described and the interpretation of their binding

structures. Moreover, it might not be possible to completely describe a material

object — e.g. a book is sometimes too damaged to show clear evidence of its ori-

ginal structure, or else, elements of the binding might not be physically visible —

e.g. spine linings — and, therefore, describable.332

A description system has to be able to accommodate such cases. Structured

descriptions can allow for three- or multi-valued logic to be embedded within

their schema, thus allowing for more than just yes or no answers. Multi-valued

logic differs from the more commonly known bivalent logic — e.g. classical or

Boolean logic — as it does not restrict the number of truth values to only two:

true or not-true (false). The simplest version of this kind of logic, three-valued

logic, provides for three possible answers to the truthfulness of a proposition:

true, unknown, false.333 The simple addition of the unknown truth value can be

used to express uncertainty, and this is easily implemented within structured de-

scriptions. For example, in describing a bookbinding, for the aforementioned

reasons, one might not be certain whether the book has a certain type of spine

lining: saying that yes it has a certain type of spine lining would introduce poten-

331. Wilkinson 2005; Wainer 2009; Gethin Powell 2012.
332. Campagnolo & Velios 2013.
333. Gottwald & Anat 2014.

125



tially erroneous data into the dataset, but so would the opposite. By offering the

possibility of saying that the type of spine lining is unknown permits one to express

the uncertainty of the information: yes, there is a spine lining, but one cannot say

for certain which type of spine lining was used on that particular binding.

5.4.2. Uncertainty in visualizations

In the literature, there have been many and various attempts to include uncer-

tainty within visualizations in a diverse range of fields: from astrophysics to met-

eorology, geography, and archaeology, to mention but a few.334 The importance

of uncertainty was historically first realized by the geovisualization community,335

followed by the scientific visualization community.336 General reviews on the

problem can be found in the contributions by Wilkinson,337 Griethe and Schu-

mann,338 Zuk and Carpendale,339 Zuk,340 and Brodlie and colleagues.341

5.4.2.1. Archaeological visualizations

Of the various fields that use visualization as research output or tool, and

communication, archaeology appears to be the closest to that of this project: both

archaeology and the archaeology of the book deal with material objects as their

data, be this a temple or a book.

334. Brodlie et al. 2012.
335. MacEachren 1992; Goodchild et al. 1994; Pang 2001; MacEachren et al. 2005; Brodlie et

al. 2012.
336. Pang et al. 1997; Thomson et al. 2005; Wilkinson 2005; Brodlie et al. 2012.
337. Wilkinson 2005.
338. Griethe & Schumann 2006.
339. Zuk & Carpendale 2006; Zuk & Carpendale 2007.
340. Zuk 2008.
341. Brodlie et al. 2012.

126



Figure 31. Diagram of the construction of the Nag Hammadi codices (Szirmai 1999, fig.
1.2, p. 8).

5.4.2.1.1. An example from archaeology of the book

Archaeological datasets are often incomplete and ambiguous.342 Notwithstanding

this, visualizations of archaeological sites or objects more often than not tend to-

wards single holistic photorealistic images.343

The same idea of one fixed and certain visualization of how a bookbinding

looked can be found in studies of the archaeology of the book. An example is the

case of the notorious Nag Hammadi Codices. These are a series of thirteen papyrus

codices, dating from the third-fourth century AD, and amongst the earliest ex-

amples of surviving structures of books in the codex form. They were found still

in their original leather bindings in 1945, buried in a jar near the Nag Hammadi

village in Egypt. The discovery was announced in 1949344 and in the same year a

short article with a few general photographs of the manuscripts was published,345

but, because of their problematic content — Gnostic texts, amongst which the

342. Zuk 2008.
343. Harrison et al. 2012.
344. Doresse 1949.
345. Doresse & Mina 1949.

127



only complete example of the non-canonical Gospel of Thomas — they were not

made available to scholars for years.346 In 1961, Doresse347 presented a first brief

study of the bindings with some sketches. However, from the description of the

bindings by Robinson done in the 1970s,348 it is sadly clear that the bindings had

already been dismounted without careful documentation, with some parts already

lost and not in their original state, leaving aspects of the original binding structure

unclear and not recoverable from the evidence remaining.349 Even with this in

mind, Szirmai’s reconstructions, reproduced in Figure 31, do not offer any indic-

ation of what is certain and what is conjectural about the binding structure of the

Nag Hammadi codices. This is a clear example of the norm within the field of the

archaeology of the book: in the literature, to this author’s knowledge, there are

no examples of visualizations that include the idea of any kind of uncertainty.

5.4.2.1.2. Uncertainty clues in archaeological visualizations

Despite the many photorealistic archaeological visualizations that do not offer

a way to distinguish between speculation and hard data, the issue of uncertainty

in archaeological visualization has been raised by many scholars in the past two

decades.350

Different approaches have been proposed to tackle the uncertainty problem.

For the most part, these try to add some visual clues to photorealistic visualizations,

like the use of colour, fog, and transparency.351 Some studies352 have opted to use

non-photorealistic methods instead, and show how sketch-like renditions coupled

with the use of other visual dimensions — e.g. transparency — are fully capable

of conveying speculative data as opposed to hard data in archaeological recon-

structions. However, more and more the appeal of photorealistic reconstructions

346. Robinson 1984; Szirmai 1999.
347. Doresse 1961.
348. Robinson 1972-1977.
349. Szirmai 1999.
350. Miller & Richards 1995; Ryan 1996; Zuk et al. 2005; Bentkowska-Kafel et al. 2012; Harrison

et al. 2012.
351. Pang et al. 1997; Zuk 2008.
352. Strothotte et al. 1994; Strothotte et al. 1999; Strothotte et al. 1999; Freudenberg et al. 2001;

Strothotte & Schlechtweg 2002; Roussou & Drettakis 2003.

128



that can be seen as photographs of a possible past, with its communicative power,

has won notwithstanding the resulting false certainty.353

5.4.2.2. Irregularities in visual propositions

Even though the archaeological visualizations mentioned in the previous sections

refer to three-dimensional reconstructions — whether photorealistic or not — it

is interesting to note how they include visual cues to the figure, and these visual

cues can be applied to a single artefact, or an entire group or scene.354 By doing

so, they manage to convey the uncertainty of certain parts of the reconstruction,

without altering the perception of the object as a whole.

It is, in fact, possible to add irregularities to a certain figure, without distorting

the overall meaning of visual propositions, for even the ‘irregularities depict what

they are intended to express’.355However, attention should be paid to avoid trig-

gering unwanted Gestalt laws because of the interplay of particular uncertainty

encodings — e.g. changing the figure/ground balance — thus shifting attention.356

5.4.2.2.1. Visual cues for uncertainty visualization

In the literature across different domains, the list of possible visual cues for

visualizing uncertainty is limited to a range of seven categories: (i) size, (ii) position,

(iii) texture, (iv) shape, (v) colour/saturation, (vi) transparency/fading, (vii) blur.357

Interestingly, these correspond, for the most part, to the retinal variable outlined

by Bertin,358 with the exception of orientation, that is never used to depict uncer-

tainty, and value that is not used in isolation, but it is rather used together with

353. Harrison et al. 2012.
354. Zuk 2008.
355. Wittgenstein 1922/2012T: 4.013.
356. Zuk 2008.
357. For reviews of different uncertainty visualization strategies see: MacEachren 1992; Good-

child et al. 1994; Strothotte et al. 1994; Pang et al. 1997; Strothotte et al. 1999; Strothotte et al.
1999; Pang 2001; MacEachren et al. 2005; Thomson et al. 2005; Wilkinson 2005; Griethe &
Schumann 2006; Zuk & Carpendale 2006; Zuk & Carpendale 2007; Zuk 2008; Boukhelifa et al.
2012; Brodlie et al. 2012.

358. Bertin 1983/2011.

129



colour or in transparency and fading. To Bertin’s variables, MacEachren359 adds

blurring and the aforementioned transparency variable.

Size and position only apply to quantitative data visualization within diagrams,360

as in iconic texts both the size of an element and its position have to be relative

to those of the other elements for their signification.

Texture, defined by Bertin361 as the number of separable marks within a unit

(point, line, area), work by altering the appearance of the unit by making it look

as if formed by a series of successive marks, which are read as pertaining to the

same semantic unit due to the Gestalt principles of continuity and grouping.362

Dashing or dotting is an example of texture applied to lines. In bookbinding line

drawings, the norm is to use a dotted or dashed line to represent elements or parts

of elements, that in reality would be partially or completely hidden and thus not

visible. Sometimes, the same convention of dashing lines is used for other reasons,

such as to indicate folds (see Figure 32). This means that the texture dimension

has already been assigned various values, and it should not be used with yet an-

other meaning.

For material object visualizations, the general shape of an item cannot be

changed arbitrarily. However, the shape of the lines making up the drawings can

be modified to carry some special meaning, in the same way as texture can change

the appearance of a line without altering its overall shape. Sketchy or wiggly lines

fall within this category. The research group of Boukhelifa, Wood, and col-

leagues,363 inspired by the work on non-photorealistic rendering reported above,

have explored the capability of sketchiness to convey visual imprecision that may

be associated with uncertainty. Their results show that whilst some users associated

sketchiness with unprofessional rendering, it is nonetheless a viable additional

choice.

Hue variations, as we have seen in §5.3.3.3, should be avoided if at all possible.

For black and white line drawings on a neutral background, the variables of grey

359. MacEachren 1992.
360. Wilkinson 2005.
361. Bertin 1983/2011.
362. Koffka 1935.
363. Boukhelifa et al. 2012; Wood et al. 2012.

130



Figure 32. Drawings showing the use of dashed lines to indicate both invisible parts and
folds (Clarkson 2005, fig. 7 and 8, p. 9).

scale saturation, transparency, and fading fall within the same category and are not

separable. These variations provide redundant encoding as they use more than

one variable and channel of vision area 1 (V1)364 at a time. As a consequence, they

are easily perceivable and good candidates for carrying extra information,365 e.g.

darker lines, being visually more prominent, suggest more certainty than lighter

lines.366 The same is valid for blurring,367 which can be defined as the removal of

spatial detail from the information.368 This variable too is redundant and acts on

the variables of grey scale saturation, transparency, and fading. But in addition, it

also banks on the V1 feature of sharpness/blurriness detection,369 which makes it

an even more redundant and distinguishable variable. Blurring has been widely

used to indicate uncertainty and ambiguity in the data across fields.370

Of the various visual cues that have been used in the literature to express un-

certainty, the methods that are applicable to line primitives can be grouped into

three main categories: (i) colour-based techniques manipulating the saturation,

or brightness dimensions; (ii) geometry-based techniques that modify the appear-

ance of line marks; and, (iii) focus-based techniques that work by modifying

364. Ware 2013.
365. Zuk & Carpendale 2006.
366. Boukhelifa et al. 2012.
367. Zuk & Carpendale 2006.
368. Boukhelifa et al. 2012.
369. Ware 2013.
370. MacEachren 1992; Pang et al. 1997; Wilkinson 2005; Zuk & Carpendale 2006; Zuk 2008;

Boukhelifa et al. 2012.

131



contour crispness, and transparency.371 To summarize, grey scale saturation and

transparency/fading, used on their own or in conjunction with blurring, and

sketchiness of lines appear to be optimum variables for this project. All of these

are also applicable selectively on specific single elements: an ideal feature for

structured visualizations.

5.4.2.3. Imprecision as uncertainty

Uncertainty in the data, that in turn should be mapped and rendered as such

in the final visualization, falls within the categories of accuracy and completeness

of recorded information, and of the precision of measurements.

The sections above have covered the problems that may lead to uncertainty in

the data for a structured description of material objects such as bookbinding

structures, covering the accuracy and completeness categories.

In regard to measurements, it should be stated again here that their precision

is not essential to the prototypification of shape, as measurements do not constitute

a prägnant feature.372 Measurement values can, in fact, be changed without psycho-

logically affecting the overall impression for the observer as long as the proportions

are left unaltered.373 Nevertheless, when the measurement of certain elements can

be compared from object to object, it is important to be able to highlight uncer-

tainty in regard to the extension of a particular line: the length or scale of an ele-

ment may, in fact, be critical to understanding certain features of the material

object described and visualized.374

5.4.2.4. Uncertainty inherent in the visualization process

Finally, as seen, in the passage from verbal descriptions to visual, some features

of the specificity of an object may be lost due to the prototypicality of verbal in-

371. Boukhelifa et al. 2012.
372. As seen in §3.4.2.2, a singular, meaningful, special value of a trait or parameter. (Palmer

1999; Sternberg & Mio 2009).
373. Goldmeier 1972.
374. Campagnolo & Velios 2013.

132



formation, the generalization inherent in defined signs, and the semantic entry

level used in the description. This does not mean that an object cannot be visual-

ized at all, but that in some cases, the prototypified shape would be too generalized

to resemble reality closely. If indeed its category was presented in the description

at the right level of abstraction, thus highlighting its parts and their arrangement,

its essence would be present;375 however, its overall shape might be overly simplified

and symmetrical.

This poses a problem of uncertainty that becomes apparent only in the visual-

ization of the data, and it is an accuracy problem. It might be linked to complex

contours of certain elements, e.g. due to highly decorative patterns, or shapes with

random and irregular outlines.

In these cases, highly prototypical shapes can be used as place-holders. These

can then be highlighted as uncertain for formal integrity.

5.4.3. Human error in the data

When inaccuracy is objectively determinable, it can be expressed as error.376

Errors are indicated by the fact that a planned sequence of activities fails to achieve

the intended outcome, without intervention of external factors.377 In electronic

databases, the encoding schema at the base of their structure allows for immediate

monitoring of data correctness and completeness during input.378 Missing and

inadmissible data is highlighted right away by the computer, prompting the user

to add or correct it. Data validation yields to a reduction in errors and acts as

quality control. However, not all mistakes can be avoided through careful database

design, and those errors that do occur are not easy to identify through automated

means.379

375. Tversky & Hemenway 1984.
376. MacEachren et al. 2005.
377. Reason 1990.
378. Mocean 2007.
379. Campagnolo 2014a.

133



Data validation is the process of ensuring that a dataset is complete, correct

and meaningful. Validation rules check for correctness or meaningfulness of data

that are input by the user.380 Just as in language, a dataset can be considered valid,

when it satisfies the validation rules put in place in the system, but this does not

necessarily mean that it is also meaningful. In fact, one should not confuse the

notion of ‘grammatically correct’ — or ‘valid according to the validation routines’

— with ‘meaningful’.381

Consider the following sentences: (i) colourless sewing passing through four sta-

tions; (ii) stations through passing four sewing colourless. Both are nonsensical, but

(i) would be recognised as grammatically correct by any English speaker. One

can, in fact, distinguish between two senses of meaningfulness or validity. A pro-

position that is ‘valid in the first sense’ is meaningful in as much as it follows the

rules of the language in which it is expressed — e.g. it follows the rules of sentence

formation set by English grammar. A proposition that is ‘valid in the second sense’

is meaningful in as much as it makes sense in the context in which it is used.382

That is to say that a statement can be ‘meaningful in the first sense’ — i.e. it makes

grammatically sense — but ‘meaningless in the second sense’ — i.e. in the context

in which it has been used. Example (i) above is ‘valid in the first sense’, but not

in the second sense, i.e. there does not seem to be a context in which such a sen-

tence would make sense, and it is not meaningful as a bookbinding sewing descrip-

tion.

In the same way, data within a database can be valid — grammatically correct

— but nonetheless meaningless. Data that is not ‘valid in the first sense’ can be

avoided through validation routines. Ambiguities due to human error that cause

‘invalidity in the second sense’ are, instead, not avoidable through data validation.

In the case of visualizations, data that is grammatically valid — i.e. ‘valid in the

first sense’ — but meaningless — i.e. not ‘valid in the second sense’ — would

translate into drawings, whose elements are all possible within the state of affairs

being described, but represent something that might not be in reality, or does not

380. Mocean 2007.
381. Chomsky 1957.
382. Wittgenstein 1969; Stoutland 1998.

134



reflect what the reality actually is. To check if a visualization is correct or incorrect,

one has to check that proposition against reality.383 Of all the possible configura-

tions of the elements of a material object, one and only one corresponds to reality,

and from a proposition it is not always possible to see that it is, in fact, false.384 A

proposition can be understood without knowing if it is true or false as it is under-

stood by virtue of understanding its constituent parts.385

Meaninglessness ‘in the first sense’ can be avoided through data validation.

Ambiguities due to human error that make propositions not ‘valid in the second

sense’ instead are not avoidable through data validation, and are, therefore, not

avoidable in visualizations either. Some mistakes, might be immediately recognised

as false propositions, because they are meaningless ‘in the first sense’, but in ac-

cordance with the rules of reality — as they do not show a configuration that

would be possible in real life — and not of the schema, i.e. the elements that

compose them are all possible according to the schema, but their coming together

in the visualization is meaningless. Other propositions instead, might not be re-

cognized as meaningless until they are indeed checked against reality, as their

being meaningless is related to ‘validity in the second sense’, and their being

meaningful is then context-dependent and the only way to understand how the

object really looked like is to check the proposition against reality. In chapter 7,

these cases are discussed with practical examples taken from the dataset.

Summary

Signs and reality are in relationship with one another. Objects and their arrange-

ment in reality are mirrored in communication as the object determines the sign

by imposing the constraints that the sign ought to meet to signify it. Controlled

vocabularies and structured descriptions have, in potency, the capacity to describe

material objects in such a way that the correspondences to reality, its objects, and

383. Wittgenstein 1922/2012T: 4.05; T: 4.06; T: 4.1.
384. Wittgenstein 1922/2012T: 2.223; T: 2.224.
385. Wittgenstein 1922/2012T: 4.024.

135



its logical form are maintained. At the same time, diagrammatic visualizations are

capable of preserving and communicating the form of the reality that they repres-

ent. It is thus possible to foresee an intersemiotic translation of propositions from

verbal structured descriptions to diagrams, with a one-to-one relationship between

the basic elements of each.

Visual signs are specific to a certain domain. It is, however, possible to draw

general considerations on the nature of visual signs and to establish general con-

ventions for the production of a successful visual communication system. In par-

ticular, it is important to fix the meaning of each sign and of the diagram in its

totality, and to avoid altering such meaning with graphical and perception/psy-

chological artefacts, such as the inclusion of information gaps within the visualiz-

ation plane, or grouping meaningful marks that should be kept visually separate

for optimum legibility.

Visualizations within scholarly research projects should convey uncertainty

where needed. Depicting uncertainty has been the object of study of many

scholars, however, its application to diagrams whose shapes are dictated by those

of the objects being represented, as in the case of material objects, poses particular

issues and constraints. Various variables have been used to depict uncertainty, of

these, blurring, grey scale variation, transparency, fading, and sketchy lines seem

to be better suited to be applied to material object visualizations.

Noteworthy is the fact that, in the literature, uncertainty has not been considered

yet for visualizations pertaining to the field of the archaeology of the book. This

project has endeavoured to amend this and brings forward a methodology to in-

clude uncertainty in a consistent manner in the visualization of historical book-

binding structures. This constitutes an important contribution of this work to the

field of the archaeology of the book.

Human errors in the data can be limited with the implementation of data valid-

ation techniques. These, however, will not eliminate the possibility of erroneous

data being input. False visualizations caused by these errors are readable, and

sometimes even identifiable as erroneous, but their validity has to be checked

against reality.

136



The following chapters will show the visual language developed for this project

and the practical application of the considerations covered so far.

137



Chapter 6. Transformation framework

The concept, element, can be understood in two different ways:

as an external, and, as an inner concept. Externally, each indi-

vidual graphic or pictorial form is an element. Inwardly, it is not

this form itself but, rather, the tension within it, which consti-

tutes the element.

Wassily Kandisky, Point and line to plane, 1947, p. 33.

This chapter introduces the technologies that have been used in this project,

the dataset, and the schema developed by Ligatus in more detail. The next chapter

gives a complete example of a transformation and introduces other bookbinding

structure transformations. A complete account of each transformation can be

found in Appendix B.

138



6.1. The technology and the description schema

6.1.1. Extensible Markup Language

This project makes extensive use of the eXtensible Markup Language (XML)386

and associated technologies.

XML was developed in 1996 by an XML Working Group formed under the

auspices of the World Wide Web Consortium (W3C)387 to meet the challenges of

large-scale electronic publishing. With time, XML has also been playing an in-

creasingly important role in the exchange of data on the Web and elsewhere.388

XML 1.0 (fifth edition) is a W3C Recommendation and is the most recent version

of the full specification.389 XML technologies are widely used to structure, store,

exchange, and process data in a system-independent way. XML, as a markup

language, is designed to process and define information in the form of text. Hy-

perText Markup Language (HTML) is an example of widely known and used

markup language for web pages. XML is a metalanguage, a language used to de-

scribe other markup languages (including XHTML, eXtensible HyperText Markup

Language, a version of HTML). In essence, XML is a grammar that specifies how

to distinguish markup from other information, and which markup is allowed and

required. XML does not specify the meaning of the markup.

In XML, markup instructions are intermixed with data in the same document.

Markup instructions, referred to as elements, are kept distinct by enclosing them

in angular brackets like this <element>.

Elements are the basic unit of XML documents. Each type of element is identi-

fied by a name — i.e. the word within the angular brackets — but XML does not

prescribe the way to express the semantic meaning linked to a particular type of

element. All XML does is express an element’s relationship to other elements.

Given an element called <foo>, all one can say about it, thanks to the XML

386. Bray et al. 2008.
387. http://www.w3.org/ (accessed April 2015).
388. Bray et al. 2008; W3C 2015.
389. Bray et al. 2008.

139



grammar ruling an XML document, is that it may (or may not) occur within ele-

ments of the type called <bar>, and that it may (or may not) contain elements

called <baz>. XML documents are in fact innately hierarchical and form a tree

structure that starts at the root and branches to the leaves.

The XML grammars describing and prescribing these kinds of relationships

and the elements types (with their names) are referred to as schemas. An XML

document is created according to a specific schema and validated against it. Every

schema has to follow the general rules set by the XML specification by the World

Wide Web Consortium that created it.

XML languages are being used to structure various other types of information

besides text, e.g. vector graphics, procedural information, and so forth.

6.1.2. Extensible Stylesheet Language Transformations

XML documents have no set function. Once a document is structured according

to a specific schema, the information is semantically highlighted and therefore

searchable, but nothing else happens. Once encoded, however, it can be transpor-

ted, exchanged, and transformed into other forms.

EXtensible Stylesheet Language Transformations (XSLT),390 part of the eXtens-

ible Stylesheet Language (XSL), a style sheet language for XML documents, is an

XML technology developed to transform XML documents — or parts of them

— into other XML documents or other formats.

XSLT is a standard of the World Wide Web Consortium. Although referred

to as stylesheet, XSLT is essentially a fully-developed programming language that

is well-suited to extrapolate information from XML documents and transform it

into whichever format is desired, be that a web page, a PDF file, or a diagram. It

is therefore an ideal language for intersemiotic translations of information encoded

within XML documents.

390. Clark 1999.

140



6.1.3. Scalable Vector Graphics

This project aims to extrapolate information from XML documents describing

bookbinding structures and transform it into diagrammatic visualizations. The

project’s dataset was already in XML and one can use an XML-based language

to transform it; it therefore makes sense to use an XML technology to visualize

the output.

Scalable Vector Graphics (SVG)391 is a language for describing two-dimensional

vector graphics in XML. Originating from the requirements for scalable graphics

for the web by Chris Lilley,392 the SVG specification is an open standard developed

by the World Wide Web Consortium (W3C) since 1999. SVG 1.1 (second edition)

is a W3C Recommendation and is the most recent version of the full specification.393

SVG uses elements to describe in text form how its vectorial shapes should be

rendered and visualized. It is scalable because it defines the shape of each graphic

element but in a scale independent way, i.e. each element can be scaled up or

down without affecting the quality of the image.394

SVG is recommended by the World Wide Web Consortium and, as such, integ-

rates with other W3C standards, e.g. XSL. Being XML-based, SVG is a good

choice for the visual representations automatically generated for this project.

6.1.4. The description schema and the dataset

In analysing the different kinds of bookbinding structure descriptions in the

literature in chapter 4, the schema developed by Ligatus was briefly introduced.

This is an example of a structured description that makes use of a purposely de-

veloped controlled vocabulary. As such, it is capable of conveying the minimum

information needed for a transformation of the information into visual represent-

ations. Through its hierarchical structure and the use of a strict controlled

391. Dahlström et al. 2011.
392. Lilley 1996.
393. Dahlström et al. 2011.
394. Terras 2008.

141



vocabulary it can portray information on the material components and form of

the object. The schema can also convey the prototypical shapes of the components

through the prototypical nature of the vocabulary used, if the terms are set at the

most useful level of abstraction. The schema also embodies the expression of the

description, and this provides a direct link with the logical form of the object being

described.

The schema is the basis of the project’s dataset. The following sections offer a

detailed introduction to its history and structure, and look at the data generated

through it.

6.1.4.1. The Ligatus schema for the description of bookbinding structures

In the introduction of The Archaeology of Medieval Bookbinding Szirmai395

points out that ‘terminological clarity is a prerequisite for precise recording of

observation in binding structures [and] the lack of an established and uniform

English terminology [prevents] from achieving the [desired] precision.’ Ligatus

has tried to tackle the terminology problem, whilst also advancing a methodology

for the description of bookbinding structures. The terminology and description

methodology were first applied during the condition survey of the library of Saint

Catherine Monastery on Mount Sinai, carried out by team of conservators396 from

2001 to 2007.

6.1.4.1.1. The manuscript collection survey

The terminology and methodology proposed by Ligatus was then developed

and applied in the condition survey of the library of Saint Catherine Monastery

on Mount Sinai. The survey project was run by the Saint Catherine Foundation397

and Ligatus. The library’s holdings are codicologically very important, especially

because of the variety of early bookbinding structures that have survived. The

main scope of the project was the preservation of the library. However, to establish

395. Szirmai 1999, p. xii.
396. A list of the visiting conservators can be found at http://www.ligatus.org.uk/people (ac-

cessed April 2015). This author was not part of the project for the duration of the survey.
397. http://www.saintcatherinefoundation.org/en/ (accessed March 2014).

142



a conservation plan for the library, a detailed survey of all the volumes was pro-

posed to assess the whole of the collection and produce a comprehensive record

of the material.398

The survey of the manuscript collection was run from 2001 to 2006, and it in-

cluded a record of the condition and the structure of the bindings in as much

detail as possible given the time constraints. Each volume was allocated about

one hour for its description. The work was done in teams of two people to avoid

as many possible errors in the data recording.399

The data collected was noted on paper forms (see Figure 33), kept aside as a

physical backup, and then automatically uploaded into a database developed for

the project. About 3,300 books have been examined. At first the data was organized

into a relational database. But given the hierarchical nature of the recorded data,

the relational model was abandoned in favour of XML records.400

6.1.4.1.2. The printed book collection survey

The way in which information was collected during the survey followed the

well-established method of following the way in which the book was put together

originally,401 starting with general observations, then progressing to the finer detail:

this is a highly hierarchical methodology that is best expressed through another

hierarchical descriptive methodology. A second development of the survey

methodology made use of XML’s inherent hierarchical nature.402

Ligatus developed an XML schema to describe bookbinding structures. A new

description schema was developed as no existing solution could describe binding

structures in such detail. This schema, based on the experience accrued during

the first phase of the survey, was devised to encompass the kind of structures

likely to be found in the collections to be surveyed, but it was not comprehensive.

398. Pickwoad 2004.
399. Pickwoad 2004.
400. Pickwoad 2004; Velios & Pickwoad 2004; Velios & Pickwoad 2005a; Velios & Pickwoad

2005b; Velios & Pickwoad 2008.
401. Sharpe 2000.
402. Velios & Pickwoad 2005a; Velios 2008; Velios & Pickwoad 2008; Velios & Pickwoad 2009.

143



Figure 33. Typical page of the paper form used in the survey of the manuscripts at the
Library of St. Catherine’s Monastery on Mount Sinai (Velios & Pickwoad 2004, p. 658).

144



In 2007, teams of conservators visited St. Catherine’s Monastery to survey the

library’s printed book collection. This time, the paper forms were substituted

with electronic forms to be filled in directly on a computer screen or tablet. The

electronic form utilized the newly-developed XML schema. Unlike the former

paper forms, the XML structured descriptions of bindings allowed for immediate

data validation during the survey, resulting in fewer errors compared to the paper-

based survey of the manuscripts. This way, 814 bookbindings from the printed

book collection were described, and the data was stored in XML files generated

from the electronic forms.403

6.1.4.1.3. Free-hand drawings

For both phases of the survey, drawing was considered an important part of

the process, as drawing demands observational accuracy, helping to focus on the

important details of the item being examined. Information was recorded in outline

and diagrammatic drawings which were fast to execute, accurate, and language

independent. The drawings, never intended to be realistic, but rather schematic

informative documentation, functioned also as a kind of supplementary detailed

record of some of the structures, adding visual information to the photographs

that were taken for each book.404

Drawings were integrated within the paper-forms for the manuscript survey as

seen on Figure 33. Whereas, a specially developed new paper form for drawings

was devised for the printed book survey: an A3 landscape form with various sec-

tions covering the required or possible drawings (see Figure 34). Paper forms

were preferred to direct digital input, as people prefer drawing on paper rather

than a computer screen.405

403. Velios & Pickwoad 2005a; Velios 2008; Velios & Pickwoad 2008; Velios & Pickwoad 2009.
404. Pickwoad 2004; Velios & Pickwoad 2004; Velios & Pickwoad 2005a; Velios & Pickwoad

2005b; Velios & Pickwoad 2008.
405. Velios & Pickwoad 2005a.

145



Figure 34. Example of the drawing paper form used during the survey of the printed
books at the Library of St. Catherine’s Monastery on Mount Sinai developed by Ligatus.

6.1.4.2. The project dataset

The transformation algorithms for this project have been based on the kind

and amount of data prescribed by the XML schema developed by Ligatus for the

printed book survey. The 814 XML records collected during the survey represent

the validation dataset for the algorithms. These records were used to test and

validate that the transformation algorithms could meet acceptable levels of accuracy

and performance. Effective data validation is typically difficult to create from

scratch; thus transformations for structures theoretically covered by the schema,

but for which no examples were found in the dataset, would be problematic to

test or develop properly. The printed book collection, however, offered a wide

range of binding styles to account for most structures. Simple structures, such as

board markers, page markers, and tacketed structures, not found in the dataset

were developed nonetheless for reasons of completeness.

146



6.2. The transformations

It is now time to examine which of the structures and structural elements ac-

counted for within the Ligatus schema are described in sufficient detail for the

information to be automatically transformed into meaningful diagrams.

Figure 35, shows a graphic representation of the hierarchical structure of the

Ligatus XML schema down to the first level of branching. The root element is

obviously the ‘book’ element. Then the elements can be grouped as follows: a first

group gathers general information on the item, like bibliographical information,

overall dimensions, and opening characteristics. A second group focuses on ma-

terial that can be added to the volume, like inserted material, and markers. The

last group instead, starting from the textblock, the covering and furniture, analyses

the formation of the binding component by component. The analysis of the text

leaves does not cover gathering formation and collation.

The schema describes the material components, and, to an extent, the form of:

markers (page markers, board markers, and bookmarks), endleaves, sewing,

boards, spine profile and joints, spine lining, endbands, covering, and furniture.

All of these structures are potentially described in sufficient detail to allow for

the information to be algorithmically analysed and transformed into diagrammatic

representations.

Chapter 7 will present the information contained in the schema component by

component. The following sections in this chapter introduce a few general con-

ventions and considerations to help the reader better understand the data

presented and the diagrams automatically generated for this project.

6.2.1. Knowledge graph schemas

Each transformation will be accompanied by a graph showing the complete set

of information made available by the schema to describe a binding structure, and

the complete set of elements that can be drawn to represent that structure. In

turn, these graphs also show what information is available to the viewer for the

147



Figure 35. Visualization of the Ligatus schema: the root and the first level branches. These
can be subdivided into three groups: group I contains the general information on the
book; group II contains the material that can be added to the volume; and group III
covers the components of the binding.

148



interpretation of the generated bookbinding structure diagrams. The diagrams

are presented as a way to visualize the information behind each transformation.

Graph schemas have been devised by this author as a means to show all the in-

formation needed for a description/transformation in a single page (or a series of

related pages for very complex structures); they visualize both the schema behind

each description and the path followed by each algorithmic transformation to

generate the diagrams for this project. They can be complicated to follow, but

they can also show in one glance the complexity of a transformation, as opposed

to having to trail through the XML schema and the XSLT transformation code.

These kind of graphs resemble Findler’s406 associative networks, and Pinker’s407

visual description graph schemas.408 These graphs schemas are helpful for repres-

enting the knowledge domain for each transformation. The graphs are formed of

nodes and links between nodes. Nodes represent variables identified as concepts,

and links represent relations between concepts. They serve also as a way to describe

how a reader of a diagram comprehends the bookbinding diagrams. Pinker,409 in

fact, considers them as memory representations embodying knowledge in some

domain, each containing slots or parameters for the information. Each graph can

both specify what information must be true for the representation of some object

of a given class, and what information varies from one exemplar of a class to an-

other.

6.2.1.1. Graphic conventions for knowledge graph schemas

The graphs show a complete visual description for each bookbinding structure

diagram. Nodes are represented by small circles, with textual notations as predic-

ates, and links by arrows linking two entities. Each node that could take different

values is represented by a small triangle branching into a series of entities, showing

the possible choices, and what information varies from record to record. Paramet-

rized predicates — i.e., here, elements that are defined by precise parameters,

406. Findler 1979.
407. Pinker 1990.
408. Pinker 1990, p. 94.
409. Pinker 1990.

149



which, in these instances, is taken to mean numbers and measurements — are

indicated by small squares linked to the entity that they define. As in Pinker,410 a

star (*) inside a node indicates that the node could be repeated more than once.

A question mark (?) inside a node indicates that the node can be repeated zero

or more times. A dagger symbol (†) indicates nodes whose visual representation

is not possible with the present level of information available in the schema. The

typical choice values of not checked (NC), not known (NK), and other are indicated,

as shorthand notation, by three small blurred filled circles placed under the relevant

choice triangle; these are blurred to indicate that the information they provide is

uncertain. Similarly, three blurred daggers are placed under a triangle where the

lack of sufficient information or the level of uncertainty renders impossible any

meaningful and not misleading visual representation. A dashed and dotted line

under a triangle, with a circle inscribed in it, is a further shorthand notation for

the customary series of choices: yes, no, NC, NK, other. The circle inscribed in

the triangle indicates that something is to be drawn.

In the interest of brevity, and if at all possible without rendering the graph too

complex, the same set of nodes is not repeated more than once. Looking at the

graphs it becomes clear that the linking lines initiate a logical path of relations

along which some information is gathered at each node. Finely dashed lines indicate

that some options are only available for a certain path along the schema. These

at first identify the beginning of the alternative path, followed by a series of nodes

in common to all paths indicated by the usual simple lines up to the point at which

the paths divert; what follows is only relevant to the path/s indicated by the first

dashed line/s.

There are cases of alternative visualizations. These are generated by the same

set of nodes, but arranged spatially in a different fashion. These alternative visual-

izations are the result of the same algorithm, but the final layout is slightly different

— e.g. diagrams for left and right endleaves are mirrored, but still composed of

the same kinds of elements. For these, one visualization is fully developed, whilst

the others are indicated with a dashed line that fades: from that point onward the

410. Pinker 1990.

150



path to be followed would be the same as the fully developed visualization, but

the final diagram would have a different spatial layout. Similarly, some structures

are particularly complex; for these a set of correlated graph schemas are presented.

The groups are identifiable by the arrow at the end of the alternative visualization

dashed line pointing to the alternative graph schema (see Figure 36).

6.2.1.2. Reading the visual description graphs

Each circle in the graphs represents an entity of the visual description. In

practical terms, this is taken to mean that for each node something is graphically

drawn in the bookbinding diagrams, either as a simple shape, or as a complex

series of lines whose final appearance will depend on the exact series of nodes

that accompany it. For this reason, the very first element in each graph is a node,

indicating the diagram as a whole, of which all the other elements are essentially

parts that contribute to its appearance.

Images, as explored in chapter 3, are to be seen and read as iconic texts. In

practice, visual description graph schemas parse these texts and clearly distinguish

the units that compose them. Each unit is understood in relation to the others for

the whole meaning to be inferred. Each node can therefore be understood as a

graphic notation corresponding to a visual sememe.

Interestingly, as predicted by Saint-Martin,411 these sememes are continuous

and spatialized topological entities. They are in fact only rarely defined by clear

boundaries, and their meaning in effect depends on their spatial arrangement and

spatial relations with other sememes. Their spatial location becomes essential to

their meaning.412

Sememes also vary greatly in size, according to the amount and level of inform-

ation sought. The largest entity being the whole diagram, a sememe comprising

the whole iconic text, which can be read by a reader presented with the diagram

in Figure 41 for example as: ‘This is the diagram of an endleaf structure.’ More

information is gathered by delving deeply into the diagram with sememes becoming

411. Saint-Martin 1990.
412. Kubovy 1981; Pinker 1990; Van Valkenburg & Kubovy 2003.

151



Figure 36. Example graph schema showing the graphical conventions.

smaller and more numerous the more information is read and infered from the

diagram.

6.2.2. Number of possible visualizations

Looking at the visual description graph schemas, it is clear that the number of

possible nodes is limited. One could argue that, at least for the simplest structures,

it would be possible to exhaust all the possibilities and resort to drawing all of

these without any transformation. Maybe for such cases going through the effort

152



of writing an algorithm to generate these diagrams is redundant. But, this is not

the case, it can be argued, for at least one good reason: even for a limited number

of options, the number of possible visualizations quickly escalates to large numbers.

Let us consider two examples: board markers, and endleaf structures.

6.2.2.1. Board marker visualizations

At first, one of the simplest components in the schema will be considered: board

markers, i.e. lengths of animal skin adhered to the inside of a board, projecting

from the edge. There are two views for each diagram: cross-sections and views

from above; they differ slightly in the information they can provide in regards to

the attachment method: for this they will be considered separately. There are two

sets of visualizations: (i) location and (ii) position. There are two possible options

for location (fore-edge left, fore-edge right), and another two for position (over

or under turnin). Finally, (iii) there are two different methods of attachment (glued,

nailed), but only the cross-section view can show them both. Having considered

all this one can calculate the number of permutations413 for each view:

(i)

n!
(n-r)!

= 120 + 60 = 180

=[
(Cross-section view + Front view) =

+ n!
(n-r)!]=

6!
(6-3)!

=[ + 5!(5-3)!]=

In principle, it would be possible to hand-draw all 180 different diagrams; it

would take some time, but it is feasible. However, should any new characteristic

of board markers need to be highlighted, they would all need to be redrawn once

more.

By considering more complex structures, it can be seen how the number of

permutations quickly escalates to unmanageable numbers.

413. In mathematics, a permutation can informally be defined as each of several possible ways
in which a set or number of things can be ordered or arranged, where the order is important.
(Weisstein [s.d.]).

153



Figure 37. Board-marker graph schema (see §6.2.1).

6.2.2.2. Endleaf structure visualizations

Endleaves are defined as the leaves of a variety of sheet materials found at the

front and back of a bookblock. There are two main types: separate and integral

endleaves. The former are added by the binder before the book is sewn and can

be described as being composed of units and components; the latter are blank

154



Figure 38. Endleaf graph schema (see §6.2.1).

155



leaves at the front and/or the back of the textblock, which are used as endleaves.

Separate endleaves and integral endleaves can be combined.

Assuming 1 integral endleaf, 1 unit, and 1 component for separate endleaves,

one can calculate the number of possible permutations taking into consideration:

different visualizations for (i) left and right endleaves, and (ii) for structure and

use; (iii) three graphical options for the outermost gathering; (iv) the two basic

types of endleaves, integral and separate. For integral endleaves there are the op-

tions of (v) the number of leaves, and (vi) the choice of pastedowns (for use visu-

alizations). Separate endleaves are described in (vii) units and (viii) components,

both of which can vary in number. For each component there are (ix) the choice

of pastedowns (for use visualizations), (x) the material, and the visualizations dis-

tinguish between paper and parchment by modifying the line thickness, (xi) the

modality of attachment (glued, sewn, NC, NK, other), and (xii) the type. Hook

types can be (xiii) doubled, and (xiv) can be further divided into endleaf hooks

and text hooks. As in most cases, when the information is marked as not checked

(NC), not known (NK), and other, the resulting visualization is dealt with according

to the standard uncertainty visualization principles, but these are not generally

differentiated amongst themselves. Having considered all this one can calculate

the number of permutations for each view:

(ii) n!
(n-r)!

× # basic units =

=
27!

(27-9)!
× # basic units =

= 1.700755056 × 1012 × # basic units

1,700,755,056,000 of possible different output diagrams is too large a number

for hand-drawing to be a feasible option. An automated visualization instead can

generate any of such possible diagrams when and if needed, based on one and

only one transformation algorithm.

156



6.2.3. Visualizations

Each visualization takes a title with information on the shelfmark of the volume

being described and the specific visualization, e.g. ‘left endleaves (use)’. A verbal

summary description is also provided giving details on the types of components

found in the structure. These textual elements act as verbal labels. As seen in

chapter 3, there are benefits in combining visual and verbal information. The labels

can fix the otherwise ambiguous meaning of visual stimuli allowing for beta reading

modality.414 Furthermore, they aid subsequent correct interpretations and recall

of the information.

Diagrams make general use of a set of conventional shapes. These shapes have

been selected and devised in such a way that what they represent should be imme-

diately clear to a viewer familiar with bookbinding structures and diagrams. The

choice of shapes is meant to be consistent and prototypical enough to be repres-

entative of any real-life shape, but they often also follow conventions found in the

literature. Conventions are both followed for the particular views that are usually

utilized to visually describe a binding structure and the shape of its components,

but not blindly: they are modified and integrated when incapable of conveying

all the information needed.

In chapter 4, looking at the different types of drawings found in the literature,

it was concluded that schematic line drawings are the most versatile and at the

appropriate visual level of abstraction to be used as visualizations of bookbinding

structures. The black and white schematic line drawings are set against a white

background and only in a few instances some surfaces are filled in with grey to

represent a reference element as background, e.g. the board for furniture.

Most visualizations make use of different views of the same structure to com-

municate all the information recorded in the schema (see Figure 39). These re-

semble projected views found in mechanical drawings (see an example in Figure

40). However, the diagrams do not follow the strict rules and conventions of

technical drawings, because the emphasis is not on the exact representation of

one object, but rather of its prototype, and also because they are generated from

414. Eco 1997.

157



a very limited number of measurements. The series of conventional lines and

symbols found in technical drawings — e.g. centrelines for circles, dimension

lines, projection lines, etc.415 — would require a reader aware of the complex set

of symbols, and they would clutter the drawings with information that, considering

the project’s scope, would be superfluous. However, following the commonly

used drafting convention for invisible edges, those structures (or part of structures)

that would be invisible from the selected view point are drawn with dashed lines

(see Figure 41).

For certain elements, the representation of their precise length does not add

much information, considering the prototypical nature of the diagrams, and would

also impede the immediate comparison between structures of different volumes.

In such cases, only the extremities of the element are drawn with a faded area in

the middle to hint at continuity of the parts. This exploits the Gestalt principle

of continuity, compelling the eye to move through one object and continue to the

other416 (see Figure 41).

6.2.3.1. Level of abstraction and semantic entry point

In the schema, there are instances of descriptions that do not convey enough

information to be usefully or successfully visualized. These are signalled in the

visual description graph schemas by a dagger symbol (†). What happens in these

cases is that the description does not lead to the correct semantic entry point for

the concept being introduced.417 A useful example to consider is the list of possible

kinds of furniture: (i) articulated metal spines, (ii) bosses, (iii) corners, (iv) full

covers, (v) plates, (vi) ties, (vii) catchplates, (viii) clasps, (ix) pins, (x) straps, (xi)

strap collars, (xii) strap plates. Types i to vi can not be usefully drawn, whilst vii

to xii can (see Figure 41). By comparing the information that is available for the

first group (see Figure 42) with the information for the second group (at least vii

to x) one would notice that these are further subdivided into more specific typo-

415. Willard 2009.
416. Koffka 1935.
417. Jolicoeur et al. 1984, see chapter 3.

158



Figure 39. Complete visualization of the left board for Vol. 18.3ig. Note the series of dif-
ferent views: head, inner and outer surface, tail, fore-edge, spine, horizontal and vertical
cross-section.

159



Figure 40. Example of mechanical drawing with three projected views (Willard 2009,
fig. 58, p. 63).

Figure 41. Complete visualization of the fastenings for Vol. 19.ιδ. Note the faded area in
the strap visualization and the dashed lines for invisible edges.

160



logies. Each subcategory type comes with a specific definition that provides details

about its functional components. Another solution would have been to describe

each main typology according to a set of components, but because the number

was limited, a specific term carries in itself enough information to allow for a

diagram to be distinguishable from others, and specific enough to resemble the

original in its formal and functional components. Type xi and xii, are a particular

case because, whilst not further subcategorized, they form in fact a category on

their own — that of a metal fixture at the end of a strap — of which each of them

is in essence a subcategory, and the definition of their descriptive naming term

carries information on their spatial and formal characteristics.

The same is not true for articulated metal spines, bosses, corners, full covers,

plates, and ties. Whilst these do hold some additional information in the schema,

this does not concern their material components or their form, but rather their

making. There is information about the material they are made of and how this

was worked, and, as for other furniture, whether they go through the pastedown

or not. In addition, bosses’ profiles are further described by means of hand

drawings, and not as XML data, and, therefore, they are not usable for this project

(see Figure 43 for some examples of bosses constructions). The terms describing

these furniture types are indeed the basic level of abstraction for their category

— i.e. bookbinding furniture — but, as predicted, this is not usually a good se-

mantic entry point for a visual representation, which needs to be a step down to-

wards the particular. Arguably, i to vi could have been rendered with general

shapes, but this would have added little useful information: all books with bosses

would have had the same drawing, adding nothing to a simple search of the

database for the word ‘bosses’. For this automated visualization methodology to

be usefully applicable, more information about their material components and

form would be needed.

6.2.3.2. Diagram elements

For each visualization, there are two main kinds of elements: the proper diagram

elements that, when read by the reader, can signify the binding structure described

161



Figure 42. Visualization of the Ligatus schema: available information for furniture deemed
not drawable or not worth visualizing.

162



Figure 43. Examples of bosses found in Gothic bindings (Szirmai 1999, fig. 9.55[1-5],
p. 264). Note how more complex the description of bosses could be and how their pecu-
liarities are not simply linked to the overall shape of the dome.

in the database, and a set of reference elements. The latter are not mentioned in

the descriptions, but help to put each element of the structure visually in context,

thus facilitating their signification and allowing for beta reading modality418 of the

diagram elements. Reference elements are also designed to be visually less prom-

inent, by greying them out or thinning their lines.

6.2.3.3. Spatial arrangement

In the diagrams, each component is clearly kept distinct from the others by

keeping them all spatially separated by a standard distance. This causes complex

structures to extend outward from the centre of each unit and from the reference

elements (see Figure 45). This convention makes it easier to identify similar

structures, and to understand the role of each component. Consider for instance

the examples [g] and [h] in Figure 44 and note how the various components are

drawn very close to each other at the fold. In example [h] this is a consequence

of the naturalistic nature of the drawing. In example [g], whilst the drawing at-

tempts to be symbolic, the distance between the components is still kept to a

minimum to try to attain a more naturalistic appearance. Inevitably, the author

is forced to fan out the structure to be able to convey the whole structure more

clearly. However, the closeness of the lines of different components causes them

to be perceived in close relationship. This is due to the Gestalt principle of prox-

imity: objects near each other tend to be grouped together whether in relationship

or not.419

418. Eco 1997 p. 336.
419. Koffka 1935.

163



[g]

[h]

legend: board
first gathering

[d] [e] [f]

[a] [b] [c]

Figure 44. Examples of schematic endleaf diagrams in the literature ([a-c] Szirmai 1999,
fig. 8.4[a,c,g], p. 147; [d-f] Carvin 1988, p. 34[0A,0B,1A] — the legend has been translated
from the original French; [g] Middleton 1996, fig. 26, p. 43; [h] Cockerell 1953, fig. 19[v],
p. 81, this example tends more towards naturalism).

164



Moving each component and unit outward avoids the risk of perceiving stronger

relationships where a weaker one would be needed to understand the structure.

This happens at the expense of naturalism, but it offers more clarity and commu-

nicability, which is a worthy trade off (see Figure 45).

Figure 45. Example of a complex endleaf structure with both integral (one leaf) and
separate (two units) endleaves (Vol. 4725.3162). Note how the units and components
are moved outward, away from the reference gathering.

6.2.3.4. Thread paths

Thread paths move and link elements in three dimensions. Diagrams always

show two-dimensional views. Whilst this is not generally a problem for sewing

diagrams, endband cross-section views present some issues.

Cross-sections of endbands are used to show the relation of cores and the

bookblock, but also to show the path of the thread around the cores and inside

the quires. In these visualizations, the third dimension (along the length of the

endband) is not shown and therefore it is impossible to depict how it moves along

the endband core (see Figures 46 and 47). In Figure 46[e1] this convention is not

followed, but the result is rather cumbersome.

Manuals of how to work endbands make scarce use of these cross-sections,420

or do not use them at all,421 whilst they are common in articles and monographs

on the history of bookbinding. The established convention of drawing the endband

in two dimensions is followed here, as these diagrams are not meant to instruct

how to work endbands. The diagrams act as a way to visualize with as few lines

as possible the essence of the thread path and the endband sewing. Complex

420. Greenfield 1990.
421. Bibliothèque Nationale (France) 1989.

165



structures become difficult to follow in these diagrams (e.g. Figure 47 [f] and [g]),

however, the diagrams as a whole can act as a symbol for the structure. In fact,

once encountered, given the distinct appearance of each path and the limited

number of options visualized,422 these are easily memorized and recognized as a

whole, without having to discern the path of the thread. Like all diagrams produced

in this project they are ruled by ratio difficilis,423 as the organization of the expres-

sion is determined by the organization of the content — i.e. they are visually

similar to what they represent — however, as representations of types, they also

constitute preformed expressions — i.e. one these types of endband thread path

diagram will always be represented in the same general manner — and can thus

also be considered as ruled by ratio facilis.424

6.2.3.5. Uncertainty

As outlined in the previous chapter, when dealing with data on historical

bookbinding structures, one has to accept that a degree of incompleteness, uncer-

tainty, and imprecision is inevitable. Given the particular nature of the automat-

ically generated diagrams for this project, this has to be generated parametrically

and applied without modifications to the overall design.

Ideally, one would not want to draw components which are not accurate, but

this is not possible because leaving out the inaccurate part of the diagram leads

to misunderstanding, and potentially it could make the whole diagram meaningless.

This is because the operation would leave gaps in the plane of the diagrams, and

zero signs would still be read as absence of phenomena, and not as missing data.425

For example, let us consider the case of double hooks (see Figure 48). As it will

be seen in detail in the next chapter, this type of hook is always drawn as being

formed out of a conjoined bifolium. However, since the information provided is

not sufficient to rule out the case of it being made out of two separate sheets, the

422. The schema distinguishes between seven different types: warps only, no bead, front bead,
no front bead, reversing twist, Greek single core, and Greek double core.

423. Eco 1975/2008; See chapter 3.
424. Eco 1975/2008.
425. Bertin 1983/2011.

166



Figure 46. Examples of cross-section visualizations for endbands found in the literature.
([a] Jäckel 1975, fig. 3, p. 210; [b] Greenfield 1990, p. 59; [c] Szirmai 1999, fig. 6.10[c],
p. 77; [d] Clarkson 2005, fig. 17, p. 16; [e1-2] Conn & Verheyen 2003b, p. 1).

167



[a] [b] [c]

[d] [e] [f]

[g]

Figure 47. Examples of cross-section visualizations for endbands: [a] warps only (Vol.
219); [b] no bead (Vol. 3793.2703); [c] front bead (Vol. 4820.3212); [d] no front bead
(Vol. 221.105α); [e] reversing twist (Vol. 2938.2149); [f] Greek single core (Vol.
2525/1814); [g] Greek double core (Vol. 2329.1633).

168



conjoined portion is drawn as uncertain. If one were to leave out the uncertain

portion of the diagram, the diagrams would always show a hook made of two

separate sheets, which might be true in some cases, but certainly not always.

Let us consider now the case of an endband cross-section diagram for which

some information is uncertain (see Figure 49). The sewing structure for this par-

ticular endband is described as no-front-bead to indicate that no front bead was

present, but also that it was not possible to check for the presence of the back

bead. Also, the schema does not have information (in XML format) in regard of

the preparation of the gatherings for the endband and the shape of the endband

core cross-section: the gatherings are always drawn as not cut at head or tail to

accommodate the endband and give cores as circular cross-section shapes, but

these are indicated as uncertain. If, as in Figure 49 [b], one were to remove all

uncertain information, the diagram might only show true and certain information,

but it has no meaning and becomes useless.

As a general rule, for this project, there are two main ways of showing that some

data is uncertain: blurring, and fading. In the examples so far, blurring was used

as the means of dealing with uncertainty.

Eco426 and Saint-Martin427 both point out how images are macroscopic blocks

of information whose meaning is identifiable with pertinent units, but these are

not independent and are essentially continuous as they cannot usually be considered

in isolation. For these reasons, the application of uncertainty needs to always take

into consideration how each unit relates to the others in their block of information.

We have seen how simply deleting them is not an option, and blurring becomes

a feasible way of showing both the shape to be read to keep the overall meaning

and highlight uncertainty at the same time.

There are cases when deleting uncertain information does not necessarily result

in the loss of the meaning of the whole information block. For example, sometimes,

it is possible to identify precise boundaries for a unit. If the uncertain information

happens to coincide with the extremities of this information block with clear

boundaries, deleting the uncertain information does not result in a nonsensical

426. Eco 1975/2008.
427. Saint-Martin 1990.

169



Figure 48. Detail of right endleaves of Vol. 2177.1482 (see Figure 72 for the whole dia-
gram) with uncertainty regarding the conjoined leaves of the double hook.

drawing. In these cases, one can avoid drawing something that might not be

completely or always true by fading the diagram where the information is uncertain

(see Figure 50). The schema does not define the typology of the shape found at

the end of board grooves towards the spine, nor the measurement of the groove.

One could draw the simplest type, a straight cut, and blur it for uncertainty.

However, since the end of the groove lies at the extremity of its information block,

one could fade it, thus indicating that it is not known how that part is to be drawn,

and at which point along the board edge. This way one could avoid drawing

something that might be correct only in some cases. The lines are also faded at

different lengths to avoid creating visual groups that might be read as coherent

and possibly misinterpreted (e.g. a gradually shallower groove).

Interestingly, units that are completely isolated and defined by clear boundaries

need to be blurred and not faded off — or in fact deleted as fading off acts by

deleting those uncertain parts of a continuous unit — as other units might depend

on it for their meaning, and the complete absence of information cannot be flagged

as uncertain.

In practice, if the uncertain part to be drawn is connected at both ends with

other units, or is a completely discreet unit, this project resorts to blurring to flag

uncertainty. When the uncertain part is only connected at one end with other

units, that part can be faded off without compromising the overall meaning of

the diagram. Both can be easily applied to any shape within SVG, thus allowing

for uncertainty to be added to a certain component parametrically when needed.

Blurring is the result of a Gaussian blur filter applied to a specific element. The

170



Figure 49. Cross-section of endband with no-front-bead sewing description for Vol.
221.105α. In [a] the diagram is complete with the uncertain information, whilst in [b]
all uncertain information has been removed, leaving a meaningless diagram.

Figure 50. Head view of the left board of Vol. 8.3γ. Note the fading off of the groove
towards the spine.

fading effect is given by a gradient fill going from black to white in the appropriate

direction.

171



It should be noted that fading is also used to signal continuing lines when the

representation view only requires a portion of the structure or element to be

visualized (see for example the boards in the cross-section view in Figure 41). In

both cases, fading indicates that the information beyond what is drawn is present

but not visualized, either because unimportant, or because uncertain. However,

the former is applied to elements that are fixed and always present in all examples

of the visualization, whereas the latter is applied to parts that vary from instance

to instance. Therefore, confusing the two should not be an issue.

6.2.3.6. Accuracy and imprecision of measurements

The Ligatus schema requires that only few measurements are taken from the

books: the dimensions of the book, the board thickness, the position of each

sewing station, and the measurement of the height of the page at the fold. However,

in the diagrams, each line has to be given a certain length. Therefore, an educated

guess has to be made for sections that had no direct measurement recorded during

the survey. For example, the thickness of the bookblock is given by subtracting

the thickness of the boards from the overall thickness of the book, and so forth.

In other cases the aim was to keep the proportions between the components correct

and meaningful, like the thickness of sewing supports and endband cores. In ad-

dition, because of the principle exposed above of moving various components

outward to keep them perceptually distinct, but at the same time grouped, some

measurements are exaggerated, such as the gaps between endleaves or the board

and the endband in greek-style bindings (see Figure 51).

172



Figure 51. Front view of the double-core greek-style endband of Vol. 2329.1633. Note
the gap between the board and the endband, and the smaller one between the two sections
of the endband.

In the previous chapters we have seen that not all physical variables are psycho-

logically perceived in the same way, and that small differences between values of

physical variables that are not prägnant features are not usually noticed. Further-

more, quantities are usually represented continuously in our minds as ratio values

as opposed to absolute values.

All this considered, and given the prototypical nature of this project’s diagrams,

precise and accurate measurement representation is not necessary. In fact, high-

lighting which measurements are not precise, would result in most lines being

flagged as imprecise, making the diagrams cumbersome whilst adding information

of little value to their power to communicate.

The only cases in which imprecision is flagged is when measurements are

sometimes given and other times not. This happens for the coverage of panels by

part of stuck-on endbands. The schema here calls not for precise measurements,

but for ratios of coverage in percentage values. It was decided to draw a diagram

173



even when the data provided was incomplete, by assuming, in such cases, a per-

centage value of 100%. In order to distinguish between diagrams of books in

which the percentage value was given and the others, imprecise stuck-on endband

lengths are flagged by creating a halo around the lines in question (see Figure 52).

Figure 52. Cross-section and front views of the stuck-on endband of Vol. 2215/1519. In
the description the surveyor gave a NK value for the percentage of the coverage of the
panel by part of the endband.

A halo is a ‘perceptual distinction’ feature428 that does not change the shape

onto which it is added, and that can be universally applied to graphics through

an ad hoc series of SVG filters. The reduced contrast with the background makes

the feature perceptually less salient, whilst the shape contour is left unaltered.

Halo differs from blurring as the line onto which it is applied remains crisp.

6.2.3.7. Place-holders for uncertainty inherent in the visualization process

One last category of uncertain visualization, as discussed in the previous chapter,

is that of components that can be characterized by convoluted contours not con-

stituting any particular function above that of decoration, or components with

random and irregular outlines. These are examples of uncertainty that result from

the visualization process: these components have been positively identified during

the survey, but their translation into a visual language leads to uncertainty. These

components are substituted in the diagrams by abstract symbolic shapes, acting

428. Ware 2013.

174



as place-holders for the component, but may not resemble what was described,

apart from functional parts.

Sketchy lines are used to flag this kind of highly abstract visualization. The

skecthy lines are attained by applying a complex series of effects and filters to the

SVG shape path.429

Figure 53, compares the diagram of a catchplate to a photograph of the same

element found on the volume: all the functional elements are present, but the

decorative outline of the metalwork is simplified in the diagram.

Figure 53. Above view of the roller-round-bar catchplate for Vol. 183.97 (left). The shape
of the catchplate could involve complex shapes and is therefore rendered with sketchy
lines. Compare with an actual photograph (right) of the volume’s catchplate.

One common feature of all of the visual effects discussed in the last few sections

is that, purposefully, all follow a modular approach. Each act on a different V1

channel and is cumulative with the others. It would be in fact possible to have,

for example, both a component drawn with sketchy lines because of the innate

nature of its complex outline and blurred because of uncertainty. For example,

patch spine linings are characterized by random and irregular outlines and are

therefore drawn with sketchy lines. When a spine lining is described as being only

on a selected number of panels, the lining has to be blurred as the schema does

not indicate exactly on which panels it has been pasted (see Figure 54).

This modular approach allows the system to be flexible and to always use the

same set of visual cues for the same kind of message.

429. Thanks go to Dr Alejandro Giacometti for his help in the development of this SVG filter.

175



Figure 54. Example of a diagram that combines sketchy lines and blurring to indicate
different kinds of uncertainty.

Summary

The Ligatus schema, at the basis of the dataset, exploits XML’s hierarchical

nature to create a formalized hierarchical structure for the description of book-

bindings. In turn, this project makes use of XML’s transferability of data to select

and present information in graphic form through other XML technologies, namely

XSLT and SVG.

Structures described at the right level of abstraction and semantic entry point

are capable of being translated into diagrams. This is true for most structures de-

scribed within the Ligatus schema; the few exceptions will be highlighted in the

following chapter.

Data is presented in graphic form — accompanied by verbal labels — following,

where possible, the typical conventions of shapes and views found in the literature.

New graphic conventions are introduced to ensure correct communication of

both material components and form, and to highlight uncertainty and imprecision.

The next chapter will examine in detail the transformation of endleaf structures

and introduce more generally all other transformations. A knowledge graph schema

will be provided for each transformation.

176



Chapter 7. Transformations & analysis

Yet this little body of thought, that lies before me in the shape

of a book, has existed thousands of years [...] To a shape like

this, so small yet so comprehensive, so slight yet so lasting, so

insignificant yet so venerable, turns the mighty activity of Homer.

Leigh Hunt, Among my books, 1823, p. 22.

This chapter discusses the transformations based on the St. Catherine’s descrip-

tion of bookbinding structures. In the previous chapter, the conventions used in

all transformations were outlined. As an example, the transformation of endleaf

structures is covered here followed by the knowledge graph schema430 and a gen-

eral introduction for the rest of the binding structure transformations. Endleaves

can be rather complex structures, but are also quite compact in their visualization,

making this transformation an ideal example. Appendix C shows the complete set

of coding for this transformation.

The last part of the chapter considers the types of human error that exist in the

dataset. Empirically, it seems that these errors could be easily identifiable once

the erroneous encoding has been visualized. Implementing transformations as

outlined in this dissertation can improve data accuracy.

430. See §6.2.1.

177



7.1. A complete visualization example: endleaves

As seen in the previous chapter, endleaves are found at the front and back of

a bookblock. There are two main types of endleaves — separate and integral —

and these types can be combined. There are normally two sets of endleaves and

the endleaf structures at either end of the bookblock are not necessarily identical.

Any part or parts of the endleaf components can be pasted to the inside of a

cover or to the boards after the book is covered. These are referred to as

pastedowns. The fact that part of an endleaf is used as a pastedown does not alter

its identification within an endleaf format, it is simply an additional function added

to that binding component.431 From one description, one can then distinguish

between a structure visualization that shows the endleaf format regardless of which

parts were used as pastedowns and a use visualization showing how the particular

endleaf format was used in the binding.

7.1.1. Visual description framework

Each endleaf visualization is drawn as a cross-section comprising two main

parts: the two outermost gatherings as reference elements, and the actual endleaves

developing from or on top of the outermost gathering.

Figure 57 reproduces the knowledge graph schema for the endleaf diagrams.432

Note that the only differences between the structure and use visualizations are the

special case of pastedowns. The descripion path reaches the pastedown level only

in use visualizations.

431. Ligatus 2013.
432. As covered in §6.2.1, graph schemas are a graphical representation of the information in-

volved in each bookbinding structure description and transformation. At a glance the reader can
appreciate the complexity of a description/transformation, and graph schemas can also be followed,
step by step, each node adding a new piece of information, each choice branching in parallel in-
formation paths. Graph schemas, in fact, present all the information needed for a description/trans-
formation in one page (or series of related pages for very complex structures).

178



Figure 55. Example of endleaf diagram for Vol. 2200.1504: left endleaves, use visualization

Left (use) Left (structure)

Right (use) Right (structure)

Figure 56. Complete set of endleaf diagrams for Vol. 2200.1504.

179



Figure 57. Graph schema for endleaf diagrams (see §6.2.1 for an explanation on how to
read graph schemas and their meaning).

180



7.1.2. Shapes: standard endleaf components

Endleaf structures are drawn diagrammatically in cross-section and in relation

to the outermost gatherings.433

Endleaf diagrams are composed of three main sets of elements: (i) the reference

elements (i.e. two outermost gatherings), (ii) the endleaves, and (iii) their attach-

ment modality.

7.1.2.1. Reference elements

A single outermost gathering as reference element would suffice for the great

majority of endleaf structures. However, for consistency sake, and to accommodate

for the rare case of two gatherings sewn through the fold inside a single text-hook

endleaf,434 two reference elements are drawn for all visualizations.

Since these are reference elements, they are, as per convention, drawn as greyed

out so to be visually less prominent. However, in the case of integral endleaves,

the last gathering is not greyed out. In the case of separate endleaves of the text-

hook type, the gathering is shortened to accommodate for the extra leaves around

it. However, the shape is not fully parametrical, i.e. it does not become shorter

and shorter depending on the number of text-hook components, as it is the case

for the endleaf components. This is to keep the element graphically distinct in its

appearance and its spatial relationships with the other endleaf components: the

gathering is shortened to accommodate for text-hook components, which are set

outwards from it, as they usually do in respect to any reference element.

Their shape is closed at the fore-edge to keep them graphically distinct from

the endleaves, which are instead open at the fore-edge. This graphic convention

helps to distinguish textblock gatherings from the endleaves. In the case of integral

endleaves, they are drawn as emerging from the outermost gathering, and kept

separate at the fore-edge (see Figure 58).

433. This projects follow here and is indebted to the convention established by Pickwoad for
the survey of the volumes of the library of St. Catherine’s Monastery (Pickwoad & Gullick 2004).

434. Ligatus 2013.

181



Figure 58. Example of integral endleaves (Vol. 235.114).

7.1.2.2. Endleaf components

The graphic representation of all endleaf structures can be thought of as being

composed of a very limited set of standard components. For separate endleaves,

five basic components can be identified: (i) a fold; (ii) a hook — be this separate

from or wrapped around the outermost gathering/s; (iii) an outside hook; (iv) a

guard; (v) a single leaf. Hooks can be doubled, i.e. they can be formed by a bifo-

lium folded around to form the hook. Integral endleaves (see Figure 58) are obvi-

ously limited to one basic shape: a leaf emerging from the outermost gathering.435

fold

hook

outside hook

guard

single leaf

Figure 59. The five basic components of separate endleaves (after Pickwoad & Gullick
2004).

7.1.2.2.1. Pastedowns in use visualizations

Each of the basic components could be pasted to the cover or boards as

pastedowns. In these cases, in the use visualizations, they are drawn open as if the

435. Pickwoad & Gullick 2004.

182



Figure 60. Examples of schematic endleaf diagrams in the literature ([a-c] Szirmai 1999,
fig. 8.4[a,c,g], p. 147; [d-f] Carvin 1988, p. 34[0A,0B,1A] — the legend has been translated
from the original French).

cover or board onto which they are pasted were open at 180°,436 also, a patterned

area is added under the pastedown symbolizing the adhesive (see Figures 55 and

56). This convention makes the need to draw the board redundant, leading to less

cluttered diagrams; also, opening the endleaves at 180° makes the diagrams more

easily comparable and separates more clearly pastedowns from flyleaves (i.e.

endleaves that have not been pasted down). Let us compare, for example, this

convention with those followed by other authors in the literature (Figure 60).

Both Szirmai437 and Carvin438 make use of the board as reference element in their

drawings and depict the structure as open at 45° or 90°. In particular, Carvin

(Figure 60[d-f]) draws the board as flat, the outermost gathering — or the first

gathering as he refers to it — straight up forming a 90° angle with the board, the

flyleaves open roughly at a 45° angle, and the pastedowns obviously flat and ad-

hered to the board. Szirmai (Figure 60[a-c]) instead draws the board open at 45°,

436. Pickwoad & Gullick 2004.
437. Szirmai 1999.
438. Carvin 1988.

183



the pastedown adhering to the board, the gathering flat, and the flyleaves either

flat or at an acute angle with the gathering.

The convention chosen for this project allows producing less cluttered visualiz-

ations. These visualizations are also more flexible as the convention permits to

draw complex structures without affecting the overall appearance of the diagram.

If one were to visualize complex structures following Szirmai’s or Carvin’s diagrams,

one would have to squeeze these between the outemost gathering and the board

that is fixed at 45° or 90°. Whereas, not having a fixed element apart from the

outermost gatherings, allows expanding without limit the visualization, and to

accommodate multiple endleaf units without having to compress these in a fixed

space. See for example Figure 65.

7.1.2.3. Attachment modalities

In separate endleaves, each component is attached to the bookblock by means

of an attachment modality: glued or sewn.

7.1.2.3.1. Sewn

Figure 61. Example of endleaf diagram with sewing thread convention (Middleton 1996,
fig. 26, p. 43).

When a component is described as sewn, a short line is drawn from the centre

of the unit towards the spine. This symbolizes the thread piercing the endleaves

and going out towards the spine through the fold.439 The same convention can be

observed in Middleton’s example in Figure 61. Note the different convention

439. Pickwoad & Gullick 2004.

184



followed by Szirmai in Figure 60).[a-c] instead, where a single dot at the centre

of the fold is drawn to symbolize the presence of the sewing thread. This is taken

as representing a cross-section of the thread. However, the convention of drawing

the line through the fold represents both the presence of the sewing thread and

the action of the thread that keeps the components together — both physically

and here visually — by crossing them all at the fold.

7.1.2.3.2. Glued

When components are described as glued, a patterned area is drawn towards

the outermost gathering symbolizing the adhesive (see Figures 62 and 63). A

similar convention is followed by Middleton in Figure 61).

Figure 62. Example of glued component (Vol. 195).

Figure 63. Example of glued components (Vol. 1332.781).

7.1.2.4. Component material

In the dataset, each component can be made of either paper or parchment.

These two materials are differentiated in the diagrams by use of a thicker line for

parchment components, to help identify and recognize similar structures (see

Figure 64).

185



Figure 64. Example of endleaf structure with internal parchment guard (Vol. 1644/1080).

7.1.2.5. Spatial arrangement

As customary, the elements in the diagrams are kept distinct and spatially sep-

arated. This convention was discussed in more detail in the previous chapter

(§6.2.3.3).

7.1.3. Shapes: different types of endleaves

Figure 65. Example of a complex endleaf structure with both integral (one leaf) and
separate (two units) endleaves (Vol. 4725.3162).

7.1.3.1. Integral endleaves

Integral endleaves are characterized by the number of leaves that are left blank

and used as endleaves. The only additional piece of information is whether a blank

sheet is used as pastedown or not. Material and attachment modalities do not

need to be specified, as they would obviously be the same as those of the gathering

to which they belong. Examples of integral endleaves can be seen in Figures 58,

65, and 66.

186



Figure 66. Example of integral four endleaves with one pastedown (Vol. 26.4β).

7.1.3.2. Separate endleaves

Separate endleaf components are differentiated into six different types: single

leaf, fold, endleaf hook, text hook, outside hook, and guard. All components,

apart from single leaves, are parametrically drawn to accommodate for any other

component they wrap around to form a unit.

7.1.3.2.1. Single leaf

Figure 67. Example of single leaf unit (Vol. 2355.1658).

Figure 68. Example of separate pastedown (Vol. 1651.1083).

A single leaf component is, as the name suggests, a single sheet pasted to the

bookblock (tipped leaf) or to the inside of a board or cover (separate pastedown),

which is not conjugate with any component in the adjacent endleaf unit. When

this constitutes a separate pastedown, it leaves an open joint, i.e. a gap between

the pastedown and the bookblock.440

440. Ligatus 2013.

187



The graphic representation of single leaves is always limited to its standard

shape (see Figure 59). The only parametrical element to their graphic representa-

tion within the diagrams is their spatial positioning.

7.1.3.2.2. Fold

A fold component is characterized by a bifolium (two complete leaves). Fold

components can be used singly (Figure 56) or in multiples (Figure 69). Each path

is split in two halves to allow more granularity with drawing uncertainty. For ex-

ample, uncertainty regarding pastedowns makes only the upper half blurred.

Figure 69. Example of fold components (Vol. 227.107).

7.1.3.2.3. Hooks

Hooks are formed by a single leaf or bifolium of sheet material (double hooks)

folded along the spine edge to create a stub on the side of the leaf towards the

textblock. Hooks may be used in combination with other endleaf components

(endleaf hook) or folded around the outermost text gathering at either end of a

texblock (text hook).441

7.1.3.2.3.1. Endleaf hook

Endleaf hooks are used alone, in multiples, or as a component within an endleaf

unit, but are not folded around the adjacent outermost text gathering.442

441. Ligatus 2013.
442. Ligatus 2013.

188



Figure 70. Example of endleaf-hook components (Vol. 2751.1991).

7.1.3.2.3.2. Text hook

Text hooks are similar to endleaf-hooks but they are folded around the outer-

most gathering/s at either end of a bookblock. Very occasionally, two gatherings

may be sewn through the fold of a single text-hook endleaf.443

Figure 71. Example of text-hook component (Vol. 103).

7.1.3.2.4. Outside hook

Outside hooks are formed in the same way as the other hooks, but the stub

they create lies on the outside of the leaf, away from the textblock. Outside hooks

can be found used on their own, in multiples or as one of several components

within a single endleaf unit.444

443. Ligatus 2013.
444. Ligatus 2013.

189



Figure 72. Example of outside-hook component in conjunction with text hooks (Vol.
2177.1482). Note the blurring of the conjoined portion of the double hook. Also, note
the blurring of the outside-hook full leaf (see page 194).

7.1.3.2.5. Guard

Guards are narrow strips of sheet material folded lengthways to create two

stubs (usually of unequal width, with the wider stub to the outside of the

bookblock) and sewn through the fold.445 See Figures 64 and 73.

Figure 73. Example of guard (Vol. 3893.2742β).

7.1.4. Visualization problems

The endleaf transformation poses, two main kinds of problems. First, the schema

used for the dataset is not always able to convey all the information that would

be needed for an accurate visualization: either it lacks some descriptive properties,

or it cannot fully deliver the form of the structure, the relationships amongst the

elements and their spatial arrangement. Second, comparing the descriptions in

445. Ligatus 2013.

190



the database with the hand drawings of the endleaf structures in the survey forms,

one can notice a somewhat poor understanding of the description conventions

set by the schema — or their inconsistent application — by part of the surveyors.

These problems are investigated in the fellowing sections.

7.1.4.1. Schema limitations

There are two main kinds of limitations in the schema: some are due to the

prescribed structure of the description, whilst others are linked to it not encom-

passing all organizational possibilities.

A first problem is found in the way in which integral endleaves are described.

The schema requires the number of integral endleaves in the gathering and

whether they have been used as pastedowns. The surveyor can thus only state that

at least one of the leaves had been pasted down. In the automated drawings, the

algorithm always interprets that only the outermost leaf is used as a pastedown,

which is arguably the most probable case. However, the hand drawing done for

Vol. 5577.3706a/3706a1 (Figure 74) shows that both leaves had been pasted down,

but the survey form did not allow for this.

The schema should instead allow the indication of exactly which leaves are

pastedowns.

Figure 74. Hand drawing for the right endleaves of Vol. 5577.3706a/3706a1 made during
the survey in 2007.

191



Figure 75. Generated drawing for the right endleaves of Vol. 5577.3706a/3706a1.

Attachment modalities are mutually exclusive — either sewn or glued — which

arguably covers the great majority of cases. However, there are examples in the

dataset where a component is both glued and sewn. The surveyor had to choose

one or the other modality, or add a composite modality for both sewn and glued.

There are 38 examples of the latter option.446 These can be identified through a

simple search within the database. However, to accommodate for inconsistencies

in the use of the form one has to visually inspect the hand drawings in the survey

forms. Compare for example the drawing in Figure 73 and its hand-drawn version

in Figure 76.

Figure 76. Hand drawing for the left endleaves of Vol. 3893.2742β: note how the guard
is drawn as both sewn and glued.

446. 2172.1477: ‘sewn and glued’, 2172.1477: ‘sewn and glued’, 2273/1582α: ‘sewn and tipped’,
2360/1663: ‘glued and possibly also sewn’, 2360/1663: ‘glued and possibly also sewn’, 2388/1691:
‘sewn and glued’, 2388/1691: ‘sewn and glued’, 2424/1725: ‘sewn and glued’, 2424/1725: ‘sewn
and glued’, 2424/1725: ‘sewn and glued’, 2424/1725: ‘sewn and glued’, 2515/1805α: ‘sewn and
glued’, 2530.1819: ‘glued and sewn’, 2542/1830: ‘sewn and tipped’, 2542/1830: ‘sewn and tipped’,
2567.1853: ‘sewn and glued’, 2598.1879: ‘glued and sewn’, 2878/2096: ‘sewn and tipped to text
quire’, 2938.2149: ‘sewn and glued’, 2938.2149: ‘sewn and glued’, 30: ‘sewn and tipped to text
block’, 30: ‘sewn and tipped to text block’, 3881.2736: ‘sewn and glued’, 3892.2742α: ‘glued and
sewn’, 3892.2742α: ‘glued and sewn’, 4051.2825: ‘sewn and glued’, 4051.2825: ‘sewn and glued’,
4596.3071c: ‘sewn and glued’, 4596.3071c: ‘sewn and glued’, 4598: ‘sewn and glued’, 5227.3471a:
‘sewn and glued’, 5227.3471a: ‘sewn and glued’, 527.268β: ‘sewn and glued’, 53.13ζ, 14ζ: ‘sewn
and glued’, 53.13ζ, 14ζ: ‘sewn and glued’, 550.278α: ‘sewn and glued’, 64.31: ‘it is glued to text
leaves, then sewn as part of the last quire’, 64.31: ‘it is glued to text leaves, then sewn as part of
the last quire’.

192



Figure 77. Automated drawing of Vol. 3893.2742β with guard component set as glued.

When a component is described as being glued, the schema lacks the ability to

indicate exactly to what the component is pasted. The assumption made in the

algorithm is that it is so to the following component or to the outermost gathering.

However, looking at the hand drawing in Figure 76, had the guard been described

as glued, still the automated drawing would have failed to accurately represent

the described structure, as the guard was actually pasted to both leaves of the

preceding component (see Figure 77). This leads to another common problem:

when single leaf components are also pastedowns, the surveyors have often indic-

ated the component as being glued. Looking at the hand drawings it is clear that

they meant ‘adhered to the board as pastedown’. Because a single leaf would not

extend to the bookblock across the joint, the attachment glued option is essentially

ignored for separate pastedowns. However, in case of unclear structures, where

a component is described as not known, not checked, or other, and its visual rep-

resentation is a separate pastedown that is faded off towards the fold, the glued

pattern is blurred and drawn towards the following component or the text gath-

ering. Compare Figure 79 and its corresponding hand drawing (Figure 78).

Figure 78. Hand drawing for the left endleaves of Vol. 1648.1082β.

193



Figure 79. Diagram for left endleaves of Vol. 1648.1082β with uncertain structure and
‘glued’ attachment modality.

Double hooks can be made either from two separate pieces of paper or a single

piece folded twice to create a folded stub out of a conjoined bifolium, but no in-

formation is given in the schema in this regard. In the diagrams the conjoined

portion of the path is always drawn and made uncertain (see Figures 72 and 80).

Figure 80. Detail of right endleaves of Vol. 2177.1482 (see Figure 72 for the whole dia-
gram) with uncertainty regarding the conjoined leaves of the double hook.

Whilst outside hooks, like the other hooks, and guards can be doubled, the

schema does not allow this option. In theory then, these should always be drawn

as uncertain, but this would overshadow any other uncertainty, and these were

felt as more important than an inaccuracy caused by a schema limitation.

Another problem with outside hooks and guards is that the schema only gives

the usual option to indicate that a component was used as a pastedown. Following

the usual interpretation, the outer part of the component is drawn as a pastedown.

However, there are frequent cases of Dutch endleaves whereby the full leaf has

been pasted down, leaving the sewing inside the pastedown.447 To allow for this,

the full leaf is then drawn as uncertain when the hook constitutes a unit on its

own (see Figure 72).

447. Pickwoad & McKitterick 2013.

194



There are also cases in the dataset of guards in which both stabs have been

pasted down, leaving the sewing inside. To allow for this, the lower stab is drawn

as uncertain. Compare Figure 82 and its corresponding hand drawing (Figure 81).

A similar problem can be observed for endleaf hooks entirely used as pastedowns

(see Figure 84 and its corresponding hand drawing in Figure 83).

Figure 81. Hand drawing for the left endleaves of Vol. 1628.1067.

Figure 82. Wrong diagram of left endleaves of Vol. 1628.1067. Note the blurring of the
bottom half of the guard and the rogue ‘glued’ pattern for the uncertain unit.

Figure 83. Hand drawing for the left endleaves of Vol. 5400. 3570.

Figure 84. Wrong diagram of left endleaves of Vol. 5400. 3570. Compare with hand
drawing in Figure 83.

195



The schema does not distinguish between endleaf guards and text guards (see

Figures 85 and 86). Therefore, they are always drawn as endleaf guards, the most

common option. Once again, in theory, guards should then always be drawn as

uncertain, at the cost of overshadowing any other uncertainty. As in the case of

double outside hooks and guards, it was felt that allowing for the visualization of

other uncertainties would be more important than an inaccuracies caused by the

current schema’s limitations.

Figure 85. Hand drawing for the left endleaves of Vol. 2529.1818.

Figure 86. Wrong diagram of left endleaves of Vol. 2529.1818. Compare with hand
drawing in Figure 85: the lack of a text guard option makes it impossible for this kind of
structure to be automatically generated.

7.1.4.2. Schema misinterpretations

If the endleaf structure being described was formed by more than one unit and

component, the surveyor was instructed to describe and number them counting

from the outside towards the textblock at each end. This was the usual convention

that had also been used in the paper forms for the manuscript survey. However,

this convention was not consistently followed. In some cases, the surveyor opted

to describe the structure starting from the textblock outward.448 See for example

the order of the component numbering in Figure 87.

448. Pickwoad & Gullick 2004.

196



Figure 87. Hand drawings for the endleaves of Vol. 5365.3471d. Note the order of the
component numbering.

The elements of the structures being described are all there, but the inconsist-

encies in their order, can create problems in the interpretation of their spatial ar-

rangement. Whilst a wrong order of components does not usually create problems

because of the symmetry of the components’ shape, a wrong unit order leads to

nonsensical diagrams. See for example the transformations for volume 5365.3471d

in Figures 88 and 89, and compare them with the hand drawings in Figure 87.

See also how a source XML file in which the order was corrected to be consistent

with the schema numbering convention generates a diagram resembling the inten-

ded endleaf structure (Figure 90).

Figure 88. Diagram of left endleaves of Vol. 5365.3471d. Compare with the hand drawing
in Figure 87: whilst the order of the components in the description was technically wrong,
the symmetry of the components’ shape makes the order irrelevant.

197



Figure 89. Diagram of right endleaves of Vol. 5365.3471d. Compare with the hand
drawing in Figure 87: the wrong order of the various units in the description leads inev-
itably to a nonsensical diagram with a pastedown in the innermost unit and a series of
other units that technically would have had to be sewn through the pastedown.

Figure 90. Diagram of right endleaves of Vol. 5365.3471d, automatically generated after
the order of the units in the source XML file was corrected to be consistent with the
schema numbering convention. Compare again with the hand drawing in Figure 87 and
how the diagram now resembles the intended endleaf structure.

Sometimes, the surveyor did not encode the structures properly. According to

the schema, an endleaf structure can be of two types — integral or separate —

but then, different groupings of separate endleaves should constitute a series of

units encoded within the same type of endleaves. Listing 2 shows an XML record

that did not follow the prescribed pattern of description: here the endleaf structure

has been described as being composed of three types of endleaves, two of which

are separate and composed of one unit each. Listing 3 shows what the correct

encoding should have been; note also the corrected order of the units. Figure 91

shows the hand drawing for these endleaves, and Figure 65 shows the diagram

198



generated by the correct XML file. Admittedly, this kind of problem can be

avoided through data validation.

Figure 91. Hand drawings for the right endleaves of Vol. 4725.3162.

Figure 92. Visual output of incorrect description for the right endleaves of Vol. 4725.3162.

Figure 93. Correct diagram for the right endleaves of Vol. 4725.3162.

199



<endleaves>
                 ...
        <right>
             <yes>
                    ...
                    <type>
                           <integral>
                                  <numberOfLeaves>1</numberOfLeaves>
                                  ...
                           </integral>
                    </type>
                    <type>
                           <separate>
                                  <units>
                                         <unit>
                                                <number>1</number>
                                                <components>
                                                       <component>
                                                              <number>1</number>
                                                              ...
                                                       </component>
                                                       <component>
                                                              <number>2</number>
                                                              ...
                                                       </component>
                                                </components>
                                                ...
                                         </unit>
                                  </units>
                           </separate>
                    </type>
                    <type>
                           <separate>
                                  <units>
                                         <unit>
                                                <number>1</number>
                                                <components>
                                                       <component>
                                                              <number>1</number>
                                                              ...
                                                       </component>
                                                </components>
                                                ...
                                         </unit>
                                  </units>
                           </separate>
                    </type>
                    ...
             </yes>
      </right>
</endleaves>

Listing 2. Snippet of XML description for the right endleaves of Vol. 4725.3162. Note
the third instance of the type element. Three dots indicate elision in the code.

200



<endleaves>
               ...
      <right>
             <yes>
                    ...
                    <type>
                           <integral>
                                  ...
                           </integral>
                    </type>
                    <type>
                           <separate>
                                  <units>
                                         <unit>
                                                <number>1</number>
                                                <components>
                                                       <component>
                                                              <number>1</number>
                                                              ...
                                                       </component>
                                                </components>
                                                ...
                                         </unit>
                                         <unit>
                                                <number>1</number>
                                                <components>
                                                       <component>
                                                              <number>1</number>
                                                              ...
                                                       </component>
                                                       <component>
                                                              <number>2</number>
                                                             ...
                                                       </component>
                                                </components>
                                                ...
                                         </unit>
                                  </units>
                           </separate>
                    </type>
                    ...
             </yes>
      </right>
</endleaves>

Listing 3. Corrected XML snippet description for the right endleaves of Vol. 4725.3162.
Three dots indicate elision in the code. The third instance of type element has been re-
moved, and the data has been reintegrated within the separate endleaf type, encoding it
correctly as a second unit.

201



7.2. The set of transformations

Following the detailed description of the endleaves transformation, this section

briefly describes the transformations of other parts of the schema. The complete

set of shapes are shown in Appendix B.

Each part of the binding structure has a separate intersemiotic translation, and

each transformation is independent from the others. This is important, because,

in a future practical application of this work, it would be desirable to run the

transformation algorithms during the surveying process, as each part of the

structure is being described. There are some cases where information present in

the description of some other part of the binding is necessary for an accurate

visualization. In the knowledge graph schemas reported here below, these nodes

are clearly identifiable as ‘additional information’ and are greyed out. In a future

application, this information might be called in as extra parameters to input for

the visualization process if not already described elsewhere.

7.2.1. Markers: page markers, board markers, bookmarks

The schema allows describing three kinds of markers: page markers, board

markers, and bookmarks. Page markers are those tabs attached to the edges of

leaves in a book to indicate important places in the text. Board markers, as anti-

cipated in the previous chapter, are lengths of animal skin pasted to the inside of

the board, projecting from the edge and most probably used as bookmarks. A

bookmark is a device used to mark temporarily particular leaves or passages of a

book; its formation can range from simple lengths of thread or textile ribbon to

multiple elements attached to bars or disks fixed to an endband.449

These are rather simple visualizations, as there is only a limited number of

parameters to be taken into account. The elements to be drawn are discrete and,

for the most part, can be drawn once, called when needed, and then spatially ar-

ranged on the plane according to the different visualizations needs. It is helpful

449. Ligatus 2013.

202



Figure 94. Example of pagemarker diagram for Vol. 26.4β.

Figure 95. Example of boardmarker diagram. No examples of boardmarkers were recorded
in the dataset.

203



Figure 96. Graph schema for pagemarker diagrams.

204



Figure 97. Graph schema for boardmarker diagrams.

205



Figure 98. Example of bookmarker diagram for Vol. 232.112.

206



Figure 99. Graph schema for bookmark diagrams.

207



to note how this approach follows to the letter the need for a one-to-one relation-

ship between verbal labels and graphics. This approach works sufficiently well,

when the element being described is only one, and the description only adds details

on its position in relation to a fixed series of accompanying or reference elements.

This was the case for page markers (see Figure 94) and board markers (see Figure

95). However, when it comes to more complex structures, as are bookmarks (see

Figure 95), the maintenance of the code becomes problematic. Because in this

approach, the only parameters are the coordinates of each component, changing

the graphic output of one component means also having to re-arrange spatially

all of the possible components that can be drawn with it. Whereas, by drawing

the components algorithmically, these problems can be avoided. It is indeed

simpler to write algorithms that work by calling a pre-drawn component, however,

due to the difficulties encountered in the maintenance of the code, this approach

was for the most part abandoned for the rest of the visualizations.

7.2.2. Sewing

Sewing indicates the process through which the sections or leaves of a book

are secured by means of thread, thus forming a consecutive and permanent unit.450

The basic description element of each sewing is the sewing station, i.e. the indi-

vidual structural point on the spine-fold used in the creation of the book struc-

ture.451 Each sewing structure is described as a series of sewing stations with a set

of specific characteristics. As the sewing develops from station to station, so does

the visualization. Books can be sewn and re-sewn various times, and usually this

means that multiple sets of sewing stations can be observed. For each sewing station

described, the surveyor is asked to specify whether the station is used by the current

sewing, or whether is was used in a previous or earlier sewing structure. Typically,

only stations of the current sewing can be described in detail, whilst for stations

450. Roberts & Etherington 1982/1994.
451. Ligatus 2013.

208



Figure 100. Examples of sewing diagrams in the literature ([a] Szirmai 1999, fig. 3.2, p.
34; [b] Szirmai 1999, fig. 2.3[d], p. 21 — note the visualization of the distribution of the
sewing thread lengths within the gatherings; [c] Szirmai 1996, fig. 1[c], p. 147; [d] Szirmai
1999, fig. 9.8[c], p. 188).

209



Figure 101. Example of sewing measurement diagram for Vol. 1671/1083β. Note the
visualization of both current and previous sewing stations.

210



Figure 102. Graph schema for sewing measurement diagrams.

belonging to the two other groups (previous and earlier sewings) one can generally

only mention the position on the fold and the kind of holes observed.

In the literature there are three main kinds of sewing structure visualizations:

(i) simple diagrams only showing the position of the sewing stations, usually with

their measurement (see Figure 100[a] and [b]); (ii) detailed diagrams showing

the development of the sewing thread along the gatherings, around the sewing

supports (when present), and from gathering to gathering (see Figure 100[c]);

(iii) cross-section diagrams showing the path of the thread for a single gathering

(see Figure 100[d]). Each kind of visualization has its purpose and is useful for

different reasons. Simple diagrams like (i) for example can show the distribution

of sewing stations along the spine of books and make it easier to compare similar

distributions among different volumes, more than merely through numeric values

of the measurements, as their relative position is more useful to identify patterns

than absolute values. Examples (ii) and (iii) instead, are able to present the path

of the thread, thus allowing for easier analysis and comparison between different

sewing structures. As seen in Figure 100[a] and [b], the simpler diagrams with

just the position of the sewing stations and their measurements can either show

the sewing station distribution along one ideal gathering or on the full thickness

of the volume. The two visualizations are similar in their scope, however, showing

211



the full thickness of the volume adds extra information on the appearance of the

sewing pattern for the whole book.

The diagrams here reflect these visualization conventions. There are two differ-

ent kinds of diagrams: one only takes the measurements of the sewing stations

and plots them on the diagrams of the spine of the book and fold of a gathering

(see Figure 101); the other visualizes the path of the sewing thread as it develops

across a gathering and through the spine (see Figure 104).

Simpler diagrams showing the position of the sewing stations set against the

full thickness of the volume indicate the distribution of the sewing thread lengths

within the gatherings (see Figure 100[b]), and this is a rather important feature

to visualize bypass sewing structures for example. However, the information

contained in the schema at the time of the survey did not allow for such detailed

visualizations, and bypass sewing structures are thus not visualizable. For similar

reasons, the sketchy information regarding stitched and longstitch sewing structures

leads to too many assumptions for a visualization to be possible at this stage of

development of the schema.

In the latter kind of diagrams, for those cases in which it is possible to visualize

the sewing thread, sewing direction arrows denote the conventional sewing direc-

tion depicted in the diagram. The direction of sewing — from the first to the last

gathering, or the reverse — provides important clues on the spatio-temporal origin

of the book. The sewing direction can be determined examining the sequence of

chainstitches or kettlestitches, when these are clearly visible. However this inform-

ation is not recorded in the St. Catherine’s schema. In the diagrams the sewing is

drawn as going from the bottom gathering to the top (or meeting in the middle)

as a standard option, which represents the common sewing direction from the

first gathering to the last, with the head of the volume placed towards the left.

Change-over station groups instead symbolically represent the passage of the

thread from one gathering to the other. These are blurred areas connecting the

thread paths in each gathering at the change-over station and take a verbal label

to explain their meaning. The verbal label is necessary because of the arbitrarity

and novelty of the convention. There are no similar conventions in the literature,

but this was deemed necessary to avoid drawing an otherwise naturalistic but

212



Figure 103. Examples of change-over stations ([a] Szirmai 1999, fig. 7.16[b], p. 116; [b-
c] Szirmai 1999, fig.2.1[e-f], p. 17).

random shape. This decision was twofold: firstly it is usually very difficult to be

able to determine the exact nature of the type of thread-link connecting the

gatherings together without taking the sewing apart (Figure 103 shows different

kinds of kettlestitch or change-over thread passages); secondly, in the Ligatus

schema, because of the difficulty just mentioned, there is no information on the

kind of link at the change-over station (apart from the fact that it is unsupported

or that it is a kettlestitch).

7.2.3. Boards

The wood, pasted paper, single- or multiple-ply sheets, or other material, used

for the covers of bound books to protect the leaves. Boards are usually used in

pairs and they never extend around the spine of the book.452

For an accurate visualization, the algorithm requires additional information,

such as the shape of the joints in the spine, and checks whether the endband’s

core attachment was described as sewn and recessed, indicating that a board with

squares is cut down to the height of the bookblock to form an endband slip recess

to sew the endband to the boards. This information is found in the descriptions

of other parts of the binding, but are essential for an accurate visualization of the

board.

Each board is visualized in eight different views (head, tail, fore-edge, spine,

inner surface, outer surface, and horizontal and vertical cross-sections) and its

452. Roberts & Etherington 1982/1994; Ligatus 2013.

213



Figure 104. Example of sewing measurement diagram for Vol. 452.221α.

214



Figure 105. Graph schema for sewing measurement diagrams.

215



shape is formed by a series of continuous lines, each defined by six different

parameters (edge treatment type, bevel type, fore-edge and spine corner type,

joint shape, and endband recess). These parameters dictate the shape of a specific

part of the board. The line drawing for each of these board features depends on

the shape of those contiguous to it. The one-to-one relationship between descrip-

tion and graphic nodes is preserved, but the overall visual effect is that of one

continuous shape onto which specific graphic sememes can be read in different

areas of diverse magnitude. Back-cornering,453 for example, is indicated in the

spine view as a single line across the thickness of the board, whereas, bevels can

extend across the whole height of the board in fore-edge views. This is where, as

mentioned in the literature and as described in the previous chapters, verbal and

visual languages differ substantially. This is also one of the facts that make auto-

mated intersemiotic translations a challenging process that may be automated,

but that is not simply a matter of mechanically matching verbal labels with

graphics and placing them on the plane.

7.2.4. Spine shape and lining

The schema describes the shape of the spine of the bookblock in an approximate

and categorized manner distinguishing between its degree of roundness (flat,

slight round, round, heavy round) or the shape of its joints (flat, slight, quadrant,

angled, square, acute). This allows inputting a record of the shape of the spine in

the database, and to then be able, for example, to search for different values of

roundness for bookblock spines.454

The spine lining is composed of one or a series of pieces of sheet material placed

on the spine of the assembled bookblock as reinforcement, and either adhered

to it or held in place without adhesive.455 There can be more than one lining, and

453. Back-cornering: the process of cutting away a small piece of the spine edge of a board at
head and tail to relieve the strain on the joints of the book when the covers are opened (Roberts
& Etherington 1982/1994).

454. Ligatus 2013.
455. Roberts & Etherington 1982/1994; Ligatus 2013.

216



Figure 106. Example of board diagram for Vol. 4689.3137 (left board).

217



Figure 107. Graph schema for board diagrams.

218



Figure 108. Example of spine shape diagram for Vol. 10.3ε.

Figure 109. Graph schema for spine shape diagrams.

219



Figure 110. Example of spine lining diagram for Vol. 184.97α.

220



Figure 111. Graph schema for spine lining diagrams.

221



these can be of different types. A common example is that of the combination of

transverse and panel linings: these would be described as two different linings at

selected location in the schema. However, the location description is only indicative

(the choices are all or selected) without a precise indication of which panels456 are

covered by which lining. In this case the form of the structure is not preserved in

the description as the schema fails to capture in its structure the logical form of

the specific lining arrangement. As a result a visualization based of this schema

will inevitably be inaccurate when different types of lining are combined. Usually,

in the schema structure, a series of successive elements indicates and is interpreted

as a spatial succession of those elements — i.e. different layers of spine lining, in

this case. However, in such combined linings, the described elements happen to

be at the same level, but in different positions, thus forming together one sole

composed layer. This problem would not have occurred, if the location of the

lining had been described in more absolute terms — i.e. panels 1 and 5 — or, less

ideally, if a complex type transverse & panel had been introduced, defined for

example as spine linings composed of transverse linings at head and tail and

panel lining for the rest of the panels (a common feature). As it stands, there is

no practical or immediate solution to this problem, apart from highlighting uncer-

tainty or opting for one separate visualization for each spine lining, while still

highlighting uncertainty. Because this combined transverse and panel lining was

only recorded once in the dataset (Vol. 2609/1886) and because the problem was

due to a schema limitation, the former solution was chosen, but further work

would need to be done on both the algorithm and the description schema for a

future application.

Like in the case of board diagrams, both spine shape and spine lining diagrams

(especially in cross-section view) are formed by a series of continuous lines defined

by a number of parameters. The formation of each portion of the line is informed

by a parameter, but depends in its final shape on the overall shape of the complete

line. In these visualizations, the parametric line generation algorithm is rendered

456. The term panel indicates the area between two sewing supports or between the head of
the spine and the top and bottom supports (Ligatus 2013).

222



more complicated by the presence of curved lines that need to be modified case

by case preserving their integrity.

7.2.5. Endbands

Endbands are cores sewn or pasted across the head and tail of a bookblock

spine. When sewn (see Figure 112), these are attached to the bookblock by a

primary sewing through the fold of the gatherings which loops around the core.

When pasted (stuck-on endbands, see Figure 114), these are essentially transverse

spine lining attached to the head and tail panels.457

Endbands are rather complex structures, but they can be described in their

essence through a series of typological parameters. The schema describes sewn

endbands according to the number and type of cores involved, the core attachment

type, the sewing type, and the class of number of tiedowns (in every gathering,

frequent, or infrequent). The drawback of this description-by-categories approach

is that it does not allow the description of unusual endbands. On the other hand,

the use of categorical descriptions permits to implement a highly symbolic visual-

ization system. Chapter 6 discusses the symbolic nature of endband cross-section

drawings. The same considerations can be extended to the rest of the views and

elements for these visualization. For example, the schema does not indicate the

exact number of gatherings in the bookblock,458 nor does it indicate which gath-

erings are used for the primary endband sewing. As a consequence, in the diagrams,

tiedowns are spaced evenly across the spine, in a variety of densities (every, fre-

quent, infrequent). This symbolism communicates the sewing pattern, without

necessarily indicating the exact positioning of the tiedowns.

457. Ligatus 2013.
458. As mentioned, it lacks any kind of collation formula or similar bookblock formation de-

scription.

223



Figure 112. Example of sewn endband diagram for Vol. 4282.2966.

224



Figure 113. Graph schema for sewn endband diagrams.

225



Figure 114. Example of stuck-on endband diagram for Vol. 2236.1539.

226



Figure 115. Graph schema for sewn endband diagrams.

227



7.2.6. Coverings

The term covering indicates those materials put over a book, with the effect of

protecting it. The schema distinguishes between four main kinds of coverings

(guard, drawn-on, case, and over-in-boards) indicating rather different constructions.

The two more simple types, guard (see Figure 116) and drawn-on, are not further

specified by other parameters. Case and over-in-boards covers (see Figures 118

and 120), instead, are further specified by means of a range of typologies and

parameters. Because of the diversity in features and the wide range of options,

the graph schema is presented here as divided in three related schemas: guard and

drawn-on together, over-in-board, and case covering respectively (see Figures 117,

119, and 121).

Amongst laced case structures, the schema proposes laced and tacketed bindings

as a subtype. These are not further described by additional parameters and are

defined as follows: ‘case-covers can be attached to a bookblock by both tackets

and sewing support slips and/or endband slips.’459 The definition is generic and

does not provide information on an exact typology of structure, describing parts

and form. The term used is not at the basic level of abstraction, and, therefore,

allows too much indeterminate information. Other structures are only described

with a single term and no additional parameters — e.g. cover lining laced cases460

— but these terms describe an entity at the correct semantic level, thus allowing

for it to be visualized.

Once again, these transformations depend on additional information found in

the description of other parts of the binding. In particular, the shape of the spine

459. Ligatus 2007.
460. ‘Cover lining: a laced-case cover with a cover lining consists of two pieces of sheet material,

one inside the other, which originally were attached to a bookblock one after the other and not
at the same time. The cover lining was the first to be attached, and was a piece of sheet material
such as cartonnage or laminated sheets of paper cut to the height of the bookblock and endbands,
folded around a bookblock and secured to it by lacing the sewing support slips through it. The
second part of the cover, usually of parchment was folded round the cover lining, turned-in around
its edges and secured by lacing the endband core slips through both parts of the cover at the head
and tail of each joint. The result is a binding in which only the endband core slips and not the
sewing support slips are visible on the outside of the cover, a fact which can make them externally
resemble contemporary Italian limp laced-case bindings in which also only the endband core slips
and not the sewing support slips are visible on the outside of the book. The cover lining is distin-
guished from boards in that both sides are part a single continuous piece of material wrapped
around the spine, as opposed to boards which must always be separate entities.’ (Ligatus 2007).

228



Figure 116. Example of guard covering diagram. No examples of guard or drawn-on
coverings were recorded in the dataset.

Figure 117. Graph schema for guard or drawn-on covering diagrams.

229



Figure 118. Example of over-in-board covering diagram for Vol. 3148.2285.

230



Figure 119. Graph schema for over-in-board covering diagrams.

231



Figure 120. Example of case covering diagram for Vol. 2269.1579.

232



Figure 121. Graph schema for case covering diagrams.

233



Figure 122. Example of furniture diagram for Vol. 3706.2677.

234



Figure 123. Graph schema for furniture diagrams.

235



for cross-section views, and the presence of protruding endbands. Also, the final

visualization of each of the covering part depends on the shape of adjacent visual

sememes. Note in Figure 118 the use of sketchy lines to signal uncertainty in the

visualization of turn-in and cap turn-in shapes.

7.2.7. Furniture

Furniture is the set of hardware that can be attached to a binding, usually with

a protective, but also often decorative, function; this includes fastenings used to

hold a book shut when not in use.461 As pointed out in the previous chapter, fur-

niture include: (i) articulated metal spine, (ii) bosses, (iii) corners, (iv) full covers,

(v) plates, (vi) ties, (vii) catchplates, (viii) clasps, (ix) pins, (x) straps, (xi) strap

collars, (xii) and strap plates. Items vii to xii can be grouped as fastening parts.

These are described in sufficient detail in the schema and their descriptions can

be turned into useful visualizations. Items i to vi, instead, are only named by their

generic name — e.g. boss or tie — leaving too much indeterminateness for the

generation of meaningful visualizations. Note in Figure 122 the use of sketchy

lines to signal uncertainty in the visualization of catchplate, clasp, and strap plate

decorative shapes, preserving only their functional shape (compare with Figure

124).

461. Ligatus 2013.

236



Figure 124. Fastening photographs of Vol. 3706.2677 taken during the survey in 2007.

7.3. Dealing with erroneous data

The resiliency of any system has to take into consideration human reliability

and the effects of the human factor on the desired outcome and its accuracy.462

When inaccuracy is objectively determinable, it can be expressed as error.463 Errors

are indicated by the fact that a planned sequence of activities fails to achieve the

intended outcome, without intervention of external factors. Human errors can

be broadly classified into three main categories: (i) skill-based errors or lapses

linked to attention or selection failures; (ii) ruled-based mistakes, linked to the

misapplication of rules; (iii) knowledge-based mistakes, linked to inaccurate or

incomplete mental models.464

In electronic databases, the encoding schema at the base of their structure allows

for immediate monitoring of data correctness and completeness during input.465

Data validation yields to a reduction in errors and acts as quality control. However,

462. Rouse & Rouse 1983.
463. MacEachren et al. 2005.
464. Reason 1990.
465. Mocean 2007.

237



not all mistakes can be impeded through careful database design, and those errors

that do occur are not easy to identify through automated means.

A schema can be highly prescriptive and as close to perfection as possible, but

when implemented in a project and used by different people problems do occur.

As shown through a selection of examples above, many mistakes in the dataset

for this project are due to misinterpretations of the schema rules, for example

problems in item numbering conventions for endleaf structures. In other cases,

typologies were not understood and were used inconsistently. See for example

the diagram of the spine shape for Vol. 5.2β in Figure 126. In the database its

joints are described as slight, which leads to an odd looking diagram (see Figure

126[a]) However, by looking at the hand drawing for the spine shape (Figure

125), it becomes apparent that the surveyor did not describe the joints correctly,

and that in fact, what it was meant was that the volume had quadrant joints. By

changing the joint typology and regenerating the diagram the shape coincides

with the hand drawing (Figure 126[b]).

Figure 125. Spine shape hand drawing for Vol. 5.2β.

or

Figure 126. Spine shape diagram for Vol. 5.2β. [a] with slight joints; [b] with quadrant
joints.

238



It became apparent, during the course of this project, that these types of errors

that are difficult to detect automatically could, however, be identified and avoided,

if the automated visualisations were to be implemented directly during the survey.

As seen, data can be valid but meaningless. One could foresee the surveyor or

a subsequent reader/editor going through the data in the database to check for

its correctness, one element at the time, diachronically and in sequence. However,

the information describing each binding structure is divided into a series of ele-

ments and parameters, which can span multiple description levels. The amount

of information needed to be kept in mind to visualise the data and analyse it syn-

chronically is unmanageable as it exceeds the limited capacity of the human

working memory.466 Mistakes, thus, easily slip through the control net and remain

unchecked.

During the development of the visualisation algorithms, at times, diagrams

would show structures that are not possible in real life. Some of these problems

were obviously coding problems in the algorithms that needed to be modified.

However, others were the result of something rather different: the coding algorithm

was functioning properly showing exactly what had been encoded in the XML

binding descriptions, but the dataset contained errors, and these errors were

translated into odd looking diagrams.

7.3.1. Error examples

Let us consider some examples of errors found in the dataset. These can be

divided into three main groups. As pointed out earlier, (i) there are cases in which

the surveyor misinterpreted the description rules and conventions set by the

schema, or else, (ii) typologies were not understood and were used inconsistently.

In addition to these, (iii) there are obvious slips in which one option in a list was

mistakenly chosen instead of the right one. Four examples of errors are reported

below; for each example are provided the incorrect and the correct diagram and

the hand drawing for the structure carried out during the survey.

466. Baddeley 2013.

239



Figure 127. Example 1. Error due to schema convention misinterpretation. Right
endleaves for volume 4725.3162.

Figure 128. Example 2. Error due to description convention misinterpretation. Right
endleaves for volume 5365.3471d.

Figure 129. Example 3. Error due to inaccurate encoding, possibly because of typology
misinterpretation. Spine shape and spine lining for volume 5.2β.

Figure 130. Example 4. Slip error during encoding. Right endleaves for volume 236.115.

240



In example 1 (Figure 127), as mentioned above, the surveyor did not encode the

structure correctly. The correct diagram was generated by encoding the description

appropriately – i.e. one integral endleaf and one group of separate endleaves

constituted by two units. This kind of problem can be avoided through data val-

idation by rendering the repetition of the same endleaf type within one structure

invalid.

In example 2 (Figure 128), the surveyor did not follow the component number-

ing convention. A wrong unit order leads to nonsensical diagrams. Tweaking the

data to follow the numbering convention generates a correct diagram.

Examples 3 (Figure 129), shows a spine lining diagram for which the spine

shape joint typology were inaccurately selected. This is possibly due to a misun-

derstanding of the joint categories set by the description schema. In the dataset

the joints are described as slight, which leads to an odd looking diagram. However,

the hand drawing shows that the surveyor picked the wrong joint type. The dis-

tinction between the various abstract joint types can be difficult to appreciate.

The volume had, in fact, quadrant joints. By changing the joint typology and re-

generating the diagram, the shape coincides with what had been drawn, and this

eliminates the gap between the bookblock and the lining at the joints.

Example 4 (Figure 130), shows an obvious slip in the selection of the endleaf

component type. The XML description indicates the endleaf as a guard . However,

the hand drawing clearly shows a fold type.

Table 2 categorises the examples according to their cause and states whether

they could have been avoided with effective data validation. Only example 1 was

avoidable through validation routines. As discussed in chapter 5, data that is not

‘valid in the first sense’ — i.e. determined by the rules of its language — can be

avoided through validation routines. Ambiguities due to human error that cause

‘invalidity in the second sense’ — i.e. determined by context — are, instead, not

avoidable through data validation, and can only be checked against reality. For

each example, the table shows whether it is not ‘valid in the first sense’ or ‘the

second’. Examples 1, 2 and 3 are not ‘valid in the first sense’. Their invalidity is

ruled (and identifiable) by the grammar of the schema (example 1, which makes

it possible to validate this error through automated routines), and by the fact that

241



Validity sense ruled byValidity sense
(I or II)

Avoidable through
data validation

Error causeExample #

schema & object formIYesschema convention
misinterpretation

1

object formINodescr. convention
misinterpretation

2

object formINoinaccurate encoding3

object contextIINoslip during encoding4

Table 2. Table showing the cause for each error example, whether it would be avoidable
through effective data validation, whether it is invalid in the first or second sense and the
language or context ruling its invalidity.

the diagrams do not show a configuration that would be possible in real life. Ex-

ample 4, instead, is not ‘valid in the second sense’ and its validity is therefore

context related: what the diagram shows is possible, but not true.

This thesis proposes the use of visual means to solve these validation problems.

Diagrams, as visual communication systems, naturally offer information in a syn-

chronic manner and could immediately highlight errors.467 The next section intro-

duces how this could be possible.

7.3.2. The feedback loop

In order to explain the value of diagrams as validation tools, it is important to

discuss the communication cycle diagram introduced in chapter 1 (reproduced

here in Figure 131). One could notice a feedback loop that connected the human

observer back to the object that started the communication cycle. The concept

of a communication feedback is a common feature in communication theories.468

467. Campagnolo 2014a.
468. Rothwell 2012.

242



Figure 131. Communication cycle for the project, highlighting the feedback loop between
the human observer at the end of the cycle and the object being described and visualized
at its beginning.

In this project, there is an extra step between the coding of the information

and its delivery, as this is re-coded by the computer into a visual message. The

feedback loop in this particular communication cycle can then take on a different

role and meaning. The feedback could in theory be connecting not a different

person to which information about the object is being communicated, but the

person who encoded the information in the first place. This was obviously not

the case for the St. Catherine’s survey. However, it is possible to foresee a system

that integrates the kind of automated transformations described in this research

directly within the surveying process. In such a case, the feedback loop would be

linking the observer with the object being described, and the observer would be

receiving the same information that was input into the system, but in a different

form. A form that being visual is more immediate and synchronic.

Although it has not been possible to test the feedback loop exploitation in a

large scale. All of the errors listed above — and many more — have been identified

thanks to the automated diagrams. It would be, however, interesting to test the

efficacy of such a system in a project and check whether the reliability of the human

surveyor does indeed increase and how much does this affect the speed of the

surveying process.

It is likely that, had this system been in place for the 2007 survey, some of the

mistakes registered in the dataset would not have occurred. The feedback loop

would have acted as a proof of truthfulness against reality for the descriptions.

243



Let us say that p is the structure to be recorded and that, for any reason, non-p is

instead input in the database. If non-p is allowed by the schema, the incorrect in-

formation will not be captured by data validation and remain in the dataset. If,

through an extra step, non-p is presented again to the encoder, it is probable that

non-p would be corrected into p and the mistake would be amended.

This system would not circumvent all kinds of errors. If an error is a knowledge-

based mistake in the interpretation of the evidence the feedback to reality would

probably still afford the same incorrect interpretation. However, based on the

empirical experience accrued during this work, it would seem probable that most

of the skill-based lapses and ruled-based mistakes presented here would be

avoidable with the application of this system to the surveying process.

Summary

This chapter presented a detailed analysis of the transformation for endleaf

structures. The rest of the other transformations have also been introduced. Some

noteworthy mistakes and problems in the schema and the dataset have also been

highlighted.

Any system has to take into consideration the human factor, and the errors that

are likely to make their way in the dataset because of it. Empirically, it seems that

the implementation of a system like the one presented in this research that is able

to take the data that has been recorded, and present it again to the surveyor in a

different form, might lead to less erroneous data in the dataset.

The following chapter draws from the thesis a series of methodological consid-

erations for automated intersemiotic translations.

244



Chapter 8. Recommendations

Und so verhält es sich in der Philosophie überhaupt: Das Einzel-

ne erweist sich immer wieder als unwichtig, aber die Möglichkeit

jedes Ein-zelnen gibt uns einen Aufschluss über das Wesen der

Welt.

And that is generally so in philosophy: again and again the indi-

vidual case turns out to be unimportant, but the possibility of

each individual case discloses something about the essence of

the world.*

Ludwig Wittgenstein, Tractatus logico-philosophhicus, 1922.

*translation by Pears & McGuinnes (1961).

Before drawing the final conclusions of the thesis, this chapter discusses the

transformation process, and considers the methodology for successful automated

intersemiotic translations based on models of material objects.

8.1. Intersemiotic translations: methodological recommendations

At the beginning of the project it was argued whether it would be possible to

gather enough information for successful automated visualizations from a model

of a material object, as opposed to produce a drawing interface for end-users

based on semantic models for.

From the analysis of the issues and considerations mentioned throughout this

dissertation, it is now possible to advance a series of methodological recommend-

245



ations that should be followed in the design of systems for the automated interse-

miotic translation and diagrammatic visualization of material objects based on

models of.

8.1.1. Intersemiotic translation elements

In intersemiotic transformations, there has to be a one-to-one relationship
between the components of the verbal propositions and those of the visual pro-
positions.

Translation from one language into any another language is attained by trans-

lating all components of the proposition a in language A into the components of

a proposition b in language B. Each component in proposition a is represented

by a sign, whose meaning is common to the sign that substitutes it in the propos-

ition b. In this project, a proposition a is encoded verbally and is translated into

a proposition b encoded graphically. It follows that each component in proposition

a finds its counterpart in proposition b, i.e. for any element described verbally,

something should be drawn.

8.1.2. Functional verbal descriptions

For an automated intersemiotic translation system to be functional, the verbal

information has to be structured. Structured descriptions can convey information

on material components of material objects, and their form, maintaining a refer-

ence to the logical form.

This is achieved through: controlled vocabularies, the hierarchical organization

of information, and the schema of the description.

Amongst the different kinds of verbal descriptions of bookbinding structures,

controlled vocabularies and structured descriptions have been found to be the

strategies most successful at conveying the minimum information.

246



The selection of precise terms and definitions, even if not referring to atomic

elements, makes them capable of conveying information on both the material

components and the form of the element being described. However, the term

should refer to the basic level of abstraction of the category that it describes, or

else indeterminateness would render its use as reference point unsuitable for

visualization purposes. In the St. Catherine’s schema there are examples of con-

trolled vocabulary terms that fail to define their category at the basic level of ab-

straction. Examples that have already been highlighted in the previous chapters

are those used for furniture items other than fastenings — e.g. bosses and ties —

or laced and tacketed case covers, as opposed to strap plates or strap collars and

cover lining case covers. The elements defined within controlled vocabularies are

prototypical examples of their categories, and, if used at the right semantic entry

point level, they maintain a reference to the logical form of the elements described.

The definitions convey information about their form and parts, and also, some-

times, about their relative position in relation to other elements. Think, for ex-

ample, of the definition of an endband’s crowing core: ‘a subsidiary core which

sits on top of the main core’.469 Here, the definition provides relative spatial in-

formation. This information is obviously domain specific. The technical vocabulary

of classical architecture is extremely well developed in these respects.470 For a

useful description, these defined elements, even if non-atomic, should still refer

to smaller parts of the whole being described, or else, if not further described

through other smaller elements of which they are composed, they risk being too

general and indeterminate.

In structured descriptions, each element is in hierarchical relationships with

the others, and its position within the hierarchy can carry information in regards

to its position in space in relation to the others. For example, the arrangement of

various endleaf components can be placed spatially based on the hierarchical de-

scription of the structure. Let us consider again the example in the previous

chapter and reported here in Figure 132. In this case the item numbering conven-

tion (from outside to the inside) was not followed. It was the incorrect hierarchy

469. Ligatus 2013.
470. Goulette 1999; Borillo & Goulette 2006.

247



caused by the erroneous encoding that produced an impossible diagram, and not

the incorrect use of some components. In other words, in the description, the

object’s components were all correct and present, but the way they had been en-

coded — i.e. their place in the description hierarchy — resulted in an impossible

diagram. Correcting the hierarchy yields a correct diagram.

Behind the hierarchical structure of the descriptions lies the model of the in-

formation to be gathered, and this model is encoded within a schema. The schema,

acting like a recipe for a particular binding component, is the expression of the

logical form of that component. Provided that the model behind the schema is

well developed, complying with the schema, a structured description is capable

of portraying any instance of the binding component.

The schema guides the human user in describing an object in accordance with

its model of. Following the schema, it is possible to algorithmically extract the

relevant information from the descriptions, and transform this into a graphic

representation. The transformation algorithms act as models for the generation

of the diagrams. The algorithms can be compiled successfully in virtue of the

schema — i.e. the model of the material object — acting as a reference to the lo-

gical form of the object. Without this, as it was the case of natural language pro-

cessing projects outlined in chapter 2, any sure link to the logical form of the object

is lost, leading to indeterminateness, especially for spatial information.

8.1.3. The nature of sememes in diagrams

In diagrams, the shape of each element depends on that of the adjacent units,

sememes are interdependent, and gaps have to be avoided, as informational la-

cunae are read as absence of phenomena, and not as missing data.

Unlike verbal propositions, the units of images are seldom discreet and with

clear boundaries. They can be generated in accordance to the one-to-one relation-

ship with the elements of the verbal description, but their final shape will generally

depend on that of the adjacent units. Working in a combinatorial way, element

by element, the whole image text can be parametrically constructed.

248



Figure 132. Errors in the hierarchical arrangement of the elements result in wrong spatial
relationships. Right endleaves for volume 5365.3471d.

Diagrams, by a semiotic point view, are visual texts, whose sememes are woven

together and are interdependent in relation to their meaning. This has important

consequences for the generation of automated intersemiotic translations, as the

generation of each visual sememe necessarily has to depend on the other sememes

present in the current diagram, and on the overall visual text.

In visualizations, one should not leave or form gaps of graphic information, or

the risk is to alter the meaning, or to render the whole visualization meaningless.

The plane on which visualizations develop is continuous as it can be infinitely

subdivided. One can distinguish between signifying and non-signifying parts of

the plane. It is difficult to disregard a part of the signifying plane, as zero signs are

read as absence of phenomena, not as missing data. Since in visual texts, sememes

are interdependent in their meaning, the absence of one part can alter the meaning

of the other parts and of the whole proposition.

8.1.4. Prototypicality of information

Words convey prototypical information. Visualizations derived from verbal de-

scriptions are necessarily prototypical in nature. Visual percepts are recognized

and labelled according to a clear-cut entity reference framework tending towards

best (regular) examples. Prototypical representations do not convey the specificity

249



of an object. The prototypicality of visualizations is however not a shortcoming

as this facilitates priming and memorization.

Material objects are recognized as part of their category through reference to

best example prototypes. Typically, these prototypical instances represent a par-

ticular level of abstraction. It is at this level that one mostly interacts with objects,

gives them labels and names. In addition, it is at this level that the parts which

constitute objects are also identifiable through simple verbal labels. Thanks to

this, at this level of abstraction it is possible to form a mental image that will re-

semble the appearance of all the members of their category. Any automated

translation of information conveyed through verbal labels will necessarily be

prototypical.

Prototypes are inclusive in nature, and specific only in regards to some funda-

mental features of the category they represent. This leads to a certain level of in-

determinateness as the specificity of an object cannot be portrayed through pro-

totypical information alone. The human mind classifies shape percepts as instances

within a categorical system in which only some features are perceptually funda-

mental. Percepts are recognized and labelled according to a clear-cut entity refer-

ence framework. Verbal descriptions, making use of these clear-cut verbal labels,

lack therefore the ability to deliver the specificity, the precise appearance and

shape of objects.

When dealing with visual perceptions, perceived shapes are matched up with

the prototypical mental images of their category. Prototypical visualizations, offer-

ing surrogate stimuli, can be used as instruments of abstract visual reasoning.

Mental representations of visual information possess structural and spatial char-

acteristics. Every part of an object and its spatial relations are encoded in mental

representations, and this information, which is not usually available to language-

base thinking, is, however, readily available for thought processing.

The prototypicality of visualizations is not a shortcoming as their standardized

and essential appearance augments their immediacy and facilitates the comparison

with other exemplars. In fact, it induces visual priming,471 as similar patterns

471. Priming: an implicit memory effect in which exposure to one stimulus influences a response
to another stimulus (Ware 2013). See Appendix A.

250



between different visualizations remain similar, whilst only patterns that signal

different information change. Priming allows comparing, identify differences, and

remember similar images even if the information is not consciously perceived.

8.1.5. Graphic prototypes

Diagrammatic representations can be generated from verbal descriptions, if the

minimum information is conveyed through the process. Graphic prototypes do

not need to be naturalistic drawings, but can be stripped down to their essence

and schematized. This is true also if they are communicated through iconic signs.

When presented as two-dimensional, patter recognition and comparison is facil-

itated.

In a communication process, information is conveyed through the cycle, from

the sender/encoder of the message, to the receiver/decoder. In the specific case,

one starts with the encoding of information regarding bookbinding structures by

part of a human interpreter within a verbal description. Then, the information is

parsed by an algorithm, which can transform it into a series of diagrams. At each

stage of the chain, the minimum information has to be preserved for the automated

diagram generation to be possible. Minimum information is information on the

material components of the object described, its form, and a reference to its logical

form. This reference to the logical form can in turn be used to inform the trans-

formation algorithm design.

A sign signifies its object only in virtue of certain features possessed by it (or

its perception) as the result of a process of selective abstraction. A sign — even

an iconic sign — can be generated through an abstraction and generalization

process.

Three-dimensional representations can make it harder to compare features

between similar items as human pattern perception resources are for the most

part devoted to planar information as opposed to depth, and shapes are distorted

when presented in three dimensions on the plane.

251



8.1.6. Fixing the meaning of visual sememes

In visual texts, the meaning of sememes can be fixed with the use of additional

information in the form of reference elements and verbal labels.

The natural tendency to polisemy of visual signs can be fixed through the integ-

ration of verbal labels, visual information within the same diagram, and the pres-

ence of graphic reference elements. These elements allow for correct reading of

the information in beta modality, i.e. to access information other than just the

immediate visual appearance of the signs.

8.1.7. Uncertainty

For scholarly research, both uncertain data and uncertainty that is the result of

the visualization process need to be flagged.

Not highlighting what is uncertain within a visualization is to express complete

conviction and absence of doubt. This would not be regarded as good practice

in any scholarly research, and yet, often uncertain information is not appropriately

flagged within visualizations. Within scholarly research projects visual information

should convey uncertainty where needed. The visualization process should allow

the recognition of what data is uncertain, and which parts are uncertain as the

result of the visualization process itself.

8.2. Recommendations for the communication of bookbinding
structures

All of the considerations presented in this chapter are applicable to any mater-

ial object and are not specific to bookbinding structures. However, considering

that this project dealt with the application of these principles to the field of

bookbinding studies, and considering the peculiarities of the field as presented

252



in chapter 4 — its lack of standards in particular — it is useful to draw some

general conclusions here.

As seen in chapters 4 and 5, unstructured verbal descriptions of bookbinding

structures, which are unaccompanied by graphic representations, lack the ability

to convey enough information for a reader to derive one and only one possible

interpretation of what was described, i.e. they are under-specified and under-de-

veloped, as these are not capable of successfully conveying the form of binding

structures. Whilst the precise canon of rules of composition of classical architecture

leads to the efficacy of its controlled vocabulary as a means of conveying also

spatial information, there is no such canon for the composition on the elements

of bookbinding structures. Structured descriptions coupled with a precise control

vocabulary allow to overcome this problem, as, by offering information in an or-

derly and prescribed manner, they provide a more complete representation of the

logical form and its inherent possibility to indicate any form. Therefore, structured

descriptions of bookbinding structures are capable of conveying the core pieces

of information for their successful communication, and, in turn, the production

of successful intersemiotic translations.

In a similar manner, the lack of standards in the graphic representation of

bookbinding structures can hinder communicability and comparative analysis

between similar structures in different books recorded by different scholars. This

problem can be overcome with the generation of graphic prototypes, i.e. exemplars

which select only the minimum amount of information needed for successful

communication of bookbinding structures. In addition to this, standardized

graphic conventions allow the integration of uncertainty visualization within

bookbinding diagrams, thus fostering scholarly transparency also within graphic

representations, and not just within textual communication.

The recommendations outlined throughout this thesis, and summarized in this

chapter, allow better communicability and shareability of information on book-

binding structures, both through verbal structured descriptions, and through

standardized graphic representations. Even if the main purpose of this project

was the evaluation of the possibility of automatically generating bookbinding

structure diagrams from verbally encoded descriptions, the analysis of the descrip-

253



tion problems encountered in the course of the project is a useful example to

follow to describe bookbinding structures in general, as successful intersemiotic

translations need, in the first place, successful descriptive models.

Szirmai472 correctly identifies the lack of a systematic vocabulary, terminological

clarity, and a precise recording system as the main problems encountered when

scholars attempt to describe bookbinding structures, and this inevitably hinders

progress in the field. This research, on the one hand, provides a reference frame-

work to develop efficient vocabulary and description systems applicable to

bookbinding structures, and on the other offers guidelines for the standardization

of graphic representations of bookbinding structures, with the inclusion of uncer-

tainty in the data and in the visualizations.

Summary

Drawing from what exposed in the course of this thesis it has been possible to

put together a series of methodological recommendations to take into consideration

for a system attempting to perform intersemiotic translations of the kind described

in this project. These consider the main steps of the visualization and communic-

ation process. They consider which verbal description is best suited to convey the

minimum information for successful visualization, the prototypicality of verbal

and graphic information, the nature of the meaning bearing parts of visual lan-

guages, the relationships between the elements in the two languages involved in

the intersemiotic translation process, and the presence of uncertainty in the data

and in the visualizations.

The next chapter draws the final conclusion for the thesis and states the contri-

butions that this project has made.

472. Szirmai 1999.

254



Chapter 9. Conclusions

The pattern of the thing precedes the thing.

Vladimir Nabokov, The Art of Fiction No. 40, The Paris Review

#41, 1967.

The novel contributions of this study include:

• methodology recommendations for successful automated intersemiotic

translations,

• the production of transformations for the domain of bookbinding

studies,

• design recommendations for the generation of standardized automated

prototypical drawings of bookbinding structures,

• the application of uncertainty visualization to the field of the

archaeology of the book.

This project investigated the transformation of verbal information, which de-

scribed material objects such as bookbinding structures, into visual representations

as part of an automated process. Intersemiotic translations pose a number of issues,

and practical considerations that have been described in the course of this thesis.

The following sections will summarize the principal points covered and state what

contributions were made during this project.

255



9.1. Successful automated intersemiotic translations

This research posed a question about the kind of information needed for a

successful visualization of bookbinding structures, or other material objects. The

literature review did not highlight any previous work on the matter. It is now

possible to identify this core information. A project that attempts a passage of

information about material objects from verbal descriptions to graphic represent-

ations should consider at first if these core pieces of information are available.

This is a primary contribution of this research project.

[C1.0] Transforming verbal descriptions to visual representations automatically

is possible, if the following core pieces of information are available:

1 whole/part relationships: the verbal description conveys information about

the material components of the object.473

2 spatial configuration of parts: the verbal description conveys information on

the form of the object.474

3 rules of composition of the object: the verbal description maintains a refer-

ence to the logical form of the object being described, and this information

can inform the transformation algorithm design.475

4 prototypical shape: it is possible to generate a prototypical shape from verbal

descriptions.476

Chapter 4 presented different types of description of historical bookbinding

structures. All of these were able to enumerate and name all the relevant material

components of the object being described, but most were not capable of commu-

nicating its form successfully through strictly verbal means, i.e. the spatiality of

the elements. This creates problems in subsequent readings and analysis. The re-

473. §3.4.1, §3.4.1.1, §4.2.1-4, §5.1, §5.1.1, §5.1.1.1-2, §5.1.2, §5.1.4, §6.2, §6.2.3.1, §7.2.6,
§8.1.2, §8.1.5. The numbers indicate the sections within the text where these points have been
described and argued.

474. §3.4.1, §3.4.1.1, §4.2.1-4, §5.1, §5.1.1, §5.1.1.1, §5.1.1.2, §5.1.2, §5.1.4, §6.2, §6.2.3.1,
§7.1.4, §7.2.4, §7.2.6, §8.1.2, §8.1.5.

475. §3.4.1.2, §4.2.3-4, §5.1, §5.1.1, §5.1.1.1-2, §5.1.2-4, §7.2.4, §8.1.2, §8.1.5.
476. §3.3.1-2, §3.4.2, §3.4.2.1-4, §4.3, §4.3.1.2, §4.4.3, §5.1.1, §8.1.4.

256



lationship between a whole and its parts, and the configuration of the parts are

all essential elements for the communication of the essence of an object. Structured

descriptions coupled with controlled vocabularies are capable of overcoming

these limitations of verbal communication.

For a successful automated translation from verbal to visual descriptions of

material objects, the capacity of conveying the form of objects is not in itself suf-

ficient. One needs to be able to refer to the ‘recipe’ of that particular object, i.e.

its logical form. Through this reference to the logical form, both kinds of propos-

itions, verbal and graphic, adhere to the same model of the material object being

described, and the automated intersemiotic translation can take place.

Percepts are recognized and labelled as tending towards clear-cut entities.

Verbal descriptions lack the ability to deliver the specificity, the precise appearance

and shape, of objects. However, this labelling system allows the successful com-

munication of general information about the visual aspects of reality. Labels at

the basic level of abstraction allow a mental image to be formed that resembles

the appearance of members of the class as a whole. When provided with labels

at the correct level of abstraction it is possible to establish and draw a prototypical

shape for objects and their components.

Based on the analysis of the issues and considerations mentioned throughout

this thesis, this work has advanced a series of methodological recommendations,

laid out in chapter 8, that should be followed in the design of systems for the

automated intersemiotic translation and diagrammatic visualization of material

objects, and of historical bookbinding structures in particular. These methodolo-

gical recommendations constitute another important contribution of this project.

9.1.1. Summary of the recommendations

Automated intersemiotic translations require that information on both the

material components of material objects, and their form may be verbally encoded.

It is also necessary that a reference to the logical form of the material object is

257



maintained and that this is available for the compilation of the transformation al-

gorithms. Structured descriptions and their schemas are capable of satisfying these

requirements and can therefore be recommended as good and feasible examples

of verbal descriptions.

In designing the transformation algorithms, one should keep a one-to-one rela-

tionship between verbal and graphic elements. It should also be noted that the

shape of the elements to be drawn is likely to depend on that of the adjacent and

concurrent elements within the visual text being generated. Care should be taken

not to leave or form gaps of graphic information in the diagrams.

The diagrams generated are necessarily prototypical in their design and shapes.

This permits the standardization of the output diagrams, and subsequent ease of

comparison between different exemplars. Prototypical shapes can be generated

from verbal labels through a regularization and generalization process.

The meaning of diagrams and their sememes can be fixed with the use of

graphic reference elements and verbal labels. Uncertainties in the data and in the

visualization have to be flagged through visual cues. These cues need to be applic-

able without disturbing the perception and identification of the diagram shapes,

and their connection with uncertainty should be easily interpreted.

9.2. Visualizations from models of

[C2.0] It is possible to draw the minimum information for successful automated

visualizations from models of a material object.477

In chapter 2, the possibility was advanced of being able to transform automat-

ically information contained in models of an object into graphic representations.

In the cultural heritage field, it has been customary to implement semantic mod-

elling interfaces that are based on models specifically designed for the task. Also,

often, these interfaces integrate direct acquisition data into the modelling process.

477. §2.1, §2.1.1, §2.2.3, §2.3, §8.1, §8.1.4.

258



These allow an artefact to be modelled without having to deal directly with geo-

metric primitives such as lines and planes.

From the experience accrued during this research project it is possible to con-

clude that models of are suitable for the generation of prototypical diagrammatic

visualizations. A model of approach cannot be regarded as a substitute for the

customary models for if the desired outcome has to reflect in detail the specificity

of an artefact. However, prototypical graphic representations, which can be gen-

erated from models of, offer advantages when the desired outcome is to be com-

pared between many instances of a class of artefacts: their prototypicality offer

standardized and essential appearance, augments their immediacy, and facilitates

the comparison with other exemplars by inducing visual priming.

9.3. Standardized automated drawings of bookbinding structures

As seen in chapter 4, there is no standardized way of depicting bookbinding

structures in drawings. This project has brought together a series of schematic

depictions of bookbinding structures that, following common design principles,

and being automated, are by nature uniform in their appearance. These design

considerations are applicable to other material objects, but considering the lack

of any standard in drawings within the field of the archaeology of the book, these

are a particularly important contribution to this research area, and not just for

automated drawings and their particular formation constraints. Following these

guidelines, illustrations of bookbinding structures depict a selection of attributes

that are of academic interest to the scholar or the conservator, while many details

not considered relevant to the research are omitted, and they can facilitate com-

parative analyses with other studies and structures, thus fostering communicability

of information regarding bookbinding structures.

259



[C3.0] Bookbinding structures are efficiently represented by prototypical line

drawings.478

Amongst the various types of drawings of bookbinding structures found in the

literature, line drawings are able to convey the minimum required amount of in-

formation, but are simple enough to be readily interpreted, understood, and re-

membered. Also, as iconic signs, they do not need to be naturalistic.

[C3.0.1] Two dimensional line drawings are sufficient for shape recognition as

they are capable of capturing the essential shape and form of an object without

the need for colour information. They can be used as iconic signs resulting from

the abstraction of shape perception information 479

There are many perceptual characteristics possessed by objects, e.g. colour,

texture, shading, etc. Amongst the various visual properties, shape is the most

prominent for unambiguous identification. Figure-ground organization is

achievable even if only the outline of the objects is preserved, and the visual in-

formation is simplified and reduced to the bare minimum. Black and white planar

line drawings are therefore sufficient for shape and pattern recognition and can

be used as iconic signs resulting from the abstraction of shape perception inform-

ation. As human pattern perception resources are for the most part devoted to

planar information as opposed to depth, for easy of feature comparison two-di-

mensional representations are to be preferred.

[C3.1] Graphic prototypes of bookbinding structures can be generated as the

result of the abstraction by simplification, regularization, symmetrization, and

preservation of other prägnant features.480

Shapes are perceived visually as tending towards more regular shapes to which

they are seen as similar or equivalent, and these more regular shapes, as seen in

C4.1, are labelled for categorical organization. Shape prototypification is a process

478. §4.4.3, §6.2.3, §8.1.5.
479. §3.4.2.3, §5.2.3.4, §8.1.5.
480. §3.4.2.2-4, §8.1.5.

260



of simplification, regularization, symmetrization, and preservation of other prägnant

features that are essential for the recognition and labelling of percepts.

[C3.2] Graphic prototypes of bookbinding structures have to preserve and

communicate their objects in a psychologically sound manner, taking into con-

sideration the mental representation capacity of the end users.481

Psychologically, there are systematic differences between perceptual displays

and their mental representations. Effective graphic representations should ensure

that the content and form of the visualization correspond to those of the desired

mental representation and that these are readily and correctly perceived and un-

derstood. For this, to preserve and communicate in a psychologically sound

manner the composition of the object represented along with its form, and the

spatial relations amongst its material components, one may need to rely on exag-

geration of some features and accurate but imprecise spatial relations, separating

information into discernible visual groups.

[C3.3] Graphic prototypes of bookbinding structures are mostly ruled by ratio

difficilis: their generation needs both to follow those few conventions established

in the literature, and to balance naturalism of shapes and symbolism. The shapes

selected need to be consistent and need to avoid creating confusion with other

shapes.482

Drawings of bookbinding structures have never been formalized, and their

appearance tends more towards iconism than symbolism. For these reasons, these

drawings are mostly ruled by ratio difficilis. As a consequence, one needs to follow

what few conventions can be seen as established in the literature — if indeed these

comply with the general design principles proposed here — as these, allowing

some degree of ratio facilis, render the diagrams easier to remember and reproduce

for the initiated. For diagrams, or parts of diagrams, that instead lack conventions,

the shapes need to tend towards iconism, i.e. generated and presenting themselves

481. §3.4.2.4, §4.5, §5.2.3.2, §6.2.3.3, §8.1.5.
482. §3.1.2.2, §4.3.1.1-2, §6.2.3, §6.2.3.2.

261



in accordance with their own content. In any case, within the whole visualization

system, shapes should be consistent, and care should be taken not to create con-

fusion mixing symbolic and iconic shapes that end up looking similar but with

substantial different meaning, e.g. symbolic square outlines for gathering pages

and iconic square outlines (see Figure 19[c]).

[C3.4] Each drawing needs to show only the minimum amount of information,

at the most useful level of detail.483

Visualizations need to draw the part of the bookbinding structure that they

represent showing only what is needed, thus not overwhelming the eye with irrel-

evant information. For example, endband views do not need to show the whole

book, but only the head or tail sections.

9.3.1. Summary of design principles

Prototypical information on bookbinding structures can be successfully com-

municated graphically through monochrome and bi-dimensional line drawings.

Being prototypical, the shapes can be generated through a prototypification process

that preserves those prägnant features that are essential for the recognition and

labelling of the objects. Information should be presented in a psychologically

sound manner: some features can therefore be exaggerated and kept separate,

and each drawing should show the most significant level of detail. Whilst the lack

of standardized symbolism means that drawings need to tend towards iconism,

any appropriate convention should be followed, but the resulting shapes should

be consistent and clearly identifiable.

483. §4.4.4, §4.5.1, §6.2.3.

262



9.4. Uncertainty visualization in bookbinding studies

[C4.0] Uncertain data need to be flagged within visualizations of historical

bookbinding structures for scholarly research.484

As noted, the field of the archaeology of the book is young, and still developing.

Certain information is scarce, and often the interpretation of bookbinding struc-

tures leaves significant space for educated guesswork and uncertainty. Nonetheless,

nowhere in the literature, one finds cases of visualizations that show what inform-

ation is certain, and what is instead a guess or completely uncertain. The problem

of the visualization of uncertainty within drawings of historical bookbinding

structures has never been addressed in a scholarly way, and this poses serious

problems of interpretation when reviewing literature sources.

The generation of the diagrams for this project posed many issues with uncertain

data, and for the first time, uncertainty has been flagged within visualizations of

historical bookbinding structures. The integration of a systematic way to show

what information is uncertain within bookbinding structure diagrams constitutes

another contribution to the field made by this project; a contribution that alone,

if followed in future scholarly work, could represent a significant improvement

and advancement for the discipline.

[C4.1] Both uncertain data, and data that acquires uncertainty in its transform-

ation to the visual need to be flagged within visualizations for scholarly research.485

Visualizations within scholarly research projects should convey uncertainty

where needed. The visualization process should allow recognition of what data

is uncertain, and which parts are uncertain as the result of the visualization process

itself.

[C4.2] Visual cues which are used to flag uncertainties within diagrams of ma-

terial objects need:486

484. §5.3, §6.2.3.5-7, §8.1.7.
485. §5.3, §6.2.3.5-7, §8.1.7.
486. §5.3.2.2, §5.3.2.2.1, §6.2.3.5-7, §8.1.7.

263



• to be applicable without changing the overall shape of the object,

• to avoid visual dimensions that have already been assigned other values

— e.g. dotted lines,

• to avoid triggering unwanted Gestalt laws because of the interplay of

particular uncertainty encodings — e.g. changing the figure/ground

balance — and shifting attention.

Depicting uncertainty in diagrams whose shapes are dictated by those of the

objects being represented, as in the case of material objects, poses particular issues

and constraints. Visual cues need to be carefully selected to avoid adding extra

coding to visual dimensions that are traditionally already used with different

meanings, or that have already been assigned other values within the same visual-

ization. The integrity of the overall perception of the shapes need not be precluded.

[C4.3] Feasible visual cues for visualizating automatically uncertainty of material

objects are: blurring, grey scale variation, transparency, fading, sketchy lines,

and halos.487

Various variables have been used to depict uncertainty; of these, blurring, grey

scale variation, transparency, fading, sketchy lines, and the addition of a halo are

those better suited to be applied to material object visualizations.

9.4.1. Summary of uncertainty visualization application

In bookbinding studies, despite the fact that uncertainty is often inherent in

the data collected from the artefacts, drawings and visualizations have never

highlighted which information is uncertain. The visualization process should in-

stead allow to recognize the parts that are uncertain. Uncertainty should be flagged

without altering the overall meaning of the visualization.

487. §5.3.2.2.1, §6.2.3.5-7.

264



9.5. Added value & benefits of visualization automation

The generation of automated visualizations offers a series of benefits of signific-

ant added value. Each of these points shows how the implementation of a system

capable of automatically visualizing information contained within structured de-

scriptions can be beneficial. Therefore, the recommendations outlined in this

project are in themselves a contribution, as, if followed for the successful generation

of automated visualizations, they can advance and ameliorate similar data-gathering

projects on material objects.

[C5.0] Automated visualizations, if integrated with the description process, can

work as a visual accuracy control system helping to identify meaningless or wrong

data immediately, thus increasing accuracy of data within a database.488

Data accuracy is an essential element for any database, but automated data

validation systems cannot avoid all kinds of errors. If one were to integrate the

generation of automated visualizations with the description process, the diagrams

thus generated can help to immediately identify meaningless or incorrect data.

They can function as a visual control of the accuracy of the data within a database,

and this can eventually lead to better data being stored within the dataset.

[C6.0] The automated nature of the visualizations, combined with the parametric

and combinatorial character of their constitutive elements, renders the high

number of possible permutations irrelevant.489

The structured description schema leads the surveyor down description paths

through choices and selections, and even for a limited number of options, this

leads to a large number of possible visualizations. The drawing algorithm, following

the same description paths, can generate suitable visualizations choosing, selecting,

and parametrically generating the necessary elements, thus rendering the number

of permutations irrelevant.

488. §5.3.3, §7.3.
489. §6.2.2.

265



[C7.0] Information within structured descriptions is fragmented and dispersed

within the database. The synchronic nature of visualizations reintegrates the in-

formation and helps in its analysis, by grouping it and freeing working memory.490

Structured descriptions by nature fragment and disperse information within

databases, or to present information in a highly condensed manner, as in the case

of collation formulas discussed below. Human working memory limitations make

it difficult to reintegrate or to reinterpret large amounts of information within the

mind. If this information is instead presented within diagrams, the synchronic

nature of visualizations naturally reintegrates or expands scattered or condensed

information. In this manner, the diagrams, acting as exograms, are capable of

helping with the grouping of the information and of freeing working memory,

thus allowing easier analysis.

[C8.0] Automated diagram generation can free survey time, increasing productiv-

ity.491

During the St Catherine’s Library printed book survey, as well as in other sur-

veys, the surveyors both recorded verbal information within the database and

sketched some structures on the survey form. The majority of these drawings are

prototypical and schematic in nature. It follows that providing the surveyor with

a system capable of generating prototypical schematic drawings automatically

would increase production speed and be beneficial even just in this respect.

[C9.0] Thinking about automated intersemiotic translations can inform schema

design.

The fact that the schema behind the dataset for this project was essentially a

work in progress did not preclude the usefulness of this exercise. The schema’s

limitations, in fact, fed into the practical considerations along with all other

problems met in the course of the XSLT code writing and the SVG generation.

Furthermore, whilst the focus of the recommendations outlined in this research

490. §1.2.2, §4.2.4, §7.3.
491. §1.2.4, §1.3, §6.1.4.1.2-3.

266



was on the generation of the automated visualizations, the analysis of the descrip-

tion problems encountered in the course of the project — e.g. too general semantic

entry point for binding elements, such as ties and other furniture elements

(§6.2.3.1) — can prove helpful in the designing of new schemas to describe ma-

terial objects, even if the generation of automated visualization is not part of the

project. In fact, to be able to accomplish a successful intersemiotic translation, as

stressed in the thesis, in the first place, one needs to be able to describe something

properly and in a complete way. The considerations outlined here are of interest

to anyone attempting to describe, to model, any material object in a useful way.

267



Chapter 10. Future work

It takes a good deal of maturity to see that every field of know-

ledge is the centre of all knowledge, and that it doesn’t matter

so much what you learn when you learn it in a structure that

can expand into other structures.

Northrop Frye, On education, 1988.

This project focussed on the generation of automated diagrams from descriptions

of bookbinding structures: what kind of information is necessary, what descriptions

can convey it, what are the implications of the transformation of the information

into a graphic representation. Such a project could be extended to cover its

practical application to the field and to other objects.

10.1. Integration of automated transformations within the data
gathering framework

In the near future, it is hoped to be able to offer the bookbinding description

schema through a web interface that would allow anyone interested to describe

their own bookbindings, ideally generating content for a central database of

bookbinding structures.

During the development phase of this research project, it has become empirically

clear that the automated visualizations can serve as data accuracy and error controls

for the database. Their development also helped a great deal this author in better

268



understanding the subtleties of the description schema. It would therefore seem

that integrating the generation of automated visualizations with the description

process could positively affect the data entry process, both helping the non-initiated

and the learner to better understand the description schema, and serving as visual

proof of the data being entered, thus increasing accuracy.

It was not possible during this research to test and measure the psychological

and cognitive effects of the automated visualizations on users and learners of the

description schema during the data entry process. Such a study would however

be highly desirable, as it could lead to even more efficient visualizations, and it

could provide a measure of the efficacy of the diagrams as data accuracy and

schema learning tools, beyond the mentioned empirical experience of these effects.

10.2. Verbal to visual and reverse: a mixed approach

This project focussed on the process of taking information from models of a

binding structure and transforming it into graphic form. Some bookbinding

components require a long series of elements in order to be described properly.

The more detail that is required, the more elements that need to be described,

the more time that is necessary for the description.

Drawings are capable of providing complex information in a synchronic manner.

It might be that complex structures could be presented to the user directly in

graphic form for the desired one to be chosen, and to subsequently feed the rel-

evant information into the database through a reverse visual-to-verbal transform-

ation. There are problems with this approach, as one would have to foresee all

possible alternative visualizations, generate them a priori, and present them to the

user. In turn, the user would have to go through the numerous graphic alternatives

and, through a spot-the-difference game, select the desired visualization. This

might also influence negatively the accuracy of the information recorded, as the

user, not guided by the schema through the option selection process, might not

understand in sufficient detail the model of the description.

269



A solution which would be helpful to investigate is a mixed approach. One

could describe a binding component down to a certain level of detail and have

the system generate a series of possible alternatives. If the description level is ap-

propriate, the number of alternative visualizations would be manageable. The in-

formation could be presented as a series of ‘small multiples’,492 small illustrations

provided with verbal labels which are positioned within the eyespan, so that the

viewer is able to make comparisons at a glance. This way, the user would not be

overwhelmed by the information and the differences between the visualizations

would be evident. By selecting the desired option, the information could be

transformed into verbal form for its inclusion in the database. Once again, there

would need to be a one-to-one relationship between visual and verbal descriptions.

Material components could take their verbal label, and spatial information could

be translated into the structured arrangement of the information.

In the same way, this mixed approach could be used to solve uncertainty caused

by insufficient schema development. The schema, for example, does not prescribe

the path of the thread around double sewing supports. The most probable path

is drawn, but then, on clicking on the uncertain path, the viewer could be

presented with alternative paths, for the desired one to be selected and recorded

in the database (see Figure 133).493

Figure 133. Small multiples to solve uncertainty due to schema limitations.

492. Tufte 1990, p. 67.
493. Campagnolo & Velios 2013.

270



10.3. Related work: Manuscript Collation Project

This project worked on XML structured descriptions of bookbinding structures.

The schema used did not cover gathering assembly structure information. Chapter

4 highlighted the fact that in recent times, people have enquired whether anyone

is working on a visualization of bookbinding structures. This project has brought

forward the idea that structured descriptions are capable of providing the right

information in a correct sign-language, i.e. one of the basic conditions for successful

intersemiotic translations. The structured descriptions presented here have been

encoded in an XML schema, and in fact all technologies involved have been strictly

XML based. This does not have to be taken to signify that only highly structured

and hierarchical XML schemas are capable of defining good structured descrip-

tions.

During the work on this research, the author has been involved in a related but

separate project on the visualization of manuscript collation formulas with col-

leagues at the Schoenberg Institute for Manuscript Studies, University of

Pennsylvania: Dot Porter and Doug Emery. In the summer of 2013, Porter494

posted on the TEI (Text Encoding Initiative)495 distlist, asking whether anyone

was visualizing with SVG collation formulas within TEI msdescription modules.496

While data on the gathering formation was not available in the St. Catherine’s

schema, the transformation principles could definitely be adapted to new data

and a different schema, so the collaboration began.

The information collected within collation formulas is highly condensed, but

it follows a strict ruling schema. Collation formulas are thus a kind of structured

description. Depending of the efficacy of the schema being implemented, collation

formulas can communicate both the material elements — i.e. the leaves, the quires

— and the form of gathering structures, and they also maintain a reference to the

logical form of quire assemblies. Therefore, even if not encoded in XML, collation

494. Porter 2013.
495. Text Encoding Initiative Consortium 2014b.
496. Text Encoding Initiative Consortium 2014a.

271



formulas meet the basic conditions for the implementation of intersemiotic

translations.

A first dataset was constituted by the Digital Walters collection.497 This digital

collection offered TEI P5498 descriptions of manuscripts with an ad hoc developed

collation formula schema.499 Whilst simplistic, this formula was still able to convey

the very basic information in regards to the gathering structure of those manu-

scripts. The formula within the TEI file was then parsed and the information was

transformed into an XML file written in accordance with a new schema. Then

from the XML, an XSLT transformation could take the information and transform

the formula into a series of SVG diagrams, one for each gathering. These trans-

formations are part of a larger Manuscript Collation Project,500 and integrated

within its webpages. The aim of the project is to provide the users with a view of

a manuscript based not on facing pages, as it is customary with online presenta-

tions, but on the physical structure of quires, thus showing conjoined and not fa-

cing pages together. The diagrams help keeping track of which pages are being

shown and their relationship with the others within the quire. At the time of

writing, the project is still at the development stage.

This project shows how the methodology developed for this research is applic-

able successfully also to other structured descriptions and datasets. In practical

terms, the methodology outlined throughout this thesis is applicable to the visual-

ization of collation formulas for a number of reasons. First of all, collation formulas

as a form of very dense structured descriptions that posses in their formation rules

a reference to the logical form of the object that they describe. Secondly, collation

formulas highlight whole/parts relationships (gatherings, bifolia, singletons) and

their spatial arrangement (which element is contained by which other elements).

Lastly, as mentioned in chapter 4, a prototypical shape for a folded bifolium can

be expressed through rounded horizontal U-like signs.

497. Walters Art Museum 2014b.
498. Sperberg-McQueen & Burnard 2005.
499. Walters Art Museum 2014a.
500. Porter et al. 2014; Emery et al. 2014.

272



Figure 134. Screenshot from the Manuscript Collation Project webpage. Reconstruction
of Quire 1 of the dismembered Galen Syriac palimpsest showing the proposed structure
of the first quire of the original Galen manuscript (Toth et al. 2010; Toth et al. 2013;
Bhayro et al. 2013). The pages shown come from the main corpus of the palimpsest and
from the Vat.sir.647 manuscript, held at the Vatican Library. On the left are visible the
diagrams generated from the collation formula.

Collation diagrams are not a novelty,501 however, the diagrams for the Manuscript

Collation project are a useful improvement. Being automated, in fact, they can act

as a way of ensuring that data collected within collation formulas — or other

collation description system — is correct, and they can also function as a simple

visual aid to understand the structure of gatherings within books as they are gen-

erated — in a consistent manner — when needed.

501. See for example: Dennison 1990, Noel 1995, or Pickwoad 2014a.

273



10.4. Final remarks

This dissertation has addressed the problem of whether and how it could be

possible to transform a detailed description of a material object, such as a book-

binding structure, into a visual diagram automatically.

This was essentially an intersemiotic translation problem. In order to investigate

the issues related to such a complex problem, the research developed in two

phases. First, it focussed on working from a theoretical perspective, attempting

to extrapolate and apply existing theories from multiple disciplines, from semiotics,

to cognitive psychology, perception, philosophy, and information visualization.

This was inevitable, since one needs broad perspectives for an investigation into

as general a problem as intersemiotic translations from verbally encoded inform-

ation to a visual output. Second, the project took a more practical approach, by

delving into the application of the theoretical framework researched in the first

part of the project to a specific domain, i.e. bookbinding structures, and producing

automated intersemiotic translations for such a domain. Although exposed here

in sequence, the two phases of this research happened in tandem and in a cyclical

way, each informing the other of the problems that needed considerations and of

the possible solutions; theoretical consideration found their application in the

visualizations, and visualization problems found their solutions in the theoretical

framework.

This research showed that verbal-to-visual transformations based on models of

are possible and that they can successfully convey the wanted information, if certain

criteria are met. This project has advanced a series of methodology recommenda-

tions for successful automated intersemiotic translations, and design recommend-

ations for the generation of standardized automated prototypical drawings of

bookbinding structures. For the first time, uncertainty has been systematically

flagged within visualization in the field of the archaeology of the book. It was also

found that the automation of the visualization brings in a series of added value

benefits of significant practical value, demonstrating the benefits of the implement-

ation of a system capable of automatically visualizing information contained

within structured descriptions.

274



Bibliography

NB: the symbol ¬ indicates that long URLs are broken across lines for justification purposes.

Adams, Thomas R. and Barker, Nicolas (1993), ‘A new model for the study of the book’, in Barker,

Nicolas (Ed.), A potencie of life: books in society, the Clark Lectures 1986-1987, London: British

Library, pp. 5-43.

Adler, Georg (2010), Handbuch Buchverschluss und Buchbeschlag: Terminologie und Geschichte

im deutschsprachigen Raum, in den Niederlanden und Italien vom frühen Mittelalter bis in die

Gegenwart, Wiesbaden: Reichert Verlag.

Adorni, Giovanni, Di Manzo, Mauro, Giunchiglia, Fausto (1984), ‘Natural language driven

image generation’, in Proceedings of the 10th International Conference on Computational

Linguistics and 22Nd Annual Meeting on Association for Computational Linguistics, Stroudsburg,

PA, USA: Association for Computational Linguistics, pp. 495–500.

Agrawala, Maneesh (2001), Visualizing route maps, PhD Thesis, Stanford University, Stanford

(CA), http://graphics.stanford.edu/papers/maneesh_thesis/thesis.pdf (Accessed 2013-02-08).

Agrawala, Maneesh, Li, Wilmot, Berthouzoz, Floraine (2011), ‘Design principles for visual

communication’, Communications of the ACM, 54, pp. 60-69.

Agrawala, Maneesh and Stolte, Chris (2001), ‘Rendering effective route maps. Improving usability

through generalization’, in SIGGRAPH 2001, SIGGRAPH, Los Angeles (CA): SIGGRAPH.

Albert, Marc K. and Hoffman, Donald D. (1995), ‘Genericity in spatial vision’, Geometric

representations of perceptual phenomena, pp. 95-112.

Albertazzi, Liliana (Ed.) (1999), Shapes of forms: from gestalt psychology and phenomenology to

ontology and mathematics, Dordrecht; Boston (MA): Kluwer Academic.

Alloway, Tracy P. and Alloway, Ross G. (Eds.) (2012), Working memory: the connected

intelligence, New York (NY); London: Psychology Press.

Allwein, Gerard and Barwise, Jon (Eds.) (1996), Logical Reasoning with Diagrams, Oxford; New

York (NY): Oxford University Press.

275



Alpers, Svetlana, Apter, Emily, Armstrong, Carol, Buck-Morss, Susan, Conley, Tom, Crary,

Jonathan, Crow, Thomas, Gunning, Tom, Holly, Michael Ann, Jay, Martin, Kaufmann,

Thomas Dacosta, Kolbowski, Silvia, Lavin, Sylvia, Melville, Stephen, Molesworth, Helen,

Moxey, Keith, Rodowick, D.N., Waite, Geoff, Wood, Christopher (1996), ‘Visual Culture

Questionnaire’, October, 77, pp. 25-70.

Anat, Biletzki and Anat, Matar (2014), ‘Ludwig Wittgenstein’, in Zalta, Edward N. (Ed.), The

Stanford Encyclopedia of Philosophy, Spring 2014 Edition, Stanford (CA): Stanford University.

Anderson, Priscilla (1999), ‘Fifteenth-century bookbinding structure in Italy and the Netherlands.

A survey of manuscripts and printed books’, The Book and Paper Group Annual, 18.

Arcelli, C., Cordella, L. P., Baja, G. S. di (Eds.) (1992), Visual form. Analysis and recognition,

New York (NY); London: Plenum Press.

Arnheim, Rudolf (1969), Visual thinking, Berkeley (CA): University of California Press.

Arnheim, Rudolf (1980), ‘A plea for visual thinking’, Critical Inquiry, 6, 3, pp. 489-497.

Asher, Nicholas and Vieu, Laure (1995), ‘Towards a geometry of common sense. A semantic and

complete axiomatization and mereotopology’, in Proceedings of the International Joint Conference

on Artificial Intelligence, IJCAI ‘95, Montréal: IJCAI, pp. 846-852.

Atkin, Albert (2010), ‘Peirce’s theory of signs’, in Zalta, Edward N. (Ed.), The Stanford

Encyclopedia of Philosophy, Winter 2010 Edition, Stanford (CA): Stanford University,

http://plato.stanford.edu/archives/win2010/entries/peirce-semiotics (Accessed 2011-11-11).

Attneave, F. (1954), ‘Some informational aspects of visual perception’, Psychological Review, 61,

3, pp. 183-193.

Aurnague, Michel and Vieu, Laure (1993), ‘Toward a formal representation of space in language.

A commonsense reasoning approach’, in Proceedings of the International Joint Conference on

Artificial Intelligence, IJCAI ‘93, Chambéry: IJCAI, pp. 123-158.

Baddeley, A. (1992), ‘Working memory’, Science, 255, 5044, pp. 556-559.

Baddeley, Alan (2000), ‘The episodic buffer: a new component of working memory?’, Trends in

cognitive sciences, 4, 11, pp. 417-423.

Baddeley, Alan (2003a), ‘Working memory and language. An overview’, Journal of Communication

Disorders, 36, 3, pp. 189-208.

Baddeley, Alan (2003b), ‘Working memory. Looking back and looking forward’, Nature Reviews

Neuroscience, 4, 10, pp. 829-839.

276



Baddeley, Alan D. (2004), ‘The psychology of memory’, in Baddeley, Alan D., Kopelman, Michael,

Wilson, Barbara A. (Eds.), The Essential Handbook of Memory Disorders for Clinicians, New

York (NY): John Wiley & Sons, pp. 1-13.

Baddeley, Alan D. (2007), Working memory, thought, and action, Oxford: Oxford University

Press.

Baddeley, Alan (2013), Essentials of human memory, New York (NY), London: Psychology Press.

Baddeley, Alan D. and Hitch, Graham J. (1974), ‘Working memory’, The psychology of learning

and motivation, 8, pp. 47-89.

Barker-Plummer, Dave (2002), ‘Diagrammatic reasoning’, in Barker-Plummer, Dave, Beaver,

David I., Benthem, Johan van, Scotto di Luzio, Patrick (Eds.), Words, proofs, and diagrams,

Stanford (CA): Center for the Study of Language and Information, pp. 1-4.

Barker-Plummer, Dave, Beaver, David I., Benthem, Johan van, Scotto di Luzio, Patrick (Eds.)

(2002), Words, proofs, and diagrams, Stanford (CA): Center for the Study of Language and

Information.

Barlow, H.B. and Reeves, B.C. (1979), ‘The versatility and absolute efficiency of detecting mirror

symmetry in random dot displays’, Vision research, 19, 7, pp. 783-793.

Barrow, Harry G. and Tenenbaum, Jay M. (1981), ‘Interpreting line drawings as three-dimensional

surfaces’, Artificial intelligence, 17, 1, pp. 75-116.

Barthes, Roland (1978), Image, music, text, translated by Heath, Stephen, New York (NY): Hill

and Wang.

Barwise, Jon and Shimoijima, Atsushi (1995), ‘Surrogate reasoning’, Cognitive Studies: Bulletin

of Japanese Cognitive Science Society, 4, 2, pp. 7-27.

Bentkowska-Kafel, Anna, Denard, Hugh, Baker, Drew (Eds.) (2012), Paradata and transparency

in virtual heritageDigital research in the arts and humanities, Farnham: Ashgate.

Benvestito, Claudia (2012), ‘Cucitura knot-tack: due esempi marciani’, in Brusegan, Marcello,

Eleuteri, Paolo, Fiaccadori, Gianfranco (Eds.), San Michele in Isola. Isola della conoscenza.

Ottocento anni di storia e cultura camaldolesi nella laguna di Venezia. Mostra organizzata in

occasione del millenario della fondazione della congregazione camaldolese, Torino: UTET, pp.

296-298.

Bertin, Jacques (1981), Graphics and Graphic Information Processing, New York (NY): Walter

de Gruyter & Co.

Bertin, Jacques (1983/2011), Semiology of graphics : diagrams, networks, maps, first English edition

with a new epilogue, Redlands (CA): ESRI Press; Distributed by Ingram Publisher Services.

277



Bhayro, S., Hawley, R., Kessel, G., Pormann, P. E. (2013), ‘The Syriac Galen palimpsest.

Progress, prospects and problems’, Journal of Semitic Studies, 58, 1, pp. 131-148.

Bibliothèque Nationale (France) (1989), Les tranchefiles brodées: étude historique et technique,

Paris: Bibliothèque Nationale.

Biederman, Irving (1985), ‘Human image understanding: recent research and a theory’, Computer

Vision, Graphics, and Image Processing, 32, 1, pp. 29-73.

Biederman, Irving (1987), ‘Recognition-by-components: a theory of human image understanding’,

Psychological Review, 94, pp. 115-147.

Biederman, Irving (1995), ‘Visual object recognition’, in Osherson, Daniel N., Kosslyn, Stephen

M., Gleitman, Lila R. (Eds.), An Invitation to Cognitive Science: Visual cognition, Cambridge

(MA): MIT Press, pp. 121-145.

Biederman, Irving (2000), ‘Recognizing depth-rotated objects: a review of recent research and

theory’, Spatial Vision, 13, 2-3, pp. 2-3.

Biederman, Irving and Gerhardstein, Peter C. (1993), ‘Recognising depth rotated objects: evidence

and conditions for three-dimensional viewpoint in variance’, Journal of Experimental Psychology:

Human perception and performance, 19, pp. 1162-1182.

Biederman, Irving and Ju, G. (1988), ‘Surface versus edge-based determinants of visual recognition’,

Cognitive Psychology, 20, 1, pp. 38-64.

Biggs, Michael A. R. (2004), ‘Visualisation and Wittgenstein’s "Tractatus"’, in Malcolm, M. (Ed.),

Multidisciplinary approaches to visual representations and interpretations, Amsterdam: Elsevier,

pp. 293-303.

Bijl, Aart (1986), ‘Designing with words and pictures in a logic modelling environment’, in

Computer-Aided Architecture Design Futures Proceedings, International Conference on

Computer-Aided Architectural Design, Delft: ICCAAD, pp. 128-145.

Bobick, Aaron F. (1987), Natural object categorization, Cambridge (MA): MIT Artificial Intelligence

Laboratory, http://dspace.mit.edu/handle/1721.1/6964 (Accessed 2011-11-28).

Boggess, Lois Carolyn (1978), Computational interpretation of English spatial prepositions, PhD

Thesis, University of Illinois at Urbana-Champaign, Champaign (IL).

Boggess, Lois Carolyn (1979), Computational interpretation English spatial prepositions, Ft. Belvoir

(VA): Defense Technical Information Center.

Bolter, Jay David and Grusin, Richard (2000), Remediation: Understanding New Media, 1st,

Cambridge, MA: The MIT Press.

278



Bor, Daniel (2012), The ravenous brain: how the new science of consciousness explains our insatiable

search for meaning, New York (NY): Basic Books.

Borillo, Mario and Goulette, Jean-Pierre (2006), ‘Computing architectural composition from

the semantics of the ‘Vocabulaire de l’architecture’’, Image, Journal of Interdisciplinary Image

Science, July.

Bosch, Gulnar K, Carswell, John, Petherbridge, Guy (1981), Islamic Bindings and Bookmaking:

A Catalogue of an Exhibition, May 18, - August 18, 1981, Chicago: Oriental Institute, University

of Chicago.

Bossche, Piet Vanden, Segers, Norbert, Wagemans, Johan, d’Ydewalle, Géry (1992), ‘Human

perception of affine equivalent polygons’, in Arcelli, Carlo, Cordella, Luigi P., Baja, Gabriella

Sanniti di (Eds.), Visual Form, New York (NY), London: Plenum Press, pp. 583-592.

Boudalis, Georgios (2007), ‘Endbands in Greek-style bindings’, The Paper Conservator, 31, 1,

pp. 29-49.

Boukhelifa, Nadia, Bezerianos, Anastasia, Isenberg, Tobias, Fekete, Jean-Daniel (2012),

‘Evaluating sketchy lines for the visualization of qualitative uncertainty’, IEEE Transactions on

Visualization and Computer Graphics, 18, 12, pp. 2769-2778.

Bowers, Fredson (1949/1994), Principles of bibliographical description, Winchester; New Castle

(DE): St. Paul’s Bibliographies; Oak Knoll Press.

Brachman, Ronald J. and Levesque, Hector J. (2004), Knowledge representation and reasoning

[electronic resource], Amsterdam ; Boston (MA): Morgan Kaufmann.

Brandimonte, Maria A. and Gerbino, Walter (1996), ‘When imagery fails: effects of verbal

recoding on accessibility of visual memories’, in Stretching the imagination: representation and

transformation in mental imagery, Oxford; New York (NY): Oxford University Press, pp. 31-76.

Brandimonte, Maria A., Hitch, Graham J., Bishop, Dorothy V. (1992a), ‘Influence of short-term

memory codes on visual image processing: evidence from image transformation tasks’, Journal

of experimental psychology: learning, memory and cognition, 18, 1, pp. 157-165.

Brandimonte, Maria A., Hitch, Graham J., Bishop, Dorothy V. (1992b), ‘Verbal recoding of

visual stimuli impairs mental image transformations’, Memory and Cognition, 20, 4, pp. 449-455.

Bray, Dirk de (1607-1658), Mensen en twee honden in een boekenwinkel (People and two dogs in

a book shop), http://www.rijksmuseum.nl/collectie/zoeken/asset.jsp?id=RP-T-1884-A-290

(Accessed 2011-12-26).

Bray, Dirk de (1658/1977), A short instruction in the binding of books, Amsterdam: Nico Israel.

279



Bray, Tim, Paoli, Jean, Sperberg-McQueen, C.M., Maler, Eve, Yergau, François (Eds.) (2008),

Extensible Markup Language (XML) Version 1.0 (Fifth edition), http://www.w3.org/TR/xml/

(Accessed 2014-03-22).

Breslauer, B. H. (1986), The Uses of Bookbinding LiteratureBook Arts Press occasional publications,

New York: Book Arts Press, School of Library Service, Columbia University.

British Census Project (1990), A guide to the census of western medieval bookbinding structures

to 1500 in British libraries, [s.l.]: [s.n.].

Brodlie, Ken, Allendes Osorio, Rodolfo, Lopes, Adriano (2012), ‘A review of uncertainty in

data visualization’, in Dill, J., Earnshaw, R., Kasik, D., Vince, J., Wong, P. C. (Eds.), Expanding

the Frontiers of Visual Analytics and Visualization, London: Springer-Verlag, pp. 81-109.

Brodribb, Conant (1970), Drawing archaeological finds for publication, London: John Baker.

Brown, Charity, Gehrke, Jürgen, Lloyd-Jones, Toby J. (2010), ‘A visual and semantic locus to

beneficial effects of verbalization on face memory’, The American journal of psychology, 123,

1, pp. 51-69.

Bryson, Norman (1991), ‘Semiology and visual interpretation’, in Bryson, Norman, Holly,

Michael Ann, Moxey, Keith P.F. (Eds.), Visual theory: painting and interpretation, New York

(NY): HarperCollins, pp. 61-73.

Burch, Robert (2010), ‘Charles Sanders Peirce’, in Zalta, Edward N. (Ed.), The Stanford

Encyclopedia of Philosophy, Fall 2010 Edition, Stanford (CA): Stanford University, http://pla¬

to.stanford.edu/archives/fall2010/entries/peirce (Accessed 2011-11-11).

Burks, Arthur W. (1949), ‘Icon, index, and symbol’, Philosophy and Phenomenological Research,

9, 4, pp. 673-689.

Campagnolo, Alberto (2014a), ‘Errata (per oculos) corrige. Visual identification of meaningless

data in database records of bookbinding structures’, in Care and conservation of manuscripts

15: proceedings of the fifteenth international seminar held at the University of Copenhagen,

2nd-4th April 2014, (in press), University of Copenhagen and the Royal Library of Denmark:

Museum Tusculanum Press; University of Copenhagen.

Campagnolo, Alberto (2014b), ‘La descrizione delle legature. Descrizioni e generazione

automatizzata di diagrammi di strutture di legature storiche’, Kermes: la rivista del restauro,

92, pp. 18-20.

Campagnolo, Alberto and Velios, Athanasios (2013), ‘Bindings of uncertainty. Visualizing

uncertain and imprecise data in automatically generated bookbinding structure diagrams’, in

Digital Humanities 2013, conference abstracts, Digital Humanities 2013, the 24th joint

international conference of the Association for Literary and Linguistic Computing and Association

for Computers and the Humanities and the 5th joint international conference of the Alliance of

280



Digital Humanities Organizations, Conference Abstracts, University of Nebraska-Lincoln, 16-19

July 2013, University of Nebraska-Lincoln (NE): Center for Digital Research in the Humanities,

pp. 135-138.

Carmichael, L., Hogen, H.P., Walter, A.A. (1932), ‘An experimental study of the effect of

language on the reproduction of visually perceived form’, Journal of Experimental Psychology,

15, 1, pp. 73-86.

Carvin, Denis (1988), La reliure médiévale du XIVeme et XVeme siècles: d’aprés les fonds des

bibliotèques d’Aix-en-Provence, Avignon, Carpentras et Marseille, Arles: Centre Interrégional

de Conservation des Livres.

Cassiodorus (1961), Cassiodori Senatoris Institutiones, Oxford: Clarendon Press.

Castellani, Enrico (1920), La legatura dei codici copti, unpublished in Henri Hyvernat files: MC

VIII.A.3, Roma.

Cave, Carolyn B. and Kosslyn, Stephen M. (1993), ‘The role of parts and spatial relations in object

identification’, Perception, 22, 2, pp. 229-248.

Chase, William G. and Simon, Herbert A. (1973), ‘Perception in chess’, Cognitive Psychology, 4,

1, pp. 55-81.

Chomsky, Noam (1957), Syntactic structuresJanua linguarum, The Hague, Paris: Mouton.

Clark, James (1999), XSL Transformations (XSLT) Version 1.0, http://www.w3.org/TR/xslt

(Accessed 2014-03-22).

Clarkson, Christopther (1978), ‘The Conservation of Early Books in Codex Form: A Personal

Approroach’, The Paper Conservator, 3, 1, pp. 33-50.

Clarkson, Christopther (1993), ‘English monastic bookbinding in the twelfth century’, in Maniaci,

Marilena and Munafò, Paola F. (Eds.), Ancient and medieval book materials and techniques:

Erice, 18-25 September 1992, Città del Vaticano: Biblioteca apostolica vaticana, pp. 181-200.

Clarkson, Christopther (1996), ‘A hitherto unrecorded English Romanesque book sewing

technique’, in Sharpe, John L (Ed.), Roger Powell, the complete binder: Liber Amicorum,

Turnhout: Brepols, pp. 215-239.

Clarkson, Christopher (1998), ‘Some representations of the book and book-making, from the

earliest codex forms to Jost Amman.’, in Sharpe, John Lawrence and Van Kampen, Kimberly

(Eds.), The Bible as book: the manuscript tradition, London; New Castle (DE): British Library;

Oak Knoll Press, in association with the Scriptorium, Center for Christian Antiquities, pp.

197-203.

281



Clarkson, Christopther (2005), Limp vellum binding and its potential as a conservation type structure

for the rebinding of early printed books: a break with nineteenth and twentieth century rebinding

attitudes and practices, Reprinted with a new introduction and corrections, Oxford: Christopher

Clarkson.

Claude-Lachenaud, Coline, Charton, Eric, Ozell, Benoit, Gagnon, Michel (2014), ‘A multimodal

interpreter for 3D visualization and animation of verbal concepts’, in Ninth International

Conference on Language Resources and Evaluation, LREC 2014, Reykjavik: European Language

Resources Association, pp. 3620-3627.

Cloonan, Michèle V. (1984), ‘Bookbinding terminology and descriptive bibliography: an analysis

of sources’, The New Bookbinder, 4, pp. 28-33.

Clowes, M.B. (1971), ‘On seeing things’, Artificial Intelligence, 2, 1, pp. 79-116.

Cockerell, Douglas (1953), Bookbinding and the care of books: a text-book for bookbinders and

librarians, fifth edition, with drawings by Noel Rooke and other illustrations, London: Sir Isaac

Pitman and Sons Ltd..

Cockerell, Sydney Morris (1958), The repairing of books, with illustrations by Joan Rix Tebbutt,

London: Sheppard Press.

Cohen, Henri and Lefebvre, Claire (2005), Handbook of categorization in cognitive science,

Amsterdam: Elsevier.

Conn, Donia and Verheyen, Peter (2003a), ‘A selected bibliography of historic endbands’, in

Standards, Seminar on Standards of Excellence in Hand Bookbinding, Denver (CO): Guild of

book workers.

Conn, Donia and Verheyen, Peter (2003b), ‘Foundational endband’, in Standards, Seminar on

Standards of Excellence in Hand Bookbinding, Denver (CO): Guild of book workers.

Cook, Jill (2013), Ice Age art: the arrival of the modern mind, London: British Museum Press.

Cornoldi, Cesare and Logie, Robert (1996), ‘Counterpoints in perception and mental imagery:

introduction’, in Stretching the imagination: representation and transformation in mental imagery,

Oxford; New York (NY): Oxford University Press, pp. 3-30.

Cowan, N (2001), ‘The magical number 4 in short-term memory: a reconsideration of mental

storage capacity’, The behavioral and brain sciences, 24, 1, pp. 87-185.

Coyne, Richard D., Roseman, M. A., Radford, A. D., Balachandran, M., Gero, John S. (1990),

Knowledge-based design systems, edited by Coyne, Richard D., Reading (MA): Addison-Wesley.

282



Coyne, Bob and Sproat, Richard (2001), ‘WordsEye. An automatic text-to-scene conversion

system’, in Proceedings of the 28th annual conference on Computer graphics and interactive

techniques, SIGGRAPH ‘01, Los Angeles (CA): ACM, pp. 487-496.

Dahlström, Erik, Dengler, Patrick, Grasso, Anthony, Lilley, Chris, McCormack, Cameron,

Schepers, Doug, Watt, Jonathan, Ferraiolo, Jon, Fujisawa, Jun (Eds.) (2011), Scalable Vector

Graphics (SVG) 1.1 (Second Edition), http://www.w3.org/TR/SVG11/ (Accessed 2014-03-22).

Darvas, György (2007), Symmetry: cultural-historical and ontological aspects of science-arts relations.

The natural and man-made world in an interdisciplinary approach, Basel: Birkhäuser Verlag.

Davis, Daniel (2013), Modelled on Software Engineering: Flexible Parametric Models in the Practice

of Architecture, PhD Thesis, School of Architecture and Design College of Design and Social

context, RMIT University, Melbourne, http://www.danieldavis.com/papers/danieldavis_thes¬

is.pdf (Accessed 2015-04-25).

De Luca, Livio (2013), ‘3D modelling and semantic enrichment in cultural heritage’, in Fritsch,

Dieter (Ed.), Photogrammetric week 2013, 53rd Photogrammetric week, Stuttgart: Institut für

Photogrammetrie, pp. 323-333.

De Luca, Livio, Véron, Philippe, Florenzano, Michel (2005), ‘Semantic-based modelling and

representation of patrimony buildings’, in Proceedings of the Workshop towards Semantic Virtual

Environments, SVE Workshop 2005, Villars: MIRAlab, University of Geneva and CUSO.

Dehaene, Stanislas (2009), Reading in the brain: the science and evolution of a human invention,

New York (NY); London: Viking.

Dennison, Lynda (1990), ‘Oxford, Exeter College MS 47. The importance of stylistic and

codicollogical analysis in its dating and localization’, in Medieval book production, assessing the

evicence. Proceedings of the 2nd conference of the Seminar in the History of the Book to 1500,

Oxford, July 1988, Los Altos Hills (CA); London: Anderson-Lovelace; Red Gull Press, pp.

41-59.

Descartes, René (1637/1985), The philosophical writings of Descartes, translated by Cottingham,

John, Stoothoff, Robert, Murdoch, Dugald, Cambridge; New York (NY): Cambridge University

Press.

Dewey, John (1946), ‘Peirce’s theory of linguistic signs, thought, and meaning’, The Journal of

Philosophy, 43, 4, pp. 85-95.

Djurcilov, S., Kim, K., Lermusiaux, P., Pang, A. (2001), ‘Volume rendering data with uncertainty

information’, in Ebert, D., Favre, J. M., Peikert, R. (Eds.), Data visualization, New York (NY):

Springer-Verlag, pp. 243-252.

283



Doerr, Martin, Plexousakis, Dimitris, Bekiari, Chryssoyla (2001), ‘A metamodel for part-whole

relationships for reasoning on missing parts and reconstruction’, in Kunii, Hideko S., Jajodia,

Sushil, Sølvberg, Arne (Eds.), Conceptual Modeling — ER 2001, Berlin; Heidelberg:

Springer-Verlag, pp. 412-425.

Donald, Merlin (1991), Origins of the modern mind: three stages in the evolution of culture and

cognition, Cambridge (MA): Harvard University Press.

Donald, Merlin (2001), A mind so rare: the evolution of human consciousness, New York (NY):

Norton.

Dondis, Donis A. (1973), Primer of Visual Literacy, Cambridge (MA): MIT Press.

Doresse, Jean (1949), ‘Nouveaux documents gnostiques coptes découverts en Haute-Égypte’,

Comptes rendus des séances de l’Académie des Inscriptions et Belles-Lettres, 93, 2, pp. 176-180.

Doresse, Jean (1961), ‘Les Reliures des manuscrits gnostiques coptes découverts à Khéoboskion’,

Revue d’Égyptologie, 13, pp. 27-49.

Doresse, Jean and Mina, Togo (1949), ‘Nouveaux textes gnostiques coptes découverts en

Haute-Egypte. La bibliotheque de Chenoboskion’, Vigiliae Christianae, 3, 3, pp. 129-141.

Dorsey, Jim (1989), ‘Techniques: knot-tack sewing in the 15th century’, Binders’ Guild Newsletter,

XII, 6, pp. 7-13.

Drap, Pierre, Merad, Djamal, Seinturier, Julien, Boï, Jean-Marc, Peloso, Daniela, Vannini,

Guido, Nucciotti, Michele, Pruno, Elisa (2012), ‘An information system for medieval

archaeology based on photogrammetry and archaeological database: the Shawbak Castle Project’,

in Ioannides, Marinos, Fritsch, Dieter, Leissner, Johanna, Davies, Rob, Remondino, Fabio,

Caffo, Rossella (Eds.), Progress in Cultural Heritage Preservation: 4th International Conference,

EuroMed 2012 : Limassol, Cyprus, October 29-November 3, 2012 : Proceedings, Berlin,

Heidelberg: Springer-Verlag, pp. 119-128.

Dylla, Kimberly, Frischer, Bernard, Mueller, Pascal, Ulmer, Andreas, Haegler, Simon (2010),

‘Rome reborn 2.0. A case study of virtual city reconstruction using procedural modeling

techniques’, in Frischer, Bernard, Webb, J., Koller, D. (Eds.), Making History Interactive.

Computer Applications and Quantitative Methods in Archaeology (CAA). Proceedings of the

37th International Conference, Williamsburg, Virginia, United States of America, March 22-26

(BAR International Series S2079), CAA 2009, Oxford: Archeopress, pp. 62-66.

Eco, Umberto (1975/2008), Trattato di semiotica generale, Milano: Bompiani.

Eco, Umberto (1979), The Role of the reader: explorations in the semiotics of texts, Bloomington

(IN): Indiana University Press.

Eco, Umberto (1993), La ricerca della lingua perfetta nella cultura europea, Bari; Roma: Laterza.

284



Eco, Umberto (1997), Kant e l’ornitorinco, Milano: Bompiani.

Eide, Øyvind (2012b), ‘Underspecified, ambiguous or formal. Problems in creating maps based

on texts’, in Meister, Jan Christoph (Ed.), Digital Humanities 2012, 24th joint international

conference of the Association for Literary and Linguistic Computing and the Association for

Computers and the Humanities and 5th joint international conference of the Association for

Literary and Linguistic Computing, the Association for Computers and the Humanities, the Society

for Digital Humanities – Société pour l’étude des médias interactifs, together with the Australasian

Association for Digital Humanities and centerNet, Conference Abstracts, University of Hamburg,

Germany, 16-22 July 2012, Digital Humanities 2012, University of Hamburg, Germany: Hamburg

University Press, Publishing House of the Hamburg State and University Library Carl von

Ossietzky, pp. 185-187.

Eide, Øyvind (2012a), The area told as a story: an inquiry into the relationship between verbal and

map-based expressions of geographical information, PhD Thesis, King’s College London, London.

Eide, Øyvind (2013), ‘Why maps are silent when texts can speak. Detecting media differences

through conceptual modelling’, in Buchroithner, Manfred F., Prechtel, Nikolas, Burghardt,

Dirk, Pipping, Karsten, Schröter, Benjamin (Eds.), From pole to pole. Proceedings of the 26th

International Cartographic Conference, Dresden, 25-30 August 2013, ICC 2013, Dresden:

International Cartographic Association, pp. 674-684.

Eide, Øyvind (2014), ‘Verbal expressions of geographical information’, in Barker, Elton,

Bouzarovski, Stefan, Pelling, Chris, Isaksen, Leif (Eds.), New worlds out of old texts. Developing

techniques for the spatial analysis of ancient narratives, in print, Oxford: Oxford University

Press.

Eiteljorg, Harrison II (2004), ‘Computing for archaeologists’, in Schreibman, Susan, Siemens,

Ray, Unsworth, John (Eds.), A companion to digital humanities, Oxford: Blackwell Publishing

Ltd, pp. 20-30.

Elleström, Lars (2010), ‘The modalities of media. A model for understanding intermedial

relations’, in Elleström, Lars (Ed.), Media borders, multimodality and intermediality,

Houndmills, Basingstoke, Hampshire; New York (NY): Palgrave Macmillan, pp. 11-48.

Emery, Doug, Porter, Dot, Campagnolo, Alberto (2014), ‘Taking manuscripts apart, and putting

them together’, in Digital Humanities 2014, the 25th joint international conference of the

Association for Literary and Linguistic Computing and Association for Computers and the

Humanities and the 6th joint international conference of the Alliance of Digital Humanities

Organizations, Conference Abstracts, EPFL-UNIL University of Lausanne, 7-12 July 2014, Digital

Humanities 2014, University of Lausanne: EPFL-UNIL University of Lausanne, pp. 458-459.

Encyclopædia Britannica (1911), Torus, http://en.wikisource.org/wiki/1911_Encyc¬

lop%C3%A6dia_Britannica/Torus (Accessed 2014-09-05 10:39:32).

285



Ericsson, K. Anders and Moxley, Jerad H. (2012), ‘Working memory that mediates experts’

performance. Why it is qualitatively different from traditional working memory’, in Alloway,

Tracy P. and Alloway, Ross G. (Eds.), Working memory: the connected intelligence, New York

(NY); London: Psychology Press, pp. 109-136.

Etemad, Elika J. (2011), Cascading Style Sheets (CSS) Snapshot 2010, http://www.w3.org/TR/css-

2010/ (Accessed 2015-05-05).

Federici, Carlo, Carvin, Denis, Houlis, Konstantinos, Pescalicchio, Francesca (1988), Scheda

di censimento delle legature medievali: protocollo di descrizione, Roma: Enidata, Istituto Centrale

per la Patologia del Libro.

Federici, Carlo and Houlis, Kostantinos (1988), Legature bizantine vaticane, Roma: Fratelli

Palombi.

Federici, Carlo and Pascalicchio, Francesca (1993), ‘A census of medieval bookbindings: early

examples’, in Maniaci, Marilena and Munafò, Paola F (Eds.), Ancient and medieval book

materials and techniques: Erice, 18-25 September 1992, Città del Vaticano: Biblioteca apostolica

vaticana, pp. 181-200.

Feldman, Jacob (1997), ‘The structure of perceptual categories’, Journal of Mathematical Psychology,

41, 2, pp. 145-170.

Feldman, Jacob (2000), ‘Bias toward regular form in mental shape spaces’, Journal of Experimental

Psychology: Human Perception and Performance, 26, 1, pp. 152-165.

Ferguson, Eugene S. (1992), Engineering and the mind’s eye, Cambridge (MA): MIT Press,.

Ferguson, Benjamin (2013), The representation of logical form: a dilemma, http://benjaminfer¬

guson.org/wp-content/uploads/2013/01/PH501-Saying-and-Showing-Paper.pdf (Accessed

2014-05-30).

Ferrier, Gillian Maver (1996), A geometry without angles. The case for a functional geometry of

spatial prepositions, PhD Thesis, University of Glasgow, , http://theses.gla.ac.uk/2045/ (Accessed

2014-09-07 08:42:11).

Findlay, John M. and Gilchrist, Iain D. (2003), Active vision: the psychology of looking and seeing,

Oxford; New York (NY): Oxford University Press.

Findlay, John M. and Gilchrist, Iain D. (2005), ‘Eye guidance and visual search’, in Underwood,

G. (Ed.), Cognitive Processes in Eye Guidance, Oxford: Oxford University Press, pp. 259-281.

Findler, N. V. (1979), Associative networks. Representation and use of knowledge by computers,

New York (NY): Academic Press.

Fine, Kit (1999), ‘Things and their parts’, Midwest Studies in Philosophy, 23, 1, pp. 61-74.

286



Fischer, Barbara (1994), ‘Old books: structural elements of their bindings. A survey’, in Di Febo,

Assunta, Putti, Maria Letizia, Istituto centrale di patologia del libro (Rome) (Eds.),

Bolletino dell’Istituto Centrale per la Patologia del Libro, 1990-1991, La legatura dei libri antichi

tra conoscenza, valorizzazione e tutela : convegno internazionale, Parma, 16-18 novembre 1989,

Parma: Istituto centrale per la patologia dei libro, pp. 251-262.

Fischer, Steven Roger (2005), A History of Writing, London: Reaktion Books.

Fish, Max H. (1978), ‘Peirce’s general theory of signs’, in Seboek, T.A. (Ed.), Sight, sound, and

sense, Bloomington (IN): Indiana University Press, pp. 31-70.

Fisler, Kathi (2002), ‘Diagrams and computational efficacy’, in Barker-Plummer, Dave, Beaver,

David I., Benthem, Johan van, Scotto di Luzio, Patrick (Eds.), Words, proofs, and diagrams,

Stanford (CA): Center for the Study of Language and Information, pp. 27-46.

Foot, Mirjam (1993), ‘Bookbinding and the History of Books’, in Barker, Nicolas (Ed.), A Potencie

of Life: Books in Society: The Clark Lectures 1986-1987, London: British Library, pp. 113-126.

Foot, Mirjam (2004), ‘Bookbinding research: pitfalls, possibilities and needs’, in Randeria, Phiroze

and Foot, Mirjam (Eds.), Eloquent witnesses: bookbindings and their history, a volume of essays

dedicated to the memory of Dr. Phiroze Randeria, London; New Castle (DE): Bibliographical

Society; British Library; Oak Knoll Press, pp. 13-29.

Ford, David (1993), ‘The nature of clarity in archaeological line drawings’, Journal of Field

Archaeology, 20, 3, pp. 319-333.

Forrest, Peter (2012), ‘The identity of indiscernibles’, in Zalta, Edward N. (Ed.), The Stanford

Encyclopedia of Philosophy, Winter 2012 Edition, Stanford (CA): Stanford University,

http://plato.stanford.edu/archives/win2012/entries/identity-indiscernible (Accessed 2013-01-03).

Freudenberg, Bert, Masuch, Maic, Röber, Niklas, Strothotte, Thomas (2001), ‘The

Computer-Visualistik-Raum. Veritable and inexpensive presentation of a virtual reconstruction’,

in Proceedings of the 2001 Conference on Virtual Reality, Archeology, and Cultural Heritage,

New York (NY): ACM, pp. 97-102.

Freyd, J. and Tversky, Barbara (1984), ‘Force of symmetry in form perception’, The American

Journal of Psychology, 97, 1, pp. 109-126.

Gabbay, Dov M. and Smets, Philippe (1998), Quantified representation of uncertainty and

imprecision, New York (NY): Springer.

Garner, Wendell R. (1974), The processing of information and structure, Potomac (MD): L.

Erlbaum Associates; distributed by Halsted Press New York.

Garrod, Simon, Ferrier, Gillian, Campbell, Siobhan (1999), ‘IN and ON. Investigating the

functional geometry of spatial prepositions’, Cognition, 72, 2, pp. 167-189.

287



Garroni, Emilio (1972), Progetto di semiotica: messaggi artistici e linguaggi non-verbali, problemi

teorici e applicativi, Bari: Laterza.

Gaskell, Philip (1972), A New Introduction to Bibliography, Oxford: Clarendon Press.

Gauthier, Isabel, James, Thomas W., Curby, Kim M., Tarr, Michael J. (2003), ‘The influence of

conceptual knowledge on visual discrimination’, Cognitive neuropsychology, 20, 3, pp. 507-523.

Geertz, Clifford (1973), The interpretation of cultures. Selected essays, New York: Basic Books.

Gethin Powell, Rees (2012), ‘Uncertain date, uncertain place. Interpreting the history of Jewish

communities in the Byzantine empire using GIS’, in Meister, Jan Christoph (Ed.), Digital

Humanities 2012, conference abstracts, Digital Humanities 2012, 24th joint international

conference of the Association for Literary and Linguistic Computing and the Association for

Computers and the Humanities and 5th joint international conference of the Association for

Literary and Linguistic Computing, the Association for Computers and the Humanities, the Society

for Digital Humanities - Société pour l’étude des médias interactifs, together with the Australasian

Association for Digital Humanities and centerNet, Conference Abstracts, University of Hamburg,

Germany, 16-22 July 2012, University of Hamburg, Germany: Hamburg University Press,

Publishing House of the Hamburg State and University Library Carl von Ossietzky, pp. 329-331.

Gnirrep, Kees, Gumbert, J.P., Szirmai, János A. (1992), Kneep en binding: een terminologie voor

de beschrijving van de constructies van oude boekbanden, Den Haag: Koninklijke Bibliotheek.

Gnoli, Claudio (2008), ‘L’organizzazione della conoscenza e i suoi fondamenti’, in Gnoli, Claudio

and Scognamiglio, Carlo (Eds.), Ontologia e organizzazione della conoscenza. Introduzioni ai

fondamenti teorici dell’indicizzazione semantica, Lecce: Pensa Multimedia, pp. 69-137.

Gobet, Fernand and Clarkson, Gary (2004), ‘Chunks in expert memory: evidence for the magical

number four … or is it two?’, Memory, 12, 6, pp. 732-747.

Goldmeier, Erich (1972), Similarity in visually perceived forms, Madison (CT): International

Universities Press.

Goldschmidt, Ernst P, (1928), Gothic and Renaissance bookbindings: exemplified and illustrated

from the author’s collection, London; Boston: E. Benn; Houghton Mifflin Co.

Goldschmidt, Ernst P. (1945), ‘The study of early bookbinding’, in The Bibliographical Society,

1892-1942: studies in retrospect, London: Printed for the Bibliographical Society, pp. 175-184.

Goodchild, M., Buttenfield, B., Wood, J. (1994), ‘On introduction to visualizing data validity’,

in Hearnshaw, H. M. and Unwin, D. J. (Eds.), Visualization in geographical information systems,

New York (NY): Wiley, pp. 141-149.

Goodman, Nelson (1976), Languages of art: an approach to a theory of symbols, Indianapolis (IN);

Cambridge (MA): Hackett Publishing Co, Inc.

288



Gorlée, Dinda L. (1989), ‘Wittgenstein, Translation, and Semiotics’, Target, 1, 1, pp. 69-94.

Gottwald, Siegfried and Anat, Matar (2014), ‘Many-valued logic’, in Zalta, Edward N. (Ed.),

The Stanford Encyclopedia of Philosophy, Spring 2014 Edition, Stanford (CA): Stanford

University.

Goulette, Jean-Pierre (1999), ‘Sémantique formelle de l’espace. Une application au raisonnement

spatial qualitatif en architecture’, Revue intellectica, 2, 29, pp. 9-34.

Greaves, Mark Thomas (2002), The philosophical status of diagrams, Stanford (CA): Center for

the Study of Language and Information.

Greenfield, Jane (1990), Headbands: how to work them, second revised edition, New Castle (DE):

Oak Knoll Books.

Greenfield, Jane (1998), ABC of bookbinding: a unique glossary with over 700 illustrations for

collectors & librarians, New Castle (DE); New York (NY): Oak Knoll Press; Lyons Press.

Greenlee, Douglas (1973), Peirce’s concept of sign, The Hague: Mouton.

Griethe, Henning and Schumann, Heidrun (2006), ‘The visualization of uncertain data. Methods

and problems’, in Proceedings of the 17th Simulation and Visualization Conference, Aarhus:

Interaction Design Foundation, pp. 143-156.

Griffiths, N., Jenner, Anne, Wilson, Christine (1990), Drawing archaeological finds: a handbook,

Revised edition, London: Archetype Publications Ltd.

Grosdidier de Matons, Dominique, Hoffmann, Philippe, Vezin, Jean (1993), ‘Le recensement

des reliure anciennes conservées dans les collections publiques de France: réflexions sur une

mèthode de travail’, in Maniaci, Marilena and Munafò, Paola F (Eds.), Ancient and medieval

book materials and techniques: Erice, 18-25 September 1992, Città del Vaticano: Biblioteca

apostolica vaticana, pp. 157-179.

Gruijs, Albert (1974), ‘Le protocole de la restauration et la description des cahiers et bifolia’, in

Les techniques de laboratoire dans l’ètude des manuscrits : [actes du colloque international] Paris,

13-115 septembre 1972, Paris: Centre national de la recherche scientifique, pp. 253-255.

Halverson, John (1992), ‘The first pictures: perceptual foundations of Paleolithic art’, Perception,

21, 3, pp. 389-404.

Hamilton, Kelly Ann (2001a), ‘Some philosophical consequences of Wittgenstein’s aeronautical

research’, Perspectives on Science, 9, 1, pp. 1-37.

Hamilton, Kelly Ann (2001b), ‘Wittgenstein and the Mind’s Eye’, in Wittgenstein: biography and

philosophy, Cambridge: Cambridge University Press, pp. 53-97.

289



Hamilton, Kelly Ann (2002), ‘Darstellungen in The Principles of Mechanics and the Tractatus:

the representation of objects in relation in Hertz and Wittgenstein’, Perspectives on Science,

10, 1, pp. 28-68.

Hamilton, Kelly Ann (2012), ‘Wittgenstein and Tufte on thinking in 3D: ‘Escaping Flatland’’, in

EVA London 2012: electronic visualisation and the arts, Electronic visualisation and the arts

(EVA 2012), London: BCS, The Chartered Institute for IT, pp. 107-118.

Hammer, Eric M. (1995), Logic and visual information, Stanford (CA): Center for the Study of

Language and Information.

Hampton, J.A. (1981), ‘An investigation of the nature of abstract concepts’, Memory & Cognition,

9, 2, pp. 149-156.

Harrison, Matthew James, Keay, Simon, Earl, Greame (2012), ‘Shape grammar modelling and

the visualisation of an uncertain past’, in CAA2012 Proceeding of the 40th Conference in

Computer Application and Quantitative Methods in Archaeology, Southampton, UK, 26-30 March

2012, CAA 2012, Southampton: CAA, pp. 895-909.

Harte, Verity (2002), Plato on parts and wholes: the metaphysics of structure, Oxford: Clarendon

Press.

Heilbronner, Edgar and Dunitz, Jack D. (1993), Reflections on symmetry: in chemistry and

elsewhere, Basel: Wiley-VCH.

Heiser, Julie and Tversky, Barbara (2002), ‘Diagrams and descriptions in acquiring complex

systems’, in Gray, Wayne and Schunn, Christian D. (Eds.), Proceedings of the meetings of

Cognitive Science Society, Twenty-Fourth Annual CSS Conference, George Mason University in

Fairfax Virginia, USA, August 7-10, 2002, Fort Belvoir (VA): Defense Technical Information

Center, pp. 447-452.

Heiser, Julie and Tversky, Barbara (2006), ‘Arrows in comprehending and producing mechanical

diagrams’, Cognitive Science, 30, 3, pp. 581-592.

Hepting, Daryl (1999), The history of a picture’s worth, http://www2.cs.uregina.ca/~hept¬

ing/research/web/words/history.html#27ad (Accessed 2013-05-26).

Herskovits, Annette (1980), ‘On the spatial uses of prepositions’, in Proceedings of the 18th annual

meeting on Association for Computational Linguistics, Philadelphia (PA): Association for

Computational Linguistics, pp. 1-5.

Hetherington, Robina, Farrimond, Brian, Presland, Steve (2006), ‘Information rich temporal

virtual models using X3D’, Computers & Graphics, 30, 2, pp. 287-298.

Hinton, Geoffrey (1979), ‘Some Demonstrations of the Effects of Structural Descriptions in

Mental Imagery’, Cognitive Science, 3, 3, pp. 231-250.

290



Hochberg, Julian (1978), Perception, second revised edition, Englewood Cliffs (NJ): Prentice Hall.

Hochberg, Julian and Brooks, Virginia (1962), ‘Pictorial recognition as an unlearned ability: a

study of one child’s performance’, The American Journal of Psychology, 75, 4, pp. 624-628.

Hoopes, James (1991), Peirce on signs. Writings on semiotic, Chapel Hill (NC): University of North

Carolina Press.

Horn, Robert E. (1998), Visual language: global communication for the 21st century, Bainbridge

Island (WA): Macrovu, Inc..

Houlis, Konstantinos and Pescalicchio, Francesca (1988), Scheda di censimento delle legature

medievali: bordereau, Roma: Enidata, Istituto Centrale per la Patologia del Libro.

Housecroft, Catherine and Sharpe, Alan G. (2012), Inorganic chemistry, 4th edition, Harlow;

New York (NY): Prentice Hall.

Howard, Ian P. and Templeton, William Brian (1966), Human spatial orientation, London: John

Wiley & Sons.

Huffman, David A. (1971), ‘Impossible objects as nonsense sentences’, Machine intelligence, 6,

1, pp. 295-323.

Hunt, Leigh (1823), ‘Among my books’, Literary Examiner, 5-12 July, http://spenseri¬

ans.cath.vt.edu/TextRecord.php?textsid=36476 (Accessed 2013-06-28).

Ishihara-Brito, Reiko (2011), Archaeological Illustration of Ceramics: drawing conventions and

practices in the Maya area. Handbook for the 1st annual Maya at the Lago Workshop, April 14-17,

2011, Washington (DC): Dumbarton Oaks Research Library and Collection.

Iversen, Margaret (1986), ‘Saussure versus Peirce: models for a semiotics of visual art’, in Rees,

A.L. and Borzello, F. (Eds.), The new art history, London: Camden Press, pp. 82-94.

Ivins, William M. (1953/1969), Prints and visual communication, Cambridge (MA): MIT Press.

Ivins, William M. (1973), On the rationalization of sight: with an examination of three Renaissance

texts on perspective, New York (NY): Da Capo Press.

Jakobson, Roman (1959), ‘On linguistic aspects of translation’, in Brower, R. A. (Ed.), Translation,

Cambridge (MA): Harvard University Press, pp. 232-239.

Jemolo, V. and Morelli, M. (Eds.) (1990), Guida a una descrizione uniforme dei manoscritti e al

loro censimento, Roma: Istituto Centrale per il Catalogo Unico.

Johansson, Richard, Berglund, Anders, Danielsson, Magnus, Nugues, Pierre (2005), ‘Automatic

text-to-scene conversion in the traffic accident domain’, in Proceedings of the Nineteenth

291



International Joint Conference on Artificial Intelligence, Edinburgh, Scotland, UK, July 30-August

5, 2005, IJCAI2005, Edinburgh: Professional Book Center, pp. 1073-1078.

Johansson, Richard, Williams, David, Nugues, Pierre (2004), ‘Converting texts of road accidents

into 3D scenes’, in López De Mántaras, R. and Saitta, L. (Eds.), Proceedings of the 16th

European Conference on Artificial Intelligence, ECAI 2004, Valencia: IOS Press, pp. 1035-1045.

Johnson, Richard (1994), Schematic Diagrams, Chennai (Tamil Nadu, India): Prompt publications.

Johnson, Chris R. and Sanderson, Allen R. (2003), ‘A Next Step: Visualizing Errors and

Uncertainty’, IEEE Computer Graphics and Applications, 23, 5, pp. 6-10.

Johnston, Mark (2002), ‘Parts and principles: false axioms in mereology’, Philosophical Topics,

30, 1, pp. 129-166.

Jolicoeur, Pierre, Gluck, Mark A., Kosslyn, Stephen M. (1984), ‘Pictures and names: making

the connection’, Cognitive psychology, 16, 2, pp. 243-275.

Jäckel, Von Karl (1974), ‘Alte Techniken des Buchbinderhandwerks in der modernen

Schriftgutrestaurierung. 1 Das Heften’, Bibliotheksforum Bayern, 2, 3, pp. 255-264.

Jäckel, Von Karl (1975), ‘Alte Techniken des Buchbinderhandwerks in der modernen

Schriftgutrestaurierung. 2 Das Kapital’, Bibliotheksforum Bayern, 3, 3, pp. 207-219.

Jäckel, Von Karl (1977), ‘Alte Techniken des Buchbinderhandwerks in der modernen

Schriftgutrestaurierung. 3 Der Holzdeckel’, Bibliotheksforum Bayern, 5, 2, pp. 122-130.

Kennedy, John M. (1974), A psychology of picture perception, San Francisco (CA): Jossey-Bass

Publishers.

Kennedy, John M. (1977), ‘Ancient and modern picture-perception abilities in Africa’, The Journal

od Aesthetics and Art Criticism, 35, 3 (Spring), pp. 293-300.

Kennedy, Maev (2011), ‘British Library seeks £9m to buy oldest book in Europe’, The Guardian,

14 July, http://www.guardian.co.uk/books/2011/jul/14/british-library-seeks-buy-oldest-book

(Accessed 2013-05-20).

Kennedy, John M. and Ross, Abraham S. (1975), ‘Outline picture perception by the Songe of

Papua’, Perception, 4, 4, pp. 391-406.

Kioussi, Anastasia, Karoglou, Maria, Bakolas, Asterios, Moropoulou, Antonia (2012), ‘Integrated

documentation protocols enabling decision making in cultural heritage protection’, in Ioannides,

Marinos, Fritsch, Dieter, Leissner, Johanna, Davies, Rob, Remondino, Fabio, Caffo, Rossella

(Eds.), Progress in Cultural Heritage Preservation: 4th International Conference, EuroMed 2012

: Limassol, Cyprus, October 29-November 3, 2012 : Proceedings, Berlin, Heidelberg:

Springer-Verlag, pp. 211-220.

292



Kirsh, David (2002), ‘Why illustrations aid understanding’, in Ploetzner, R. (Ed.), Proceedings

of the international workshop on dynamic visualizations and learning, Tubingen, Germany,

Tubingen: Knowledge Media Research Center, pp. 1-9.

Kitcher, Patricia (1988), ‘Marr’s computational theory of vision’, Philosophy of Science, 55, 1, pp.

1-24.

Klatzky, Roberta L., Martin, G.L., Kane, R.A. (1982), ‘Semantic interpretation effects on memory

for faces’, Memory & cognition, 10, 3, pp. 195-206.

Klee, Eleonore (1978), ‘Knotenheftung des 15. Jahrhunderts’, Codices manuscripti: Zeitschrift für

Handschriftenkunde, 4, 3, pp. 89-92.

Koffka, Kurt (1935), Principles of Gestalt psychology, Oxford: Harcourt Brace.

Koslicki, Kathrin (2007), ‘Towards a Neo-Aristotelian mereology’, Dialectica, 61, pp. 127-159.

Koslicki, Kathrin (2008), The structure of objects, Oxford; New York (NY): Oxford University

Press.

Kosslyn, Stephen M. (1980), Image and mind, Cambridge (MA): Harvard University Press.

Kosslyn, Stephen M. (1986), Image and mind, Cambridge (MA): Harvard University Press.

Kosslyn, Stephen M., Alpert, Nathaniel M., Thompson, William L., Maljkovic, Vera, Weise,

Steven B., Chabris, Christopher F., Hamilton, Sania E., Rauch, Scott L., Buonanno, Ferdinando

S. (1993), ‘Visual mental imagery activates topographically organized visual cortex: PET

investigations’, Journal of Cognitive Neuroscience, 5, 3, pp. 263-287.

Kosslyn, Stephen M. and Ganis, Giorgio (2010), The case for mental imagery, Oxford; New York

(NY): Oxford University Press.

Kosslyn, Stephen M. and Thompson, William L. (2003), ‘When is early visual cortex activated

during visual mental imagery?’, Psychological bulletin, 129, 5, pp. 723-746.

Kotiadis, K. and Robinson, S. (2008), ‘Conceptual modelling. Knowledge acquisition and model

abstraction’, in Mason, J.S., Hill, R.R., Mönch, M., Rose, O., Jefferson, T., Fowler, J.W.

(Eds.), Proceedings of the 2008 Winter Simulation Conference, Winter Simulation Conference

2008, Miami (FL): Institute of Electrical and Electronics Engineers, pp. 951-958.

Kress, Gunther (2000), ‘Text as the punctuation of semiosis: pulling at some threads’, in Meinhof,

Ulrike Hanna and Smith, Jonathan (Eds.), Intertextuality and the media: from genre to everyday

life, Manchester; New York (NY): Manchester University Press; St. Martin’s Press, pp. 132-154.

293



Kubovy, Michael (1981), ‘Concurrent-Pitch Segregation and the Theory of Indispensable Attributes’,

in Kubovy, Michael and Pomerantz, James R (Eds.), Perceptual organization, Hillsdale (NJ):

Lawrence Erlbaum Associates, pp. 55-98.

Kurata, Yohei and Egenhofer, Max J. (2005a), ‘Semantics of simple arrow diagrams’, in

Barkowsky, Thomas, Freksa, Christian, Hegarty, Mary, Lowe, Ric (Eds.), Reasoning with

mental external diagrams: computational modeling and spatial assistance. Technical report SS-05-06,

2005 AAAI Spring Symposium, Menlo Park (CA): AAAI Press, pp. 101-104.

Kurata, Yohei and Egenhofer, Max J. (2005b), ‘Structure and semantics of arrow diagrams’,

Lecture notes in computer science, no. 3693, pp. 232-250.

Köhler, Wolfgang (1929), Gestalt psychology, Oxford: Liveright.

Lakoff, George P. (1987), Women, fire and dangerous things: what categories reveal about the

mind, Chicago (IL): University of Chicago Press.

Latour, Bruno (1986), ‘Visualization and cognition: thinking with eyes and hands’, Knowledge

and society, studies in the sociology of culture past and present, 6, 1, pp. 1-40.

Latta, Greg ([s.n.]), The AA8V Twinplex Regenerative Receiver. Schematic diagrams and circuit

descriptions, http://faculty.frostburg.edu/phys/latta/ee/twinplex/schematic/twinplexschemat¬

ic.html (Accessed 2014-09-04).

Leeuwenberg, Emanuel Laurens Jan (1968), Structural information of visual patterns. An efficient

coding system in perception, The Hague, Paris: Mouton.

Leeuwenberg, Emanuel Laurens Jan (1969), ‘Quantitative specification of information in sequential

patterns’, Psychological Review, 76, 2, pp. 216-220.

Leeuwenberg, Emanuel Laurens Jan (1971), ‘A Perceptual Coding Language for Visual and

Auditory Patterns’, The American Journal of Psychology, 84, 3, pp. 307-349.

Lemon, Oliver (2002), ‘Comparing the efficacy of visual languages’, in Barker-Plummer, Dave,

Beaver, David I., Benthem, Johan van, Scotto di Luzio, Patrick (Eds.), Words, proofs, and

diagrams, Stanford (CA): Center for the Study of Language and Information, pp. 47-70.

Levey, Martin (1962), Mediaeval Arabic Bookmaking: And Its Relation to Early Chemistry and

Pharmacology [by] Martin LeveyTransactions of the American Philosophical Society, Philadelphia:

The American Philosophical Society.

Lewis, Philippa and Darley, Gillian (1986), Dictionary of ornament, New York (NY): Pantheon

Books.

Leyton, Michael (1984), ‘Perceptual organization as nested control’, Biological Cybernetics, 51,

3, pp. 141-153.

294



Leyton, Michael (1986a), ‘A theory of information structure I: general principles’, Journal of

Mathematical Psychology, 30, 2, pp. 103-160.

Leyton, Michael (1986b), ‘A theory of information structure II: a theory of perceptual organization’,

Journal of Mathematical Psychology, 30, 3, pp. 257-305.

Leyton, Michael (1987), ‘A limitation theorem for the differentiable prototypification of shape’,

Journal of Mathematical Psychology, 31, 4, pp. 307-320.

Leyton, Michael (1992), Symmetry, causality, mind, Boston (MA): MIT Press.

Leyton, Michael (2001), A generative theory of shape, Berlin; Heidelberg: Springer.

Ligatus (2007), Survey schemas. Saint Catherine’s project, http://www.ligatus.org.uk/stcather¬

ines/node/1052 (Accessed 2014-08-22).

Ligatus (2013), Ligatus bookbinding glossary, http://www.ligatus.org.uk/glossary/ (Accessed

2013-07-28).

Lilley, Chris (1996), W3C scalable graphics requirements, http://www.w3.org/Graphics/Scalable¬

Req.html (Accessed 2015-04-25).

Little, Martha (2010), ‘Evidence of structure and procedure in books. Selected examples of

potential clues’, in Standards, Seminar on Standards of Excellence in Hand Bookbinding, Tucson

(AZ): Guild of book workers.

Lockwood, Kate, Forbus, Ken, Halstead, D., Usher, Jeffrey (2006), ‘Automatic categorization

of spatial prepositions’, in Proceedings of the 28th annual conference of the Cognitive Science

Society, Conference of the cognitive sciences, Wheat Ridge (CO): Cognitive Science Society, pp.

1705-1710.

Lopes, Dominic (1996), Understanding Pictures, Oxford: Clarendon Press.

Lu, Zhong-Lin and Dosher, Barbara (2013), Visual psychophysics. From laboratory to theory,

Cambridge (MA): MIT Press.

MacEachren, A. M (1992), ‘Visualizing uncertain information’, Cartographic Perspectives, 13, pp.

10-19.

MacEachren, Alan M., Robinson, Anthony, Gardner, Steven, Murray, Robert, Gahegan, Mark,

Hetzler, Elisabeth (2005), ‘Visualizing geospatial information uncertainty. What we know

and what we need to know’, Cartography and Geography Information Science, 32, 3, pp. 139-160.

Mach, Ernst (1897/1959), The analysis of sensations, and the relation of the physical to the psychical,

New York: Dover Publications.

295



Madan, Falconer, Turbutt, Gladwyn, Gibson, Strickland (1905), The original Bodleian copy of

the first folio of Shakespeare (The Turbutt Shakespeare), Oxford: Clarendon Press.

Malik, Jitendra (1987), ‘Interpreting line drawings of curved objects’, International Journal of

Computer Vision, 1, 1, pp. 73-103.

Marr, David (1982), Vision: a computational investigation into the human representation and

processing of visual information, San Francisco (CA): W. H. Freeman & Company.

Marr, David and Nishihara, H.K. (1978a), ‘Representation and recognition of the spatial

organization of three-dimensional shapes’, Proceedings of the Royal Society of London. Series

B, Biological Sciences, 200, 1140, pp. 269-294.

Marr, D and Nishihara, H K (1978b), ‘Representation and recognition of the spatial organization

of three-dimensional shapes’, Proceedings of the Royal Society of London. Series B, Containing

papers of a Biological character. Royal Society (Great Britain), 200, 1140, pp. 269-294.

Martens, Femke, Muši , B., Poblome, Jeroen, Waelkens, Marc (2012), ‘The integrated urban

survey at Sagalassos’, in Vermeulen, Frank, Burgers, Gert-Jan, Keay, Simon, Corsi, Cristina

(Eds.), Urban landscape survey in Italy and the Mediterranean, Oxford: Oxbow Books, pp.

84-93.

Maury, André (1983), ‘Reality and logical form’, Synthese, 56, 2, pp. 171-180.

McCall, Matthew (2009), ‘The points of the picture: Hertz and Wittgenstein on the picture

theory’, Papers from the 32nd International Wittgenstein Symposia, Kirchberg am Wechsel, pp.

269-271.

McCarty, Willard (2005), Humanities computing, Houndmills, Basingstoke, Hampshire; New

York (NY): Palgrave Macmillan.

McCollough, Andrew and Vogel, Edward (2007), ‘Visual chunking allows efficient allocation

of memory capacity’, Journal of Vision, 7, 9, pp. 861-861.

McKenzie, Donald F. (1999), Bibliography and the sociology of texts, Cambridge; New York (NY):

Cambridge University Press.

McKie, Robin (2012), ‘When Homo sapiens hit upon the power of art’, The Guardian, 9 December,

http://www.guardian.co.uk/science/2012/dec/09/ice-age-art-exhibition-british-museum

(Accessed 2013-06-05).

McKitterick, David (2007), ‘Perspectives for an international history of the book’, in Eliot,

Simon, Nash, Andrew, Willison, Ian Roy (Eds.), Literary cultures and the material book,

London: British Library, pp. 413-429.

296



McLuhan, Marshall (1964/2001), Understanding media: the extensions of man, London; New

York (NY): Routledge.

McLuhan, Marshall and Powers, Bruce R. (1989), The global village: transformations in world

life and media in the 21st century, Oxford; New York (NY): Oxford University Press.

Mermet, Jean P. (Ed.) (1993), Fundamentals and standards in Hardware Description Languages,

Dordrecht; Boston (MA): Kluwer Academic Publishers.

Mervis, Carolyn B. and Rosch, Eleanor (1981), ‘Categorization of natural objects’, Annual Review

of Psychology, 32, 1, pp. 89-115.

Middleton, Bernard C. (1996), A history of English craft bookbinding technique, fourth rev. ed.,

New Castle (DE); London: Oak Knoll Press; British Library.

Miller, George (1956), ‘The magical number seven, plus or minus two: Some limits on our capacity

for processing information’, The psychological review, 63, pp. 81-97.

Miller, George A. (2003), ‘The cognitive revolution: a historical perspective’, TRENDS in Cognitive

Sciences, 7, 3, pp. 141-144.

Miller, George Armitage and Johnson-Laird, Philip Nicholas (1976), Language and perception,

Cambridge (MA): Belknap Press of Harvard University Press.

Miller, P. and Richards, J. (1995), ‘The good, the bad and the downright misleading.

Archaeological adoption of computer visualization’, in Hugget, J. and Ryan, N. (Eds.), Computer

Applications and Quantitative Methods in Arcgaeology, Glasgow, 1994, CAA1994, Oxford:

Tempus Reparatum, pp. 249-254.

Mitchell, William J.T. (1980), The Language of images, Chicago (IL): University of Chicago Press.

Mitchell, William J.T. (1986), Iconology: image, text, ideology, Chicago (IL); London: University

of Chicago Press.

Mitchell, William J.T. (1994), Picture theory: essays on verbal and visual representation, Chicago

(IL); London: University of Chicago Press.

Mocean, Loredana (2007), ‘Internet data validation’, Economy Informatics, 1, 1, pp. 96-99.

Monedero, Javier (2000), ‘Parametric design: a review and some experiences’, Automation in

Construction, 9, 4, pp. 369-377.

Monelle, Raymond (1991), ‘Music and the Peircean trichotomies’, International Review of the

Aesthetics and Sociology of Music, 22, 1, pp. 99-108.

Moretti, Franco (2005), Graphs, maps, trees: abstract models for a literary history, London; New

York (NY): Verso.

297



Moriarty, Sandra (1994), ‘Visual communication as a primary system’, Journal of Visual Literacy,

14, 2, pp. 11-21.

Muzerelle, Denis (1985), Vocabulaire codicologique: répertoire méthodique des termes français

relatifs aux manuscritsRubricae: histoire du livre et des textes, Paris: Éditions CEMI.

Muzerelle, Denis (2002-2003), Vocabulaire codicologique: répertoire méthodique des termes français

relatifs aux manuscrits avec leurs équivalents en anglais, italien, espagnol. Edition hypertextuelle,

Version 1.1, http://vocabulaire.irht.cnrs.fr/ (Accessed 2011-12-29).

Müller, Pascal, Wonka, Peter, Haegler, Simon, Ulmer, Andreas, Van Gool, Luc (2006),

‘Procedural modeling of buildings’, in ACM SIGGRAPH 2006 Papers, SIGGRAPH 2006, New

York (NY): ACM, pp. 614-623.

Neisser, Ulric (1967/2014), Cognitive Psychology, New York (NY): Psychology Press.

Nishimoto, Takehiko, Ueda, Takashi, Miyawaki, Kaori, Une, Yuko, Takahashi, Masaru (2010),

‘A normative set of 98 pairs of nonsensical pictures (droodles)’, Behavior Research Methods,

42, 3, pp. 685-691.

Noel, William (1995), ‘The division of work in the Harley Psalter’, in Brownrigg, Linda L. (Ed.),

Making the medieval book: techniques of production, Proceedings of the fourth conference of the

seminar in the history of the book to 1500, Oxford, July 1992, Los Altos Hills (CA); London:

Anderson-Lovelace; Red Gull Press, pp. 1-15.

Nordmann, Alfred (2002), ‘Another new Wittgenstein: the scientific and engineering background

of the Tractatus’, Perspectives on Science, 10, 3, pp. 356-384.

Ogle, William (Ed.) (1882), Aristotle: On the parts of animals, translated by Ogle, William,

London: K. Paul, Trench & Co.

Online Etymology Dictionary (2014), diagram, n., http://www.etymon¬

line.com/index.php?term=diagram (Accessed 2014-08-09).

Oxford English Dictionary (2014a), diagram (2), n., http://www.oed.com/view/Entry/51854

(Accessed 2014-08-09).

Oxford English Dictionary (2014b), listel, n., http://www.oed.com/view/Entry/109006 (Accessed

2014-09-05).

Oxford English Dictionary (2014c), morpheme, n., http://www.oed.com/view/Entry/122336

(Accessed 2014-08-03).

Oxford English Dictionary (2014d), percept, n., http://www.oed.com/view/Entry/140554

(Accessed 2014-08-03).

298



Oxford English Dictionary (2014e), phoneme, n., http://www.oed.com/view/Entry/142626

(Accessed 2014-08-03).

Oxford English Dictionary (2014f), proposition, n., http://www.oed.com/view/Entry/152801

(Accessed 2014-08-03).

Oxford English Dictionary (2014g), sememe, n., http://www.oed.com/view/Entry/175577

(Accessed 2014-08-03).

Oxford English Dictionary (2015a), verbal, adj., http://www.oed.com/view/Entry/222360

(Accessed 2015-05-05).

Oxford English Dictionary (2015b), visual, adj., http://www.oed.com/view/Entry/224004

(Accessed 2015-05-05).

Paivio, Allan (1971), Imagery and verbal processes, New York (NY): Holt, Rinehart and Winston.

Paivio, Allan (1975), ‘Imagery and long-term memory’, in Kennedy, Alan and Wilkes, Alan (Eds.),

Studies in long term memory, London: John Wiley & sons, pp. 57-85.

Paivio, Allan (1986), Mental representations: a dual coding approach, Oxford; New York (NY):

Oxford University Press; Clarendon Press.

Paivio, Allan (1991), Images in mind: the evolution of a theory, New York (NY): Harvester

Wheatsheaf.

Paivio, Allan (2007), Mind and its evolution: a dual coding theoretical approach, Mahwah (NJ): L.

Erlbaum Associates.

Paivio, Allan and Clark, J.M. (1990), ‘Static versus dynamic imagery’, in Cornoldi, Cesare and

McDaniel, Mark A. (Eds.), Imagery and cognition, New York (NY): Springer-Verlag.

Paivio, Allan and Csapo, Kalman (1973), ‘Picture superiority in free recall: imagery or dual

coding?’, Cognitive Psychology, 5, 2, pp. 176-206.

Palmer, Stephen E. (1975), ‘The effects of contextual scenes on the identification of objects’,

Memory & Cognition, 3, 5, pp. 519-526.

Palmer, Stephen E. (1999), Vision science: photons to phenomenology, Cambridge (MA): MIT

Press.

Palmer Elbridge, Betsy (1993), ‘Sewing variations outline’, in Standards, Seminar on Standards

of Excellence in Hand Bookbinding, Boston (MA): Guild of book workers.

Palmer Elbridge, Betsy and Prior, Anna (2008), ‘A sewing presentation: stitches and sewing for

bookbinding structures’, in Standards, Seminar on Standards of Excellence in Hand Bookbinding,

Toronto, Canada: Guild of book workers.

299



Pang, A. (2001), ‘Visualizing uncertainty in geo-spatial data’, Paper prepared for a committee of

the Computer Science and Telecommunications Board, National Research Council. Computer

Science Department, University of California, Santa Cruz, pp. 1-14.

Pang, Alex T., Wittenbrink, Craig M., Lodha, Suresh K. (1997), ‘Approaches to uncertainty

visualization’, The Visual Computer, 13, 8, pp. 370-390.

Pazzi, Luca (1999), ‘Implicit versus explicit characterization of complex entities and events’, Data

& Knowledge Engineering, 31, 2, pp. 115-134.

Peirce, Charles S. (1931-1935/1958), Collected papers of Charles Sanders Peirce: science and

philosophy and reviews, correspondence and bibliography, vols. 1-6, 1931-1935, edited by Charles

Hartshorne and Paul Weiss, vols. 7-8, 1958, edited by Arthur W. Burks, Cambridge (MA):

Harvard University Press.

Peirce, Charles S. (1992-1998), The essential Peirce: selected philosophical writings, edited by

Houser, Nathan, Kloesel, Christian J. W., Peirce Edition Project, Bloomington (IN): Indiana

University Press.

Petersen, Theodore C. (1948), Coptic bindings in the Pierpont Morgan Library, unpublished

manuscript in the Pierpont Morgan Library, New York (NY).

Petersen, Theodore C. (1954), ‘Early Islamic bookbindings and their Coptic relations’, Ars

Orientalis, 1, pp. 41-64.

Petherbridge, Guy (1987), ‘Introduction’, in Petherbridge, Guy (Ed.), Conservation of library

and archive materials and the graphic arts, London: Butterworths.

Petrucci, Armando (2001), La descrizione del manoscritto: storia, problemi, modelli, Roma: Carocci

Editore.

Pettitt, Paul and Pike, Alistair (2007), ‘Dating European palaeolithic cave art: progress, prospects,

problems’, Journal of Archaeological Method and Theory, 14, 1, pp. 27-47.

Pickwoad, Nicholas (1995), ‘The interpretation of bookbinding structure: an examination of

sixteenth-century bindings in the Ramey Collection in the Pierpont Morgan Library’, The

Library, 17, 3, pp. 209-249.

Pickwoad, Nicholas (2000), ‘Tacketed bindings: a hundred years of European bookbinding’, in

Foot, Mirjam and Pearson, David (Eds.), "For the Love of the Binding": studies in bookbinding

history presented to Mirjam Foot, London; New Castle (DE): British Library; Oak KnollPress,

pp. 119-167.

Pickwoad, Nicholas (2004), ‘The condition survey of the manuscripts in the monastery of Saint

Catherine on Mount Sinai’, The Paper Conservator, 28, 1, pp. 33-61.

300



Pickwoad, Nicholas (2008), ‘Reading Bindings: Bindings as evidence of the culture and business

of books’, in Panizzi Lectures, London: British Library.

Pickwoad, Nicholas (2011a), ‘Library or Museum? The Future of Rare Book Collections and its

Consequences for Conservation and Access’, in New approaches to book and paper

conservation-restoration, Horn/Wien: Berger.

Pickwoad, Nicholas (2011b), ‘The Place of Bindings in the Study of the Book’, Sinaiticus - Bullettin

of the Saint Catherine Foundation, pp. 2-3.

Pickwoad, Nicholas (2012), ‘An Unused Resource: Bringing the Study of Bookbindings out of

the Ghetto’, in Mouren, Raphaële (Ed.), Ambassadors of the book: competences and training

for heritage librarians, Berlin: De Gruyter Saur, pp. 83-94.

Pickwoad, Nicholas (2014b), ‘The evidence of the forged SNML Sammelband book structure’,

in Bredekamp, Horst, Brückle, Irene, Needham, Paul (Eds.), A Galileo forgery: unmasking the

New York Sidereus Nuncius, Berlin: De Gruyter, pp. 61-70.

Pickwoad, Nicholas (2014a), ‘Binders’ Gatherings’, lbt The Library: The Transactions of the

Bibliographical Society, 15, 1, pp. 63-78.

Pickwoad, Nicholas and Gullick, Michael (2004), Assessment manual. A guide to the survey forms

to be used in St Catherine’s Monastery, http://www.ligatus.org.uk/sites/default/files/manu¬

al20050110.pdf (Accessed 2012-04-20).

Pickwoad, Nicholas and McKitterick, Rosamond (2013), ‘A Carolingian manuscript fragment

from the ninth century in Amsterdam University Library, used as the binding for ‘Band 1 e

22’’, Quaerendo, 43, 3, pp. 185-213.

Pietroski, Paul (2014), ‘Logical Form’, in Zalta, Edward N. (Ed.), The Stanford Encyclopedia of

Philosophy, Fall 2014, Stanford (CA): Stanford University.

Pinker, Steven (1990), ‘A theory of graph comprehension’, in Freedle, Roy (Ed.), Artificial

intelligence and the future of testing, Hillsdale (NJ): Lawrence Erlbaum Associates, pp. 73-126.

Pitt, David (2012), ‘Mental Representation’, in Zalta, Edward N. (Ed.), The Stanford Encyclopedia

of Philosophy, Winter 2012 edition, Stanford (CA): Stanford University, http://plato.stan¬

ford.edu/archives/win2012/entries/mental-representation (Accessed 2013-05-13).

Pizlo, Zygmunt (2008), 3D shape: its unique place in visual perception, Cambridge (MA): MIT

Press.

Poblome, Jeroen and Theelen, Joeri (2014), ‘The digital excavation: best practices and lesson

learnt in the field’, in , Digital Humanities Summer School, University of Leuven, Leuven: .

Polidoro, Piero (2008), Che cos’è la semiotica visiva, Roma: Carocci Editore.

301



Pollard, Graham (1956), ‘Changes in the Style of Bookbinding, 1550–1830’, The Library, s5-XI,

2, pp. 71-94.

Pollard, Graham (1962), ‘The Construction of English Twelfth-Century Bindings’, The Library,

s5-XVII, 1, pp. 1-22.

Pollard, Graham and Potter, Esther (1984), Early Bookbinding Manuals: An Annotated List of

Technical Accounts of Bookbinding to 1840Occasional pubication / Oxford Bibliographical Society,

Oxford: Oxford Bibliographical Society.

Polyak, Stephen (1958), Vertebrate visual system, Chicago (IL): University of Chicago Press.

Porter, Dot (2013), ‘Visualizations of TEI Ms Descriptions’, forum post, 2013,

https://listserv.brown.edu/archives/cgi-bin/wa?A2=tei-l;775d4091.1307 (Accessed 2014-07-27).

Porter, Dot, Emery, Doug, Campagnolo, Alberto (2014), The Manuscript Collation Project,

http://dorpdev.library.upenn.edu/collation/ (Accessed 2014-07-27).

Price, Roger (1959), Droodles, London: Transworld Publishers.

Price, Roger (1976), Droodles #1, Los Angeles (CA): Price/Sterne/Sloan.

Pugliese, Silvia (2001), ‘Stiff-board vellum binding with slotted spine’, Papier Restaurierung.

Mitteilingen der IADA, 2, Jahrbuch / supplement, pp. 93-101.

Pylyshyn, Zenon W. (1976), Imagery and artificial intelligence, London (ON): University of

Western Ontario.

Pylyshyn, Zenon W. (1977), ‘What the mind’s eye tells the mind’s brain: a critique of mental

imagery’, in Nicholas, John M. (Ed.), Images, perception, and knowledge, Philosophy of Science

Workshop at Ontario, Canada, May 1974, Ontario: Springer Netherlands, pp. 1-36.

Pylyshyn, Zenon W. (1979), ‘The rate of “mental rotation” of images: a test of a holistic analogue

hypothesis’, Memory & cognition, 7, 1, pp. 19-28.

Pylyshyn, Zenon W. (1981), ‘The imagery debate: analog media vs. tacit knowledge’, Psychological

Review, 88, December, pp. 16-45.

Pylyshyn, Zenon W. (2003), Seeing and visualizing: it’s not what you think, Cambridge (MA):

MIT Press.

Pylyshyn, Zenon W. (2011), Things and places: how the mind connects with the world, Cambridge

(MA): MIT Press.

Rains, Mike, Clarke, Amanda, Fulford, Mike (2014), IADB - Integrated Archaeological Database

system, http://www.iadb.org.uk/ (Accessed 2014-07-08).

302



Ramsay, Stephen (2004), ‘Databases’, in Schreibman, Susan, Siemens, Ray, Unsworth, John

(Eds.), A companion to digital humanities, Oxford: Blackwell Publishing Ltd, pp. 20-30.

Rausch, Edwin (1966), ‘Das Eigenschaftsproblem in der Gestalttheorie der Wahrnehmung’, in

Metzger, Wolfgang and Erke, H. (Eds.), Handbuch der Psychologie. Wahrnehmung und

Bewußtsein, Göttingen: Verlag für Psychologie; Hogrefe, pp. 866-953.

Ravenberg, Heather (2012), A data model to describe book conservation treatment activity, MPhil

Thesis, University of the Arts London, London.

Reason, James (1990), Human Error, Cambridge: Cambridge University Press.

Regemorter, Berthe van (1992), Binding structures in the middle ages: a selection of studiesStudia

Bibliothecae Wittockianae, Brussels; London: Bibliotheca Wittockiana; Maggs Bros.

Roberts, Matt and Etherington, Don (1982/1994), Bookbinding and the conservation of books:

a dictionary of descriptive terminology, electronic edition, edited by Henry, Walter, Stanford

(CA): Preservation Department, Stanford University Libraries.

Robinson, James M. (Ed.) (1972-1977), The facsimile edition of the Nag Hammadi codices, Leiden:

Brill.

Robinson, James M. (Ed.) (1984), The facsimile edition of the Nag Hammadi codices. Introduction,

Leiden: Brill.

Rosch, Eleanor (1973), ‘On the internal structure of perceptual and semantic categories’, in

Moore, T.E. (Ed.), Cognitive Development and the Acquisition of Language, New York (NY):

Academic Press, pp. 111-144.

Rosch, Eleanor (1975), ‘Cognitive reference points’, Cognitive psychology, 7, 4, pp. 532-547.

Rosch, Eleanor (1977), ‘Human categorization’, in Warren, Neil (Ed.), Studies in cross-cultural

psychology, London: Academic Press.

Rosch, Eleanor (1978), ‘Principles of categorization’, in Rosch, Eleanor and Lloyd, Barbara

(Eds.), Cognition and categorization, New York (NY): John Wiley & Sons Inc, pp. 27-48.

Rosch, Eleanor and Mervis, Carolyn B. (1975), ‘Family resemblances: studies in the internal

structure of categories’, Cognitive Psychology, 7, 4, pp. 573-605.

Rosch, Eleanor, Mervis, Carolyn B., Gray, Wayne, Johnson, David, Braem, Penny (1976), ‘Basic

objects in natural categories’, Cognitive Psychology, 8, 3, pp. 382-439.

Rose, Gillian (2012), Visual methodologies: an introduction to researching with visual materials,

London; Thousand Oaks (CA): SAGE.

303



Rosen, Joe (2008), Symmetry rules: how science and nature are founded on symmetry, Berlin;

Heidelberg: Springer.

Rothwell, J. Dan (2012), In the company of others: an introduction to communication, Oxford;

New York (NY): Oxford University Press.

Rouse, William B. and Rouse, Sandra H. (1983), ‘Analysis and classification of human error’,

IEEE Transactions on Systems, Man and Cybernetics, SMC-13, 4, pp. 539-549.

Roussou, Maria and Drettakis, George (2003), ‘Photorealism and non-photorealism in virtual

heritage representation’, in Proceedings of the 4th International Conference on Virtual Reality,

Archaeology and Intelligent Cultural Heritage, Aire-la-Ville, Switzerland: Eurographics

Association, pp. 51-60.

Rubin, Edgar (1915), Synsoplevede Figurer: studier i psykologisk analyse, København, og Kristiania:

Gyldendal, Nordisk forlag.

Russell, Bertrand (1914), Our knowledge of the external world as a field for scientific method in

philosophy, London; New York (NY): Routledge.

Ryan, N. (1996), ‘Computer-Based Visualisation of the Past: Technical ‘Realism’ and Historical

Credibility’, in Higgins, T., Main, P., Lang, L. (Eds.), Imaging the Past: Electronic Imaging and

Computer Graphics in Museums and Archaeology, London: British Museum Press, pp. 95-108.

Ryan, Marie-Laure (2014), ‘Story/World/Media. Tuning the instruments of a media-conscious

narratology’, in Ryan, Marie-Laure and Thon, Jan-Noël (Eds.), Storyworlds across media.

Toward a media-conscious narratology, Lincoln (NE); London: University of Nebraska Press,

pp. 25-49.

Saaty, T.L. and Ozdemir, M.S. (2003), ‘Why the magic number seven plus or minus two’,

Mathematical and Computer Modelling, 38, 3-4, pp. 233-244.

Saint-Martin, Fernande (1990), Semiotics of visual language, Bloomington (IN); Indianapolis

(IN): Indiana University Press.

Saldana, Marie (2014), ‘An integrated approach to the procedural modeling of ancient cities and

buildings’, in Digital Humanities 2014, the 25th joint international conference of the Association

for Literary and Linguistic Computing and Association for Computers and the Humanities and

the 6th joint international conference of the Alliance of Digital Humanities Organizations,

Conference Abstracts, EPFL-UNIL University of Lausanne, 7-12 July 2014, Digital Humanities

2014, University of Lausanne: EPFL-UNIL University of Lausanne, pp. 342-343.

Salthe, Stanley N. (1985), Evolving hierarchical systems. Their structure and representation, New

York: Columbia University Press.

304



Sandywell, Dr Barry (2012), Dictionary of visual discourse. A dialectical lexicon of terms, Aldershot;

Burlington (VT): Ashgate.

Sayim, Bilge and Cavanagh, Patrick (2011), ‘What line drawings reveal about the visual brain’,

Frontiers in Human Neuroscience, 5, 118, pp. 1-4.

Schiano, Diane J. and Tversky, Barbara (1992), ‘Structure and strategy in encoding simplified

graphs’, Memory & Cognition, 20, 1, pp. 12-20.

Schmidt-Künsemüller, Friedrich-Adolf (1987), Bibliographie Zur Geschichte Der Einbandkunst

Von Den Anfängen Bis 1985, Wiesbaden: L. Reichert.

Sebeok, Thomas A (1991), A sign is just a sign, Bloomington: Indiana University Press.

Sebeok, Thomas (1994), Signs. An introduction to semiotics, Toronto; Buffalo (TX): University of

Toronto Press, Scholarly Publishing Division.

Shah, Priti and Miyake, Akira (Eds.) (2005), The Cambridge handbook of visuospatial thinking,

Cambridge: Cambridge University Press.

Shannon, Claude E. (1948a), ‘A mathematical theory of communication’, Bell System Technical

Journal, 27, 4, pp. 623-656.

Shannon, Claude E. (1948b), ‘A mathematical theory of communication’, Bell System Technical

Journal, 27, 3, pp. 379-423.

Sharpe, John L. (Ed.) (1996), Roger Powell, the Compleat Binder: Liber Amicorum Bibliologia,

Turnhout: Brepols.

Sharpe, John L. (2000), ‘Observations on data collection: drawing and recording information’,

in Misiti, Maria Cristina and Fondazione per la Conservazione e il Restauro dei Beni

Librari Spoleto (Eds.), Contributi e testimonianze, Spoleto: Accademia Spoletina, pp. 103-133.

Shen, Dandan and Walther, Dirk B. (2012), ‘Categorization of line drawings of natural scenes

using non-accidental properties matches human behavior’, Journal of Vision, 12, 9, pp. 593-640.

Sheppard, Jennifer M. (1995), ‘Some twelfth-century monastic bindings and the question of

localization’, in Brownrigg, Linda L. (Ed.), Making the medieval book: techniques of production,

Proceedings of the fourth conference of the seminar in the history of the book to 1500, Oxford,

July 1992, Oxford: Anderson-Lovelace; Red Gull Press, pp. 181-198.

Shiffrin, Richard M. and Schneider, Walter (1977), ‘Controlled and automatic human information

processing: II. Perceptual learning, automatic attending and a general theory’, Psychological

Review, 84, 2, pp. 127-190.

Shin, Sun-Joo, Lemon, Oliver, Mumma, John (2013), ‘Diagrams’, in Zalta, Edward N. (Ed.), The

Stanford Encyclopedia of Philosophy, Fall 2013 Edition, Stanford (CA): Stanford University.

305



Short, Thomas L. (2007), Peirce’s theory of signs, Cambridge; New York (NY): Cambridge

University Press.

Simons, Peter (1987), Parts: a study in ontology, Oxford; New York (NY): Oxford University

Press.

Smith, E.E., Bolzano, G.J., Walker, J. (1978), ‘Nominal, perceptual, and semantic codes in

picture categorization’, in Cotton, John W. and Klatzky, Roberta L. (Eds.), Semantic Factors

in Cognition, Hillsdale (NJ): Erlbaum, pp. 137-167.

Snyder, Joel (1980), ‘Picturing vision’, Critical Inquiry, 6, 3, pp. 499-526.

Sperberg-McQueen, C.M. and Burnard, L. (Eds.) (2005), Text Encoding Iniziative P5: guidelines

for electronic text encoding and interchange, P5 revised and re-edited edition, Oxford; Providence

(RI); Charlottesville (VA); Nancy (KY): Text Encoding Initiative Consortium.

Spitzmueller, Pamela (1982), ‘A trial terminology for describing sewing through the fold’, The

Paper Conservator, 7, 1, pp. 44-46.

Spitzmueller, Pamela and Frost, Gary (1982), ‘A trial terminology for sewing through the fold’,

American Institute for Conservation of Historic and Artistic Works, Book and Paper Group, 10th

Annual Meeting Postprints, 1, p. 3.

Springer, Sally P. and Deutsch, George (1997), Left brain, right brain: perspectives from cognitive

neuroscience, New York (NY): Freeman.

Stenning, Keith and Lemon, Oliver (1999), ‘Aligning logical and psychological perspectives on

diagrammatic reasoning’, Artificial Intelligence Review, 15, pp. 29-62.

Sternberg, Robert J. and Mio, Jeffery Scott (2009), Cognitive psychology, fifth edition, Australia;

Belmont (CA): Cengage Learning; Wadsworth.

Stilla, Uwe and Michaelsen, Eckart (1997), ‘Semantic modelling of man-made objects by

production nets’, in Gruen, Armin, Baltsavias, Emmanuel P., Henricsson, Olof (Eds.),

Automatic extraction of man-made objects from aerial and space images (II), Birkhäuser Basel:

Springer, pp. 43-52.

Stiny, G. (1980), ‘Introduction to shape and shape grammars’, Environment and Planning B:

Planning and Design, 7, 3, p. 343 – 351.

Stiny, George (2006), Shape: talking about seeing and doing, Cambridge (MA): MIT Press.

Stiny, George and Gips, James (1972), ‘Shape grammars and the generative specification of

painting and sculpture’, in Freiman, C. V. (Ed.), Information Processing 71, IFIP Congress71,

Ljubljana, Yugoslavia: Amsterdam: North- Holland, pp. 1460-1465.

306



Stoutland, Frederick (1998), ‘Wittgenstein: on certainty and truth’, Philosophical Investigations,

21, 3, pp. 203-221.

Strothotte, Thomas, Masuch, Maic, Isenberg, Tobias (1999), ‘Visualizing knowledge about

virtual reconstructions of ancient architecture’, in Proceedings of the International Conference

on Computer Graphics, Washington (DC): IEEE Computer Society, pp. 36-43.

Strothotte, Thomas, Preim, Bernhard, Raab, Andreas, Schumann, Jutta, Forsey, David R. (1994),

‘How to render frames and influence people’, Computer Graphics Forum, 13, 3, pp. 455-466.

Strothotte, Thomas, Puhle, Matthias, Masuch, Maic, Freudenberg, Bert, Kreiker, Sebastian,

Ludowici, Babette (1999), ‘Visualizing uncertainty in virtual reconstructions’, in Proceedings

of Electronic Imaging & the visual Arts, Berlin, EVA Europe 1999, Berlin: Vasari, GFaI, pp. 16.

Strothotte, Thomas and Schlechtweg, Stefan (2002), Non-photorealistic computer graphics.

Modeling, rendering, and animation, San Francisco (CA): Morgan Kaufmann Publishers Inc..

Szczepanowska, Hanna M. (2013), Conservation of Cultural Heritage: Key Principles and Approaches,

London; New York (NY): Routledge.

Szirmai, János A. (1988), ‘The girdle book of the museum Meermanno-Westreenianum’, Quaerendo,

18, 1, pp. 17-34.

Szirmai, János A. (1996), ‘The archaeology of bookbinding and book restoration’, Quaerendo, 26,

2, pp. 144-164.

Szirmai, János A. (1999), The archaeology of medieval bookbinding, Aldershot; Burlington (VT):

Ashgate.

Talmy, Leonard (2000), ‘How language structures space’, in Toward a cognitive semantics. Concept

structuring systems, Cambridge (MA): MIT Press, pp. 177-254.

Tanselle, G. Thomas (2002), Introduction to bibliography: seminar syllabus for English/Comparative

Literature, Columbia University, Charlottesville (VA): Book Arts Press; University of Virginia.

Terras, Melissa M. (2008), Digital images for the information professional, Aldershot; Burlington

(VT): Ashgate Publishing Limited.

Text Encoding Initiative Consortium (2014a), 10 Manuscript description. TEI P5 Guidelines,

http://www.tei-c.org/release/doc/tei-p5-doc/en/html/MS.html (Accessed 2014-07-27).

Text Encoding Initiative Consortium (2014b), TEI, http://www.tei-c.org/index.xml (Accessed

2014-07-27).

Thomas, Nigel J.T. (2013), ‘Mental imagery’, in Zalta, Edward N. (Ed.), The Stanford Encyclopedia

of Philosophy, Spring 2013 edition, Stanford (CA): Stanford University, http://plato.stan¬

ford.edu/archives/spr2013/entries/mental-imagery (Accessed 2013-05-13).

307



Thomson, J., Hetzler, E., MacEachren, A., Gahegan, M., Pavel, M. (2005), ‘A typology for

visualizing uncertainty’, in Proceedings of Visualization and Data Analysis, IS&T/SPIE Symposium

on Electronic Imaging, 16-20 January 2005, San Jose (CA): , pp. 146-157.

Thouless, Robert H. (1931a), ‘Phenomenal regression to the real object I’, British Journal of

Psychology, 21, pp. 339-359.

Thouless, Robert H. (1931b), ‘Phenomenal regression to the real object II’, British Journal of

Psychology, 22, pp. 1-30.

Toth, Mike, Bhayro, Siam, Emery, Doug (2010), Syriac palimpsest digital release README

document, http://digitalgalen.net/0_ReadMe.html (Accessed 2014-07-27).

Toth, Mike, Bhayro, Siam, Emery, Doug (2013), The digital Galen Syriac palimpsest,

http://digitalgalen.net/ (Accessed 2014-07-27).

Tufte, Edward R. (1990), Envisioning information, Cheshire (CT): Graphics Press.

Tufte, Edward R. (1997), Visual explanations: images and quantities, evidence and narrative,

Cheshire (CT): Graphics Press.

Tufte, Edward R. (2001), The visual display of quantitative information, Second Edition, Cheshire

(CT): Graphics Press.

Tufte, Edward R. (2006), Beautiful evidence, Cheshire (CT): Graphics Press.

Tversky, Barbara (1989), ‘Parts, partonomies, and taxonomies’, Developmental Psychology, 25,

6, pp. 983-995.

Tversky, Barbara (1991), ‘Distortions in memory for visual displays’, in Ellis, S.R., Kaiser, M.K.,

Grunwald, A. (Eds.), Pictorial communication in virtual and real environments, London: Taylor

& Francis, pp. 61-75.

Tversky, Barbara (1997), ‘Memory for pictures, maps, environments and graphs’, in Payne, D.

and Conrad, F. (Eds.), Intersections in Basic and Applied Memory Research, Mahwah (NJ):

Erlbaum, pp. 257-277.

Tversky, Barbara (2001), ‘Spatial schemas in depictions’, in Gattis, Merideth (Ed.), Spatial schemas

and abstract thought, Cambridge (MA): MIT Press, pp. 79-111.

Tversky, Barbara (2002), ‘What do sketches say about thinking?’, in Stahovic, T., Landay, J.,

Davis, R. (Eds.), Proceedings of AAAI spring symposium on sketch understanding, Menlo Park

(CA): AAAI Press, pp. 148-151.

Tversky, Barbara (2005c), ‘Some ways images express and promote thought’, in Grialou, P.,

Longo, G., Okada, M. (Eds.), Images and Reasoning, Tokyo: Keio University, pp. 15-29.

308



Tversky, Barbara (2005a), ‘Functional significance of visuospatial representations’, in Shah, Priti

and Miyake, Akira (Eds.), The Cambridge handbook of visuospatial thinking, Cambridge; New

York (NY): Cambridge University Press, pp. 1-34.

Tversky, Barbara (2005b), ‘Prolegomenon to scientific visualizations’, in Gilbert, John K. (Ed.),

Visualization in Science Education, Dordrecht: Springer Netherlands, pp. 29-42.

Tversky, Barbara (2006), ‘Visuospatial reasoning’, in Holyoak, Keith James (Ed.), The Cambridge

handbook of thinking and reasoning, Cambridge; New York (NY): Cambridge University Press,

pp. 209-240.

Tversky, Barbara (2007), ‘Gestalts of thought’, in Albertazzi, Liliana (Ed.), Visual thought: the

depictive space of perception, Amsterdam: Benjamins, pp. 155-163.

Tversky, Barbara, Agrawala, Maneesh, Heiser, Julie, Lee, Paul, Hanrahan, Pat, Phan, Doantam,

Stolte, Chris, Daniel, Marie-Paule (2007), ‘Cognitive design principles for automated generation

of visualizations’, in Allen, Gary L. (Ed.), Applied spatial cognition from research to cognitive

technology, London; New York (NY): Routledge, pp. 53-74.

Tversky, Barbara and Hemenway, Kathleen (1984), ‘Objects, parts, and categories’, Journal of

Experimental Psychology. General, 113, 2, pp. 169-197.

Tversky, Barbara, Zacks, Jeff, Lee, Paul, Heiser, Julie (2000), ‘Lines, blobs, crosses and arrows:

diagrammatic communication with schematic figures’, in Anderson, Michael, Cheng, Peter,

Haarslev, Volker (Eds.), Theory and application of diagrams, Berlin; Heidelberg:

Springer-Verlag, pp. 221-230.

Tyler, Andrea and Evans, Vyvyan (2003), The semantics of English prepositions. Spatial scenes,

embodied meaning, and cognition, Cambridge; New York (NY): Cambridge University Press.

Ullman, David G (2010), The mechanical design process, 4th edition, international edition, Boston

(MA): McGraw-Hill.

Van Valkenburg, David and Kubovy, Michael (2003), ‘In Defense Of The Theory Of Indispensable

Attributes’, Cognition, 87, pp. 225-233.

Varela, Francisco J., Thompson, Evan T., Rosch, Eleanor (1991), The embodied mind: cognitive

science and human experience, Cambridge (MA): MIT Press.

Varzi, Achille (2014), ‘Mereology’, in Zalta, Edward N. (Ed.), The Stanford Encyclopedia of

Philosophy, Fall 2014 Edition, Stanford (CA): Stanford University, http://plato.stan¬

ford.edu/archives/fall2014/entries/mereology (Accessed 2014-09-09).

Velios, Athanasios (2008), ‘Hierarchical recording of binding structures’, in The Book and Paper

Group Annual, American Institute for Conservation of Historic & Artistic Works 36th Annual

309



Meeting, Denver, CO: American Institute for Conservation of Historic & Artistic Works, pp.

97.

Velios, Athanasios (2014), ‘Beyond databases. Linked open data for bookbinding descriptions’,

in Men and books, From microorganisms to mega-organisms, St. Pölten, Austria: Archive of the

Diocese of St. Pölten in cooperation with ICARUS (International Centre for Archival Research)

and the ERC (European Research Centre for Book and Paper Conservation).

Velios, Athanasios and Pickwoad, Nicholas (2004), ‘Collecting Digital Data on Paper’, in Second

international conference of museology - Technology for cultural heritage, 28th June - 2nd July

2004, Mitilini: Καλειδοσκόπιο, pp. 650-658.

Velios, Athanasios and Pickwoad, Nicholas (2005a), ‘Current use and future development of the

database of the St Catherine’s Library Conservation Project’, Paper Conservator, 29, pp. 39-53.

Velios, Athanasios and Pickwoad, Nicholas (2005b), ‘The Database of the St Catherine’s Library

Conservation Project in Sinai, Egypt’, in , IS&T Archiving 2005 Conference, April 26-29,

Washington (DC): Society for Imaging Science and Technology, pp. 73-78.

Velios, Athanasios and Pickwoad, Nicholas (2008), ‘Collecting and managing conservation survey

data’, in Fellows-Jensen, Gillian and Springborg, Peter (Eds.), Care and conservation of

manuscripts 10: proceedings of the tenth international seminar held at the University of

Copenhagen, 19th-20th October 2006, Care and conservation of manuscripts 10, Copenhagen:

Museum Tusculanum Press; University of Copenhagen, pp. 172-188.

Velios, Athanasios and Pickwoad, Nicholas (2009), ‘An optimised workflow for large-scale

condition surveys of book collections’, in Driscoll, Matthew James and Mósesdóttir,

Ragnheidur (Eds.), Care and conservation of manuscripts 11: proceedings of the eleventh

international seminar held at the University of Copenhagen 24th-25th April 2009, Care and

conservation of manuscripts 11, Copenhagen: Museum Tusculanum Press, University of

Copenhagen, pp. 269-290.

Velios, Athanasios and Pickwoad, Nicholas (2012), ‘The digitization of bookbindings’, in Nelson,

Brent and Terras, Melissa M. (Eds.), Digitizing medieval and early modern material culture,

Tempe (AZ): Iter: Gateway to the Middle Ages and Renaiisance Toronto, Ontario; Arizona

Center for Medieval and Renaissance Studies (ACMRS), pp. 201-228.

W3C (2003), XML development history, http://www.w3.org/XML/hist2002 (Accessed 2015-04-25).

W3C (2010), Secret origin of SVG, http://www.w3.org/Graphics/SVG/WG/wiki/Secret_Ori¬

gin_of_SVG (Accessed 2015-04-25).

W3C (2015), Extensible Markup Language (XML), http://www.w3.org/XML/ (Accessed

2015-04-25).

310



Wainer, Howard (2009), Picturing the uncertain world. How to understand, communicate, and

control uncertainty through graphical display, Princeton (NJ): Princeton University Press,.

Wakeman, Geoffrey (1993), Functional Developments in Bookbinding, Kidlington, Oxford ; New

Castle, Delaware: Plough Press.

Walters Art Museum (2014a), Describing manuscripts with TEI - Collation, http://thedigitalwal¬

ters.org/Supplemental/ManuscriptDescription.html#collation (Accessed 2014-07-27).

Walters Art Museum (2014b), The Digital Walters, http://www.thedigitalwalters.org/ (Accessed

2014-07-27).

Walther, Dirk B., Chai, Barry, Caddigan, Eamon, Beck, Diane M., Fei-Fei, Li (2011), ‘Simple

line drawings suffice for functional MRI decoding of natural scene categories’, Proceedings of

the National Academy of Sciences, 108, 23, pp. 9661-9666.

Ware, Colin (2008), Visual thinking for design, Amsterdam: Morgan Kaufmann.

Ware, Colin (2013), Information visualization: perception for design, Amsterdam: Morgan Kaufmann.

Warwick, Claire, Fisher, Claire, Terras, Melissa, Baker, Mark, Clarke, Amanda, Fulford, Mike,

Grove, Matt, O'Riordan, Emma, Rains, Mike (2009), ‘iTrench: A study of user reactions to

the use of information technology in field archaeology’, Literary and Linguistic Computing, 24,

2, pp. 211-223.

Weale, W.H. James (1898), Bookbindings and rubbings of bindings in the National Art Library,

South Kensington Museum, London: Printed for H. M. Stationery Off., by Eyre and

Spottiswoode.

Webopedia ([s.d.]), Markup language, http://www.webopedia.com/TERM/M/markup_lan¬

guage.html (Accessed 2014-03-22).

Webster, Graham (1964), Romano-British coarse pottery: a student’s guide, York: Council for

British Archaeology, http://archaeologydataservice.ac.uk/catalogue/adsdata/arch-281-1/dis¬

semination/pdf/cba_rr_006.pdf (Accessed 2011-12-15).

Weisstein, E. W. ([s.d.]), Permutation. From MathWorld - A Wolfram Web Resource,

http://mathworld.wolfram.com/Permutation.html (Accessed 2014-03-19).

Weyl, Hermann (1952), Symmetry, Princeton (NJ): Princeton University Press.

Wilkinson, Leland (2005), ‘Uncertainty’, in The Grammar Of Graphics, Berlin; Heidelberg:

Springer, pp. 451-488.

Willard, William Franklin (2009), The art of mechanical drawing: a practical course for drafting

and design, New York (NY): Hearst Books.

311



Winston, P. H. (1975), ‘Learning structural descriptions from examples’, in Winston, D. H.

(Ed.), The psychology of computer vision, New York (NY): McGraw-Hill.

Wiseman, Sandor, MacLeod, Colin M., Lootsteen, P.J. (1985), ‘Picture recognition improves

with subsequent verbal information’, Journal of experimental psychology. Learning, memory,

and cognition, 11, 3, pp. 588-595.

Wittgenstein, Ludwig (1922/2012), Tractatus Logico-Philosophicus, Side-by-side edition, version

0.29 (November 17, 2012), containing the original German, alongside both the Ogden/Ramsey,

and Pears/McGuinness English translations, Amherst (MA): University of Massachusetts,

http://people.umass.edu/phil335-klement-2/tlp/tlp.pdf (Accessed 2013-05-13).

Wittgenstein, Ludwig (1967), Philosophical investigations, third edition, translated by Anscombe,

G. E. M., Basil: Blackwell Publishing Ltd.

Wittgenstein, Ludwig (1969), On certainty, edited by Anscombe, G. E. M. and Wright, G. H.

von, Oxford: Blackwell.

Wood, Jo, Isenberg, Petra, Isenberg, Tobias, Dykes, Jason, Boukhelifa, Nadia, Slingsby, Aidan

(2012), ‘Sketchy rendering for information visualization’, IEEE Transactions on Visualization

and Computer Graphics, 18, 12, pp. 2749-2758.

Wood, Phyllis and McDonnell, Patrick (1994), Scientific illustration: a guide to biological,

zoological, and medical rendering techniques, design, printing, and display, second edition, New

York (NY): John Wiley & Sons.

Wragg, Richard (2012), ‘Anybody know cheap/free software that’s good for producing a diagram

of a MS volume’s bifolia / quires / construction? #medievalistproblems’, @richdwragg twitter

post, 22 November, https://twitter.com/richdwragg/status/271614026415497216 (Accessed

2013-06-28).

Wright, Holly Ellen (2011), Seeing triple. Archaeology, field drawing and the Semantic Web, PhD

Thesis, University of York, York, http://etheses.whiterose.ac.uk/2194/1/WrightThesis.pdf

(Accessed 2014-03-05).

Wurfel, Cliff (1989), ‘Knot-tack sewing’, Binders’ Guild Newsletter, XII, 7, p. 14.

Yong, Liu, Congfu, Xu, Zhigeng, Pan, Yunhe, Pan (2006), ‘Semantic modeling for ancient

architecture of digital heritage’, Computers & Graphics, 5, pp. 800-814.

Zappella, Giuseppina (1996), Manuale del libro antico, Milano: Editrice Bibliografica.

Zhu, Xiaojin, Goldberg, Andrew B., Eldawy, Mohamed, Dyer, Charles R., Strock, Bradley

(2007), ‘A text-to-picture synthesis system for augmenting communication’, in The Integrated

Intelligence Track of the Twenty-Second AAAI Conference on Artificial Intelligence, AAAI-07,

Menlo Park (CA): AAAI Press, pp. 1590-1595.

312



Zimmermann, M. (1989), ‘The nervous system in the context of information theory’, Human

physiology, 2, pp. 166-175.

Zuk, Torre D. (2008), Visualizing uncertainty, PhD Thesis, Department of Computer Science,

University of Calgary, Calgary, Alberta, Canada, http://innovis.cpsc.ucalgary.ca/innov¬

is/uploads/Publications/Publications/Zuk_2008_Visualizing_Uncertainty.pdf (Accessed

2014-06-12).

Zuk, Torre D., Carpendale, S., Glanzman, W. D. (2005), ‘Visualizing temporal uncertainty in

3D virtual reconstructions’, in Proceedings of the 6th International Conference on Virtual Reality,

Archaeology and Intelligent Cultural Heritage, Aire-la-Ville, Switzerland, Switzerland:

Eurographics Association, pp. 99-106.

Zuk, Torre D. and Carpendale, Sheelagh (2006), ‘Theoretical analysis of uncertainty visualizations’,

in Erbacher, Robert F., Roberts, Jonathan C., Gröhn, Matti T., Börner, Katy (Eds.),

Proceedings of SPIE Conference Electronic Imaging, Visualization and Data Analysis 2006, San

Jose Marriott and San Jose Convention Center, San Jose, California, USA: SPIE.

Zuk, Torre D. and Carpendale, Sheelagh (2007), ‘Visualization of uncertainty and reasoning’, in

Butz, Andreas, Fisher, Brian, Krüger, Antonio, Olivier, Patrick, Owada, Shigeru (Eds.),

Smart Graphics, Berlin, Heidelberg: Springer-Verlag, pp. 164-177.

313



Illustration credits

Figure 1. Campagnolo 2015, after Donald 2001....................................................5
Figure 2. Campagnolo 2015....................................................................................9
Figure 3. Zhu et al. 2007......................................................................................21
Figure 4. De Luca et al. 2005...............................................................................26
Figure 5. Campagnolo 2015..................................................................................33
Figure 6. Amigos 2015, www.amigos.co.uk/Our-Philosophy.aspx.....................35
Figure 7. Campagnolo 2015, after Eco 1997........................................................37
Figure 8. Campagnolo 2015..................................................................................39
Figure 9. Campagnolo 2015..................................................................................46
Figure 10. Campagnolo 2015................................................................................47
Figure 11. Campagnolo 2015; Saint Catherine monastery, Sinai, Egypt 2007.....57
Figure 12. Petersen 1948......................................................................................72
Figure 13. Dorsey 1989; Wurfel 1989..................................................................73
Figure 14. Benvestito 2012...................................................................................74
Figure 15. Spitzmueller 1982...............................................................................75
Figure 16. Saint Catherine monastery, Sinai, Egypt 2007....................................78
Figure 17. Webster 1964; Brodribb 1970............................................................83
Figure 18. Gaskell 1972; Federici & Houlis 1988; Sheppard 1995; Clarkson 1996;
Regemorter 1992...................................................................................................84
Figure 19. Muzerelle 1985; Noel 1995; Szirmai 1999.........................................85
Figure 20. Campagnolo 2015................................................................................88
Figure 21. Szirmai 1988; Clarkson 1993; Adler 2010..........................................89
Figure 22. Petersen 1954; Clarkson 1993; Boudalis 2007....................................91
Figure 23. Clarkson 2005; Pickwoad 2000; Szirmai 1999...................................93
Figure 24. Greenfield 1998..................................................................................94
Figure 25. Bray 1607-1658, Rijksmuseum Amsterdam.......................................96
Figure 26. Szirmai 1999........................................................................................98
Figure 27. Carvin 1988..........................................................................................98
Figure 28. Greenfield 1990..................................................................................99
Figure 29. Greenfield 1990..................................................................................99
Figure 30. Bibliothèque Nationale (France) 1989.............................................100
Figure 31. Szirmai 1999......................................................................................127
Figure 32. Clarkson 2005....................................................................................131
Figure 33. Velios & Pickwoad 2004...................................................................144
Figure 34. Ligatus 2007......................................................................................146
Figure 35. Campagnolo 2015..............................................................................148
Figure 36. Campagnolo 2015..............................................................................152
Figure 37. Campagnolo 2015..............................................................................154
Figure 38. Campagnolo 2015..............................................................................155
Figure 39. Campagnolo 2015..............................................................................159
Figure 40. Willard 2009......................................................................................160
Figure 41. Campagnolo 2015..............................................................................160
Figure 42. Campagnolo 2015..............................................................................162

314



Figure 43. Szirmai 1999......................................................................................163
Figure 44. Szirmai 1999; Carvin 1988; Middleton 1996; Cockerell 1953..........164
Figure 45. Campagnolo 2015..............................................................................165
Figure 46. Jäckel 1975; Greenfield 1990; Szirmai 1999; Clarkson 2005; Conn &
Verheyen 2003b..................................................................................................167
Figure 47. Campagnolo 2015..............................................................................168
Figure 48. Campagnolo 2015..............................................................................170
Figure 49. Campagnolo 2015..............................................................................171
Figure 50. Campagnolo 2015..............................................................................171
Figure 51. Campagnolo 2015..............................................................................173
Figure 52. Campagnolo 2015..............................................................................174
Figure 53. Campagnolo 2015; Saint Catherine monastery, Sinai, Egypt 2007....175
Figure 54. Campagnolo 2015..............................................................................176
Figure 55. Campagnolo 2015..............................................................................179
Figure 56. Campagnolo 2015..............................................................................179
Figure 57. Campagnolo 2015..............................................................................180
Figure 58. Campagnolo 2015..............................................................................182
Figure 59. Campagnolo 2015, after Pickwoad & Gullick 2004.........................182
Figure 60. Szirmai 1999; Carvin 1988................................................................183
Figure 61. Middleton 1996.................................................................................184
Figure 62. Campagnolo 2015..............................................................................185
Figure 63. Campagnolo 2015..............................................................................185
Figure 64. Campagnolo 2015..............................................................................186
Figure 65. Campagnolo 2015..............................................................................186
Figure 66. Campagnolo 2015..............................................................................187
Figure 67. Campagnolo 2015..............................................................................187
Figure 68. Campagnolo 2015..............................................................................187
Figure 69. Campagnolo 2015..............................................................................188
Figure 70. Campagnolo 2015..............................................................................189
Figure 71. Campagnolo 2015..............................................................................189
Figure 72. Campagnolo 2015..............................................................................190
Figure 73. Campagnolo 2015..............................................................................190
Figure 74. Saint Catherine monastery, Sinai, Egypt 2007..................................191
Figure 75. Campagnolo 2015..............................................................................192
Figure 76. Saint Catherine monastery, Sinai, Egypt 2007..................................192
Figure 77. Campagnolo 2015..............................................................................193
Figure 78. Saint Catherine monastery, Sinai, Egypt 2007..................................193
Figure 79. Campagnolo 2015..............................................................................194
Figure 80. Campagnolo 2015..............................................................................194
Figure 81. Saint Catherine monastery, Sinai, Egypt 2007..................................195
Figure 82. Campagnolo 2015..............................................................................195
Figure 83. Saint Catherine monastery, Sinai, Egypt 2007..................................195
Figure 84. Campagnolo 2015..............................................................................195
Figure 85. Saint Catherine monastery, Sinai, Egypt 2007..................................196
Figure 86. Campagnolo 2015..............................................................................196

315



Figure 87. Saint Catherine monastery, Sinai, Egypt 2007..................................197
Figure 88. Campagnolo 2015..............................................................................197
Figure 89. Campagnolo 2015..............................................................................198
Figure 90. Campagnolo 2015..............................................................................198
Figure 91. Saint Catherine monastery, Sinai, Egypt 2007..................................199
Figure 92. Campagnolo 2015..............................................................................199
Figure 93. Campagnolo 2015..............................................................................199
Figure 94. Campagnolo 2015..............................................................................203
Figure 95. Campagnolo 2015..............................................................................203
Figure 96. Campagnolo 2015..............................................................................204
Figure 97. Campagnolo 2015..............................................................................205
Figure 98. Campagnolo 2015..............................................................................206
Figure 99. Campagnolo 2015..............................................................................207
Figure 100. Szirmai 1999; Szirmai 1996............................................................209
Figure 101. Campagnolo 2015............................................................................210
Figure 102. Campagnolo 2015............................................................................211
Figure 103. Szirmai 1999....................................................................................213
Figure 104. Campagnolo 2015............................................................................214
Figure 105. Campagnolo 2015............................................................................215
Figure 106. Campagnolo 2015............................................................................217
Figure 107. Campagnolo 2015............................................................................218
Figure 108. Campagnolo 2015............................................................................219
Figure 109. Campagnolo 2015............................................................................219
Figure 110. Campagnolo 2015............................................................................220
Figure 111. Campagnolo 2015............................................................................221
Figure 112. Campagnolo 2015............................................................................224
Figure 113. Campagnolo 2015............................................................................225
Figure 114. Campagnolo 2015............................................................................226
Figure 115. Campagnolo 2015............................................................................227
Figure 116. Campagnolo 2015............................................................................229
Figure 117. Campagnolo 2015............................................................................229
Figure 118. Campagnolo 2015............................................................................230
Figure 119. Campagnolo 2015............................................................................231
Figure 120. Campagnolo 2015............................................................................232
Figure 121. Campagnolo 2015............................................................................233
Figure 122. Campagnolo 2015............................................................................234
Figure 123. Campagnolo 2015............................................................................235
Figure 124. Saint Catherine monastery, Sinai, Egypt 2007................................237
Figure 125. Saint Catherine monastery, Sinai, Egypt 2007................................238
Figure 126. Campagnolo 2015............................................................................238
Figure 127. Campagnolo 2015; Saint Catherine monastery, Sinai, Egypt 2007..240
Figure 128. Campagnolo 2015; Saint Catherine monastery, Sinai, Egypt 2007..240
Figure 129. Campagnolo 2015; Saint Catherine monastery, Sinai, Egypt 2007..240
Figure 130. Campagnolo 2015; Saint Catherine monastery, Sinai, Egypt 2007..240
Figure 131. Campagnolo 2015............................................................................243

316



Figure 132. Campagnolo 2015............................................................................249
Figure 133. Campagnolo 2015............................................................................270
Figure 134. Schoenberg Institute for Manuscript Studies, Manuscript Collation
Project 2015........................................................................................................273
Figure 135. Campagnolo 2015............................................................................319

317


	Title page
	Abstract
	Table of contents
	Table of figures
	Table of tables
	Table of listings
	Acknowledgements
	Chapter 1. Introduction
	1.1. The cognitive importance of diagrams
	1.1.1. Diagrams as visual proofs

	1.2. Problems with the adoption of diagrams
	1.3. Verbal to visual: intermediation & intersemiotic translations
	1.3.1. The essential elements of the communication cycle

	1.4. Bookbinding descriptions and the generation of diagrams
	1.5. Terminological issues
	1.6. Aims and objectives
	1.7. Chapter overview

	Chapter 2. Modelling & related work
	2.1. Modelling
	2.1.1. Modelling of and modelling for

	2.2. From verbal to visual: related projects and approaches
	2.2.1. Hardware description languages
	2.2.2. Natural language processing
	2.2.3. Cultural heritage modelling

	2.3. From verbal to visual through models of
	Summary

	Chapter 3. Signs & material objects
	3.1. Signs, signification, and communication
	3.1.1. Signs
	3.1.2. Reading and remembering iconic signs
	3.1.2.1. Reading modalities of icons
	3.1.2.2. Formalization of hypoicons

	3.1.3. Visual texts

	3.2. Multimodal communication
	3.2.1. Dual-coding representation
	3.2.2. Verbal recoding of ambiguous visual stimuli

	3.3. Categorization and prototypification
	3.3.1. Prototypes
	3.3.2. Basic level of abstraction

	3.4. Objects: their form and their shape
	3.4.1. The form of objects
	3.4.1.1. Material components and form
	3.4.1.2. Logical form and expression

	3.4.2. The shape of objects
	3.4.2.1. The nature of shape
	3.4.2.2. Regularities for the prototypification of shape
	3.4.2.3. Line drawings
	3.4.2.4. Perceptual biases


	Summary

	Chapter 4. Bookbinding descriptions
	4.1. Books as artefacts
	4.1.1. Binding structures and decoration
	4.1.1.1. The evolution of bookbinding studies

	4.1.2. Terminology

	4.2. Verbal and visual descriptions of bindings
	4.2.1. A description of Coptic bindings
	4.2.2. A particular sewing pattern
	4.2.3. A controlled vocabulary description
	4.2.4. Structured and controlled vocabulary descriptions

	4.3. Prototypical visualization of bookbinding elements
	4.3.1. Gathering assembly diagrams
	4.3.1.1. Naturalistic representations of gatherings
	4.3.1.2. Schematic representations of gatherings


	4.4. A categorization of bookbinding illustrations
	4.4.1. Archaeological-style drawings
	4.4.2. Naturalistic drawings
	4.4.3. Line drawings
	4.4.4. Generic-shape drawings
	4.4.5. Scenes

	4.5. Clarity of information in bookbinding line drawings
	4.5.1. Sketchiness and detail views

	Summary

	Chapter 5. Communicating & translating objects
	5.1. Communicating material objects through verbal means
	5.1.1. Communicating through controlled vocabularies
	5.1.1.1. Names and prototypical material components
	5.1.1.2. Sewing structure controlled vocabulary

	5.1.2. Communicating through structured descriptions
	5.1.3. Diagrams as support for natural language descriptions

	5.2. Automated intersemiotic translations
	5.2.1. Translating material components, their form, and shape
	5.2.2. Translating material objects’ shapes
	5.2.3. Translating an object’s form
	5.2.4. The reference to the logical form

	5.3. Communicating material objects through visual means
	5.3.1. Marks on a background
	5.3.2. Continuity of the plane
	5.3.3. Visualization conventions
	5.3.3.1. Fixing meaning through a reference system
	5.3.3.1.1. Titling and verbal labelling
	5.3.3.1.2. Graphic reference elements

	5.3.3.2. Separation of information
	5.3.3.3. Use of colour
	5.3.3.4. Two-dimensional vs. three-dimensional visualizations

	5.3.4. Structured visualizations as output

	5.4. The uncertainty of reality
	5.4.1. Uncertainty in structured descriptions
	5.4.2. Uncertainty in visualizations
	5.4.2.1. Archaeological visualizations
	5.4.2.1.1. An example from archaeology of the book
	5.4.2.1.2. Uncertainty clues in archaeological visualizations

	5.4.2.2. Irregularities in visual propositions
	5.4.2.2.1. Visual cues for uncertainty visualization

	5.4.2.3. Imprecision as uncertainty
	5.4.2.4. Uncertainty inherent in the visualization process

	5.4.3. Human error in the data

	Summary

	Chapter 6. Transformation framework
	6.1. The technology and the description schema
	6.1.1. Extensible Markup Language
	6.1.2. Extensible Stylesheet Language Transformations
	6.1.3. Scalable Vector Graphics
	6.1.4. The description schema and the dataset
	6.1.4.1. The Ligatus schema for the description of bookbinding structures
	6.1.4.1.1. The manuscript collection survey
	6.1.4.1.2. The printed book collection survey
	6.1.4.1.3. Free-hand drawings

	6.1.4.2. The project dataset


	6.2. The transformations
	6.2.1. Knowledge graph schemas
	6.2.1.1. Graphic conventions for knowledge graph schemas
	6.2.1.2. Reading the visual description graphs

	6.2.2. Number of possible visualizations
	6.2.2.1. Board marker visualizations
	6.2.2.2. Endleaf structure visualizations

	6.2.3. Visualizations
	6.2.3.1. Level of abstraction and semantic entry point
	6.2.3.2. Diagram elements
	6.2.3.3. Spatial arrangement
	6.2.3.4. Thread paths
	6.2.3.5. Uncertainty
	6.2.3.6. Accuracy and imprecision of measurements
	6.2.3.7. Place-holders for uncertainty inherent in the visualization process


	Summary

	Chapter 7. Transformations & analysis
	7.1. A complete visualization example: endleaves
	7.1.1. Visual description framework
	7.1.2. Shapes: standard endleaf components
	7.1.2.1. Reference elements
	7.1.2.2. Endleaf components
	7.1.2.2.1. Pastedowns in use visualizations

	7.1.2.3. Attachment modalities
	7.1.2.3.1. Sewn
	7.1.2.3.2. Glued

	7.1.2.4. Component material
	7.1.2.5. Spatial arrangement

	7.1.3. Shapes: different types of endleaves
	7.1.3.1. Integral endleaves
	7.1.3.2. Separate endleaves
	7.1.3.2.1. Single leaf
	7.1.3.2.2. Fold
	7.1.3.2.3. Hooks
	7.1.3.2.3.1. Endleaf hook
	7.1.3.2.3.2. Text hook

	7.1.3.2.4. Outside hook
	7.1.3.2.5. Guard


	7.1.4. Visualization problems
	7.1.4.1. Schema limitations
	7.1.4.2. Schema misinterpretations


	7.2. The set of transformations
	7.2.1. Markers: page markers, board markers, bookmarks
	7.2.2. Sewing
	7.2.3. Boards
	7.2.4. Spine shape and lining
	7.2.5. Endbands
	7.2.6. Coverings
	7.2.7. Furniture

	7.3. Dealing with erroneous data
	7.3.1. Error examples
	7.3.2. The feedback loop

	Summary

	Chapter 8. Recommendations
	8.1. Intersemiotic translations: methodological recommendations
	8.1.1. Intersemiotic translation elements
	8.1.2. Functional verbal descriptions
	8.1.3. The nature of sememes in diagrams
	8.1.4. Prototypicality of information
	8.1.5. Graphic prototypes
	8.1.6. Fixing the meaning of visual sememes
	8.1.7. Uncertainty

	8.2. Recommendations for the communication of bookbinding structures
	Summary

	Chapter 9. Conclusions
	9.1. Successful automated intersemiotic translations
	9.1.1. Summary of the recommendations

	9.2. Visualizations from models of
	9.3. Standardized automated drawings of bookbinding structures
	9.3.1. Summary of design principles

	9.4. Uncertainty visualization in bookbinding studies
	9.4.1. Summary of uncertainty visualization application

	9.5. Added value & benefits of visualization automation

	Chapter 10. Future work
	10.1. Integration of automated transformations within the data gathering framework
	10.2. Verbal to visual and reverse: a mixed approach
	10.3. Related work: Manuscript Collation Project
	10.4. Final remarks

	Bibliography
	Illustration credits