This book evaluates the results of two decades of research in

‘history and computing’. In spite of the fact that a lot has been

past, present and future of historical information science

accomplished, the report indicates critical places for improve-
ment. Many historians and other humanities scholars seem
satisfied with standard office tools, which do not always suit
Past, present and future
their complex sources and research questions. While more
and more archival sources have become available in digital of historical information science
form, there has not been enough attention to the development
of computational methods to process and analyse them. As a
result, the level of technical sophistication in many historical
studies has remained rather low.
onno boonstra, leen breure and peter doorn
It is argued that the concept of ‘history and computing’ under-
scores the duality expressed in the phrase itself. Is it aimed at
historians with an interest in using the computer as a practical
tool in their research, or is it aimed at information scientists
working in the subject area of history? As a consequence, nei-
ther of the needs and priorities of these parties have been fully
accommodated, and communication, both with ‘traditional’
history and with general computer and information science,
has remained less developed than might be possible.

The authors propose a new research agenda and provide

requirements for an adequate research infrastructure for fu-
ture historical information science.

International acclaim for the Past, present & future:

‘An extremely interesting report’.

boonstra, breure & doorn

Manfred Thaller, Universität zu Köln, Germany.

‘My high appreciation . . . a comprehensive analysis of the

past, present and future of Historical Computing . . . a right
diagnosis of the critical situation in the field’.
Leonid Borodkin, Moscow State University, Russia.

‘a lot of work and a great work . . . on the past, the present

and the future of Historical Information Science’.
Ingo Kropac, Karl-Franzens-Universität Graz, Austria.

‘This is really very good indeed... a ground breaking piece of

work... you have produced a landmark piece’.
Seamus Ross, University of Glasgow, uk.
niwi-knaw

niwi-knaw 2004
 past, present and future of historical information science



 data archiving and networked services

Past, present and future of

historical information science

onno boonstra, leen breure and peter doorn

amsterdam, 2006


 Second edition 2006

© 2006 dans
No part of this publication may be reproduced, stored in a retrieval system or
transmitted in any form or by any means, electronic, mechanical, photo-copying,
recording or otherwise, without the prior written permission of the publisher.

DANS – Data Archiving and Networked Services

Postbus 93067
2509 AB The Haque
T 070 3494450
F 07 3494451
[email protected]
www.dans.knaw.nl

isbn 90-6984-413-3

The paper in this publication meets the requirements of ∞ iso-norm 9706 (1994) for
permanence.

Design
Ellen Bouma, Edita-knaw, Amsterdam
www.knaw.nl/edita

This report is also published in the journal Historical Social Research / Historische
Sozialforschung, Vol. 29 (2004), No. 2.
First edition 2004 of the report was published for niwi-knaw


 Contents

Preface 7

1 Introduction   9

2 Historical information science   11

2.1 E-science, e-humanities, e-history  11
2.2 The life cycle of historical information  21

3 The past   25
3.1 The beginning of history and com­puting  25
3.2 Take off: Manfred Thaller’s clio  26
3.3 Getting organised  27
3.4 Ideas, claims and convictions  28
3.5 Main topics in historical information science  36
3.6 Final remarks   83

4 The present   85
4.1 Conclusions from the past   85
4.2 The lost topics   86
4.3 A failing infrastructure   88
4.4 The failing relation between history and computing and information science   88

5 The future   91
5.1 Conclusions from the present: A paradox   91
5.2 Relevant research lines   93
5.3 A future infrastructure for historical information science   95
5.4 Is that all there is?   99

Appendix   101

References   113

Acknowledgements   125

 Contents
 Preface

The idea of writing a report on the past, present and future of historical information
science arose when the knaw, the Royal Netherlands Academy of Arts and Sciences,
decided to close down the niwi, the Netherlands Institute for Scientific Informa-
tion Services. One Phoenix rising from niwi’s ashes would be a new e-Science
programme, oriented towards the exploration of the use of information technology
in the humanities and social sciences. Some of the themes that might be explored
in such a programme are revealed by examining what historical information science
has – or has not – accomplished over the last few years. This background also allows
the formulation of the requirements of an infrastructure in which historical informa-
tion science can thrive.
A second motive for the project was our feeling, based on the literature and a series
of discussions with scholars both within the Netherlands and abroad, that there was
a need for a new impetus to ‘history and computing’ internationally. The roots of
historical information science are grounded in quantitative socio-economic history
on the one hand and in computerised analysis of historical texts on the other. In the
second half of the 1980s ‘history and computing’ got a strong impulse from the ad-
vent of the pc. In the late 1980s and early 1990s the debates on history and comput-
ing flourished. The Internet stimulated many heritage institutes (archives, libraries
and museums) to digitise their collections. But since the late 1990s and in the first
years of the 21st century, the ‘history and computing movement’seems to have lost
momentum: humanities’ computing courses and departments in universities are
under pressure, journals or series of specialised publications are being discontinued
or diminishing, and conferences of the international Association for History and
Computing have become less frequent and attract less participants than before. Is
there a ‘crisis’ in historical computing? And if this is so, what is necessary to change
the tide?
Normally, educational responsibilities and other duties are too pressing to reflect
thoroughly on questions such as these. The extraordinary situation of the niwi made
it possible to spend some extra effort on this study. The Department of History of the
niwi created the opportunity and provided the financial resources. The Faculteit der
Letteren of the University of Nijmegen and the Institute for Information and Comput-
ing Sciences of Utrecht University made it possible that two of their senior lecturers
could work on this project for almost a year. We thank our institutes for giving us this
freedom.
Many people have discussed with us the scope, preliminary results, and implica-
tions of the project and report. We are very grateful to all who gave opinions, made
suggestions for improvement and criticized our endeavour. A more detailed list of
acknowledgements can be found at the end of this book. Of course, only the authors
can be held responsible for the views represented in this work.

April 2004

 Preface

1 Introduction

‘The historian who refuses to use a computer as being unnecessary, ignores vast
areas of historical research and will not be taken serious anymore’ (Boonstra,
Breure and Doorn, 1990)

When we wrote the lines above, fifteen years ago, we sensed that, with the coming
of the computer, not only new areas of historical research would be opened, but also
that computers would be able to help find solutions to many of the information prob-
lems that are so distinctive to historical science.
Nowadays, information problems in historical research still exist and are still vast
and very varied. They range from textual problems (what is the word that is written
on this thirteenth-century manuscript? what does it mean? to which issue does it
relate? why was it put there? why was the text written? who was the author? who was
supposed to read the manuscript? why has it survived?) and linkage problems (is this
Lars Erikson, from this register, the same man as the Lars Eriksson, from this other
register?), to data structuring problems (how can historical contextual information
be kept as metadata in a xml-database?), interpretation problems (from this huge
amount of digital records, is it possible to discern patterns that add to our knowledge
of history?) and visualisation problems (how do you put time-varying information on
a historical map?).
But this does not mean that nothing has been achieved over the last two decades.
On the contrary, hundreds of research projects have been initiated to tackle problems
like these. Historians, linguists, literary scholars, information scientists, they all have
done their share in making historical information science grow and flourish.
Nevertheless, if we look back at what ‘history and computing’ has accomplished,
the results are slightly disappointing. They are not disappointing because ‘comput-
ing’ failed to do what it intended to do, which was to provide ‘history’ with computer-
ised tools and methods historians could use to expand the possibilities and to im-
prove the quality of their research, but because ‘history’ failed to acknowledge many
of the tools ‘computing’ had come up with.
The primary aim of this report is to find out what, when and why things went
wrong. A major chapter therefore is dedicated to the Past (Chapter 3), and the way it
effected the Present (Chapter 4). In both chapters, attention is focused on content as
well as infrastructure, because both elements – the content of ‘history and computing’
research, and the infrastructure in which this research has been done – have had an
impact on the present situation of historical information science.
But disappointment has not been the major incentive to write this report. It is also
written to show how much has been accomplished within the field of ‘history and
computing’ and what great opportunities lie ahead for further research in computer-
ised methods to be used in historical science.

 Introduction
 As a consequence, the report ends with a few suggestions about the future of histori-
cal information science. Again, its future is not only a matter of generating new con-
tent for historical information science, but also about setting up a new infrastructure.
Both issues will be discussed in Chapter 5.
At this point, the concept of ‘historical information science’ is introduced instead of
‘history and computing’. This is done deliberately so. ‘History and computing’ is a
very vague and confusing term. Historical information science is neither ‘history’
nor ‘computing’. It is a science of its own, with its own methodological framework.
The object of historical information science is historical information, and the various
ways to create, design, enrich, edit, retrieve, analyse and present historical informa-
tion with help of information technology. In this way, historical information can be
laid out as a sequential phases of a ‘historical information life cycle’. In Chapter 2, a
definition of historical information science is given, as well as a short description of
the life cycle of historical information.

10 Introduction
2 Historical information science

2.1 E-science, e-humanities, e-history

Computing in the sciences and humanities has developed enormously over the past
decades. There is hardly a scientist or scholar remaining who does not use a com-
puter for research purposes. There are different terms in use to indicate the fields
that are particularly oriented to computing in specific disciplines. In the natural and
technical sciences, the term ‘e-Science’ has recently become popular, where the ‘e’
of course stands for ‘electronic’. Although there are now hundreds if not thousands
of compositions of ‘e’ cum substantive (e-Business, e-Culture, e-Learning, e-Social
Science, and even e-Love, e-Death and e-Pizza) in use, the term ‘e-Humanities’ is less
common than ‘humanities computing’ and ‘e-History’ is usually called simply ‘his-
tory and computing’.
Computer science or informatics as an independent discipline dates to only about
1960 It is often defined as the study of the systematic processing of information by
computer systems. In many scientific fields the application of computers has been
introduced as an auxiliary or subdiscipline.
Information science is the discipline that deals with the processes of gathering, ma-
nipulating, storing, retrieving, classifying and transferring information. It attempts
to bring together concepts and methods from various disciplines such as library sci-
ence, computer science and communication science.
Is it possible to discuss (historical) information science without first defining the
term ‘information’? Many definitions of the concept exist, but it is not our intention
to give an extensive overview here. However, a few words on how ‘information’ is
generally understood and used may be useful. All definitions agree that information is
something more than data and something less than knowledge. In the Open Archival
Information System (oais), information is defined as any type of knowledge that can
be exchanged, and this information is always expressed (i.e., represented) by some
type of data.1 According to McCrank, ‘data are what is given in the smallest units [...]
These are the raw material for building information’. He gives an admirable list of
definitions (‘in ascending order’), ranging from data, via information, facts, evidence,
and proof, to knowledge, belief, and ultimately wisdom. (McCrank, (2002)627-628).
In this section, we will go into some conceptual aspects of computing in the natural
and technical sciences, in the humanities and in history. Although there appears to
be some ambiguity in the usage of terms, we will try to present some common defini-
tions.

1
The oais reference model is approved as iso Standard 14721:2002.

11 Historical information science

 2.1.1 E-science and e-Social Science
The term E-Science is predominantly used in Great Britain, where an official and
generally used definition appears to exist:

‘science increasingly done through distributed global collaborations enabled by the

Internet, using very large data collections, tera-scale computing resources and high
performance visualisation.’

This definition by the Department of Trade and Industry is supported by the Re-
search Council e-Science Core Programme. Many researchers and institutes, such as
the Particle Physics and Astronomy Research Council (pparc) interpret this defini-
tion widely, to include computational and data grid applications, middleware develop-
ments and essential hardware procurement.2 The Oxford e-Science Centre uses the
same definition to describe its core activities.The World Wide Web gave us access to
information on Web pages written in html anywhere on the Internet. A much more
powerful infrastructure is needed to support e-Science. Besides information stored in
Web pages, scientists will need easy access to expensive remote facilities, to comput-
ing resources – either as dedicated Teraflop computers or collections of cheap pcs
– and to information stored in dedicated databases.
The Grid is an architecture proposed to bring all these issues together and make
a reality of such a vision for e-Science. Ian Foster and Carl Kesselman, inventors of
the Globus approach to the Grid define the Grid as an enabler for Virtual Organisa-
tions: ‘An infrastructure that enables flexible, secure, coordinated resource sharing
among dynamic collections of individuals, institutions and resources’ (Foster and
Kesselman, 1999). It is important to recognize that resources in this context include
computational systems and data storage and specialized experimental facilities.3 The
computational grid is the next-generation computing infrastructure to support the
growing need for computational based science. This involves utilization of widely
distributed computing resources, storage facilities and networks owned by differ-
ent organisations but used by individuals who are not necessarily members of that
organisation.
A descriptive way to explain computational grids is by analogy to the electric power
grid. The latter provides us with instant access to power, which we use in many dif-
ferent ways without any thought as to the source of that power. A computational grid
is expected to function in a similar manner. The end user will have no knowledge of
what resource they are using to process their data and, in some cases, will not know
where the data itself came from. Their only interest is in the results they can obtain
by using the resource. Today computational grids are being created to provide acces-
sible, dependable, consistent and affordable access to high performance computers,
databases and even people across the world. It is anticipated that these new grids will

2
See www.pparc.ac.uk
3
See: http://www.nesc.ac.uk/nesc/define.html

12 Historical information science

 become as influential and pervasive as their electrical counterpart.
In the social sciences, a National Centre for eSocial Science (ncess) has recently been
formed by the esrc (Economic and Social Research Council) in the uk. In parallel
to the developments in the natural sciences, the ncess aims to stimulate the uptake
and use by social scientists of new and emerging Grid-enabled computing and data
infrastructure, both in quantitative and qualitative research.4

2.1.2 E-humanities or humanities computing?

‘Humanities computing is an academic field concerned with the application of

computing tools to arts and humanities data or to their use in the creation of these
data.’ (McCarty, 1999)

Humanists have used computers since the 1950s, but until the 1980s usage could
be described as occasional (Feeney and Ross, 1993). It is clear from the literature and
on-line resources that, especially since the 1980s, computing has pervaded every
conceivable field of the humanities, although in some areas the role of computers has
become more important than in others.
The introduction of computing in the humanities was not universally met with
enthusiasm by researchers. There have been debates on the use of computers in the
humanities ever since they were introduced. Even today there are pockets of stub-
born resistance against computing. At the same time we can see that, although basic
computing skills of word processing, e-mailing and web browsing are nowadays
omnipresent among humanities scholars, their methodical and technical skills for
computerised research are fairly limited. In 2004 the steep learning curve of such
techniques, which was already observed in a report by the Commission for the Hu-
manities of the knaw in 1997, is as steep as ever.
It is not questioned that the electronic media are extremely important for opening
up sources for research that would otherwise remain hidden, inaccessible and impos-
sible to analyse. Digital media are undoubtedly more suitable for source publica-
tions than paper, and in many respects also more than microfilm. It is therefore not
surprising that many source publishers have turned digital.
Many fields in the humanities are based on the study of documents, hand-written
or printed, consisting of text, numbers and images. Other fields are based on oral
sources (speech) or on sound (music), on material objects (works of art, archaeologi-
cal objects, realia), or on visual information (photographs, film).
It is a matter of epistemological debate how fundamental the rise of computing
(and more in particular of the Internet) is for the ways in which knowledge is pro-
duced and used in the humanities. Clearly, the growth of the Web is changing the
behaviour and priorities of scholars in a number of respects, but the significance of
these changes is only partly understood. Although the importance of the Internet

4
ncess will consist of a co-ordinating Hub and a set of research-based Nodes distributed across the uk. The
Hub will be based at the University of Manchester, with support from the uk Data Archive at the University
of Essex.

13 Historical information science

 can hardly be overestimated, it is not sure that what we are seeing is a fundamental
change rather than the adaptation of what researchers have been doing before they
used computers to a new environment.
Nevertheless, information technology has brought changes in certain disciplines of
the humanities that can be labelled revolutionary. For example in linguistics, research
of language and speech has developed into a sophisticated, thoroughly technological
field. In archaeology the application of geographical information systems is now rule
rather than exception. Many historical dissertations are literally computer-based, that
is: grounded on sources that have first been converted to databases. Text corpora have
become indispensable in a lot of literary analysis.
Access to research data was and is a central issue in the humanities. The digitisa-
tion of catalogues and other access tools by libraries, archives, museums and other
heritage institutions was a first step to improved access to holdings. There is of
course a virtually endless amount of potentially relevant material for humanities re-
search, that currently is in analogue form and that could be digitised. Given the vari-
ety of disciplines within the humanities and the infinite number of subjects that can
be (and are) studied, it is unavoidable that any digitisation programme is selective.
Centres for humanities computing, which are focal points in research, teaching
and services on ict and the humanities, have been created in many countries. Again,
we limit ourselves to a few examples. In the uk, mention should be made of the Arts
and Humanities Data Service (ahds), which consists of five specialised centres for a
variety of humanities fields distributed over the country.5 The Office for Humanities
Communication at King’s College, London, is an umbrella organisation that fosters
communication among scholars and others involved in computer-related projects
and activities. It has published a series of monographs and collected papers con-
cerned with the impact of computers in humanities scholarship and higher educa-
tion.6 The Humanities Computing Unit at Oxford carries out research and develops
resources in many areas of humanities computing. The unit provides support for
academics in the humanities applying new technologies to their research and teach-
ing. Among other things, the unit includes the Centre for Humanities Computing,
the Oxford Text Archive, and the Humbul Humanities Hub.7 The mission of the
Humanities Advanced Technology and Information Institute (hatii) at Glasgow Uni-
versity is to actively encourage the use of information technology and information to
improve research and teaching in the arts and the humanities.8 Recently, the Arts and
Humanities Research Board has announced the establishment of an ict Methods
Network. The aim of this network is to promote and disseminate the use of ict in uk
Arts and Humanities research in order to enhance, develop and make more effec-

5
See: http://ahds.ac.uk/
6
http://www.kcl.ac.uk/humanities/cch
7
http://www.oucs.ox.ac.uk/; In 2006, Humbul has merged form the new Intute: Arts and Humanities
service: http://www.intute.ac.uk/artsandhumanities/. The Oxford Text Archive hosts the ahds Language,
Literature & Linguistics: http://ota.ahds.ac.uk/
8
http://www.hatii.arts.gla.ac.uk/

14 Historical information science

 tive the process of research, and to communicate research outcomes more widely
and efficiently. It will focus on new developments and advanced methodologies, on
research processes, questions, and methods, and on uses of data, rather than on data
creation and preservation or access to resources.9
In the usa, we limit ourselves to mentioning the Centre for Electronic Text in the
Humanities at Princeton and Rutgers Universities and the Institute for Advanced
Technology in the Humanities at the University of Virginia.10 In Canada, the mandate
of the Computing in the Humanities and Social Sciences facility at the University
of Toronto is to promote computing in research and teaching within the humani-
ties and social sciences.11 The list could continue for many pages with references to
centres and institutes in humanities computing in many European countries and
on other continents. The developments within the Netherlands were comparable to
those abroad, and in some fields Dutch ‘humanities computing’ was comparatively
strong.
The journal Computers and the Humanities, which was founded as a newsletter
as early as 1966, is a good source to get an insight in the broad field of humanities
computing. There are also many journals on this subject in the various humanities
disciplines, such as History and Computing. A series of monographs in the British
Library Research series give an excellent overview of humanities computing until the
mid-1990s (Katzen, 1990; Kenna and Ross, 1995; Mullings, 1996). A book edited by
Terry Coppock (Coppock, 1999), who as distinguished geographer and Fellow of The
British Academy did much to promote the use of information technology in scholar-
ship, extends the picture until the end of the last century. For more recent additions
and reflections on the use of ict in the humanities, some articles, conference papers
and, of course, Internet sources are relevant. For instance, the electronic publication
Humanist is an international electronic seminar on the application of computers to
the humanities. It provides a forum for discussion of intellectual, scholarly, pedagogi-
cal, and social issues and for exchange of information among members.12
Humanities computing is interdisciplinary by definition. The first thing that is
striking when we look at definitions of ‘humanities computing’ is that humanities
scholars tend to spend a lot of words on the concept, and that they approach the
subject with great care, if not with many detours. They even tend to discuss at great
length whether Humanities Computing is an academic field or not (see the confer-
ence papers from a seminar held at the University of Virginia in December 1999).13
Geoffrey Rockwell describes his quest about this question as a Table of Digressions:

9
http://www.methodsnetwork.ac.uk/.
10
http://www.ceth.rutgers.edu/index.htm and http://jefferson.village.virginia.edu/home.html and http://
www.iath.virginia.edu/.
11
http://www.chass.utoronto.ca/
12
Humanist is allied with the Association for Computers and the Humanities and the Association for
Literary and Linguistic Computing. It is an affiliated publication of the American Council of Learned
Societies and a publication of the Office for Humanities Communication (uk). See http://www.princeton.
edu/~mccarty/humanist/
13
‘Is Humanities Computing an Academic Discipline?’ The Institute for Advanced Technology in the
Humanities, The University of Virginia. http://www.iath.virginia.edu/hcs/

15 Historical information science

 Naming the Child
What is Multimedia

Conclusions
The Question
Multimedia at McMaster

Bibliography
Socratic Administration Implementation

Figure 2.1 ‘Is humanities computing an academic discipline?’ (Rockwell, 1999).

This figure illustrates the detours that humanities scholars tend to make when
they attempt to define the position of computing in their field.

In a terminological paper, Willard McCarty describes the historical development of

the term ‘humanities computing’, which was preceded by computing and the humani-
ties (as in the name of the professional journal) and by computing in the humanities.
He also considers humanistic informatics, but instead of arriving at a firm conclusion
he leaves us with two questions:
‘The basic question [...] however, is independent of its idiosyncrasies, namely,
what kind of a practice do we want? What approach to the other humanistic practices
is best for it?’ (McCarty, 1999).In Susan Hockey’s view, the core of the research
agenda of humanities computing is knowledge representation and manipulation,
or, to use a term that has a broader application in the world beyond academia, it is
about information management. In the humanities we are dealing with a variety of
extremely complex information (Hockey, 1999). This information can have a very
long life span and so electronic representations of it must also last for a long time, far
longer than the computer system on which they were created. The information can
also be studied for many different purposes and we therefore need to research and
develop computer tools not just to deliver the information to people but also to help
them study and manipulate the information in different ways. We need also to find
out how effective these tools are and to feed the results of evaluations of the tools into
new and better systems.
Espen Aarseth tries to find a good balance between what goes inside and outside
of the field of humanistic informatics by defining an upper and a lower threshold of
the field (Aarseth, 1998). He visualizes this as a pyramid of three levels, where the
top represents each traditional departmental field and its specific methodologies. The
middle level comprises the activities and applications that do not exclusively belong
to any specific field, but are relevant to several. This is where humanistic informatics
is located in Aarseth’s view. Beneath the lower threshold we find the applications,
method and perspectives that do not stand in any particular relationship with hu-
manities research or its research objects.

16 Historical information science

 Field-
specific IT

Upper threshold

Dicipline specific IT
Humanistic Informatics

Lower threshold

General IT

Figure 2.2 Aarseth’s three levels of information technology (Aarseth, 1998)

The Department of Humanistic Informatics at the University of Bergen keeps a focus

on the methodological tradition, while developing the fields of digital culture analysis
and the exploration of new media technologies, with the following directions as key
areas:
1. Humanistic it-methods. Studies of how humanities research applies new digital
methods to solve problems within the various disciplines. Examples of this are data
analysis by explorative (and traditional) statistics, systems for machine assisted
translation, text corpus, dictionaries, database applications (such as lexicography,
terminology), tagging and mark-up, geographical information systems, use of
simulation and dynamic models in the study of cultural processes, three-dimensio-
nal graphical presentation of objects and phenomena.
2. Multimedia and hypermedia research. Understanding and development of multi-
media-applications; distributed multimedia platforms and network communica-
tion, www-programming, hypertext-development and research on standards such
as xml, vrml, hytime, etc.
3. Pedagogical software and the development and use of network communication for
pedagogical purposes, such as distance learning. Information- and communication
technology (ict) -based didactics.
4. Digital culture and digital rhetoric and aesthetics. The study of digital modes of
communication and topics like computer art, digital literature, Internet cultures,
virtual reality, computer games, gender/identity and ict, through cultural and com-
munication theories.

By combining explorative, methodological, and critical approaches, the Department

of Humanistic Informatics aims to offer a uniquely humanistic perspective on ict,
and to create an independent theoretical framework for the study of the new digital
fields.

17 Historical information science

 2.1.3 E-history, ‘history and computing’ and ‘historical information science’

‘History as science is an information science. Its specialty is informing the present

about the past’ (McCrank, 2002)

The discussions and developments concerning the definition of ‘history and comput-
ing’ can be envisaged as a particular subset of (or parallel to) those in ‘humanities
computing’. Also here, there are different names for the field in use. The names and
definitions used are partly dependent on ideas about the field and partly dependent
on language. In Dutch, apart from ‘geschiedenis en informatica’ the term ‘historische in-
formatiekunde’ is common. The latter term refers to applied informatics and informa-
tion science in the historical discipline. Most historians regard computing in histori-
cal research as a technical and auxiliary trade. In English, ‘history and computing’
is the most neutral and encompassing term, while ‘historical information science’
refers to the specific field of history in the more general discipline of information sci-
ence. ‘Cliometrics’ is oriented on historical econometrics (or quantitative economic
history; Clio being the muse of history). In German the term ‘historische Fachinforma-
tik’ is in use, while in Russia ‘istoricheskaya informatika’ is used to indicate the field.
Also the term ‘historical information processing’ is used internationally.In a recent
survey of the literature, Lawrence McCrank proposes to define historical information
science as a hybrid field of study:

‘Historical information science integrates equally the subject matter of a historical

field of investigation, quantified social science and linguistic research methodolo-
gies, computer science and technology, and information science, which is focused
on historical information sources, structures, and communications.’ (McCrank,
2002)

In contrast to this, according to Charles Harvey, historical computing must be

concerned with the creation of models of the past or representations of past realities.
It cannot be defined simply in terms of areas of application or applied information
technology. Database systems or expert systems might happen to be of tremendous
interest, but there is nothing specifically historical about such things. They are just
general tools. Historical computing can only be defined in terms of the distinctive
contribution it can make to historical research. As a subject, it exists on the methodo-
logical plane, and none of its historical methods owes anything to computers as such:
historical computing can be done without computers. Computers merely make opera-
tional the concepts and methods that are the product of historical computing. His-
torical computing is a formal approach to research, that requires data and algorithms
to be made explicit, and, as such, it is part of scientific history.14
Historical Informatics has been defined as a new field of interdisciplinary specializa-
tion dealing with pragmatic and conceptual issues related to the use of information
and communications technologies in the teaching, research and public communi-
14
Italics are all ours.

18 Historical information science

 cation of history.15 It reflects the reality that the history-based disciplines are being
transformed through the impact of new technologies, to the point where new inter-
disciplinary norms and practices have emerged. As such, Historical Informatics
recognizes that the traditional division of labour between professional historians and
information specialists fails to meet the needs of either group.
To better cope with the demands of practitioners in the digital publishing era, his-
torians need to become more information literate, while information specialists need
to better understand and the specific information needs of historians. Arguably, what
is needed is a new generation of practitioners who are highly trained in the craft of
history as well advanced information skills, such as computer programming, data-
base development and multimedia production.
However, history and computing is not only about historical research, but also
about historical resource creation (Woollard, 1999). More and more archival sources
are becoming available in digital form. On the one hand, researchers are transform-
ing archival sources into digital files like modern monks (Doorn, 2000). Part of this
material is made available for secondary analysis by other researchers through data
archives. Reasoned and commented source editions, which in the past were only pub-
lished as books, are now increasingly being published in electronic form. Archives
and other heritage institutes are also digitising parts of their collections. Presently,
only a fraction of the archival holdings are available in digital form, but the amount is
constantly growing.
The growing availability of digital historical sources is bringing about a change
in the set-up and organisation of historical research. By bringing sources virtually
together that are physically stored scattered in archives around the globe (such as the
voc archives in Europe, Asia and Africa), new opportunities for comparative research
emerge, that were unfeasible in the past. Digital source collections can be studied
from different perspectives by larger groups of researchers in the form of virtual ‘col-
laboratories’ on the Web. Others can also check research if the digitised sources are
made available.
Another area in which new research possibilities are being created by the growing
availability of vast quantities of digital material, is that of electronic image repositor-
ies (of photographs, paintings, etc.). Historians traditionally use pictures as illustra-
tive material, but now systematically unlocked historical image collections can be
used for direct analysis. In addition to this, there is an exponentially growing modern
archive of data banks, e-mail correspondences, and specialised information systems
that only exist in digital form (Higgs, 1998). The historical research of the future will
increasingly depend on the processing of this ‘digitally born’ material. Of course,
there is a whole range of issues related to the problems of ‘digital longevity’, many of
which require further research.16

15
Welling has proposed to use the term ‘Computational History’ in an analogy of ‘Computational Linguis-
tics’ (Welling, 1998). However, computational history is often understood as the history of computing.
16
There is a fast growing body of literature on digital longevity, to which archivists, information scientists
and digital librarians contribute. Also in some commercial sectors the digital preservation of documents
and research materials plays an important role, e.g., in the pharmaceutical sector.

19 Historical information science

 In this report, we propose to define historical information science (historische infor-
matiekunde) as the discipline that deals with specific information problems in histori-
cal research and in the sources that are used for historical research, and tries to solve
these information problems in a generic way with the help of computing tools.
This definition not only excludes specific information problems outside the field
of historical research, but also general information problems that are not specific for
historical research. Our definition of historical information science is compatible
with the Aspen’s definition of humanistic informatics.
It is important to dwell upon the clause ‘in a generic way’ in the above definition.
Historical information science is not about finding specific solutions for specific his-
torical problems, but about finding solutions that transcend the specific problem. To
reach that goal, specific problems need to be formalised or generalised, before more
generic solutions can be found. Of course, the more general solution in the form of a
tool or demonstrator will also be applicable in the specific case.
So, ‘historical information science’ is about determining and solving information
problems. The information problems in historical research can be divided into four
categories:

− information problems of historical sources

Sources are the basis of historical research. They have their own particularities:
they contain inconsistencies, they are incomplete or incompletely preserved, and
they are unclear or ambiguous. Another characteristic of sources is that they may
contain information about different units of analysis: a population register con-
tains information about individuals, about the families to which they belong, and
the households in which they live. This type of multi-level problem requires further
investigation.
Moreover, not all historians have a clear research question at the start of their inves-
tigation: it is therefore impossible and undesirable to model an information system
on the basis of an information requirement that has not yet crystallised.
Finally, historical research aims to unravel ‘wie es damals gewesen ist’. This means
that the meaning of a certain piece of data cannot exist without interpretations.
These interpretations therefore need to be added to the information system that
has been built on the basis of the source. However, interpretations are by definition
subjective. They need to be added, but in such a way that they can be separated
from the original data in the source.

− information problems of relationships between sources

When several sources are used in historical research additional problems will arise.
In the first place there will be the linking of the data: how can we establish whether
a person in source A is the same as another person mentioned in source B? This
is particularly problematic if there are spelling variations or more persons with the
same name (more on nominal record linkage, including literature references, in
section 3.5.1.6). In the second place there is the problem of changes over time and
place. The meaning of data from a source is dependent on the spatial and temporal
context. When linking such a source to another one, potential changes in meaning

20 Historical information science

 need to be considered. For instance, an occupational title such as ‘labourer’ can
mean ‘agricultural worker’ at one point in time and ‘factory worker’ at another
point in time.

− information problems in historical analysis

Historical research deals with changes in time and space. Therefore, historians
need analysis tools that take these changes into account. Many statistical analysis
techniques that historians use are borrowed from the social sciences. The social-
scientific analysis methods that measure changes over time are only suitable in an
experimental setting. The other techniques are at best usable for cross-sectional
analysis, but not for diachronic analysis. Several techniques have been developed
that seem useful for historical research (such as event history analysis), but the ap-
plicability of these techniques still needs to be determined.
The same is true for techniques that process information from different units and
levels of aggregation. Multilevel regression techniques seem appropriate. However,
applications of ecological inference techniques in historical research are yet to be
explored.

− information problems of the presentation of sources or analysis

In accordance with the lack of specific historical analysis tools, presentational tech-
niques for historical data also require research. For instance, methods to represent
changes in time and space, to visualise multi-level linkages, etc.

Historical information problems are different at various stages of historical infor-

mation. Therefore, an important classification scheme is the life cycle of historical
information.

2.2 The life cycle of historical information

Like many objects that take part in a process of production and consumption his-
torical information will go through several distinct stages, each related to a specific
transformation, which produces a desired quality. Each of them represents a major
activity in the historical research process. Together the stages are referred to as a
life cycle. The idea of an information life cycle is derived from records management
(sometimes referred to as ‘document control’), where the idea of document life cycle
is central to the overall process: design and creation of records, approval, retrieval,
circulation, access and use, archiving and destruction.
Historical information is to be distinguished from raw data in historical sources.
These data are selected, edited, described, reorganized and published in some form,
before they become part of the historian’s body of scientific knowledge. Informa-
tion once presented may give cause for the creation of other information, in this way
closing the circle. The stages of the life cycle are not always passed through in a strict
sequence, and some of them may even be skipped under certain circumstances.

21 Historical information science

 Enrichment Editing

Durability Retrieval
Creation Usability
Modelling

Presentation
Analysis

Figure 2.3 The life cycle of historical information

The edge of the life cycle shows six stages:

1 Creation. Creation comprises not only the physical production of digital data, but
also the design of information structure, (e.g., through data modelling, text model-
ling), and, in a broader sense, the project’s design as well. The practical aspects
regard data entry and text entry tools, digitization (like ocr) and considering appro-
priate database software.
2 Enrichment. The bare data, images and texts once created will need to be enriched
with metadata, which describes the historical information itself in more detail,
preferably in a standardized way (e.g., by means of a system as Dublin Core), and
intelligible to (retrieval) software. It comprises also to the linkage of individual data
that belongs together in the historical reality, because these data refer to the same
person, place or event (so-called nominal record linkage).
3 Editing. Enrichment changes into editing, which includes the actual encoding
of textual information by inserting mark-up tags, or by entering data in the fields
of database records. Enhancement may be considered as a separate phase of the
editing process. It is the process by which data is transformed through algorithmic
processes preliminary to analysis. Editing extends also to annotating original data
with background information, bibliographical references, and links to related pas-
sages.
4 Retrieval. When information has gone through the previous stages, it is ready to be
selected, looked up and used (i.e. retrieval). In some projects this will happen only
after a formal publication, which moves retrieval to a later point in the life cycle.
Retrieval itself is based on mechanisms of selection and look-up, like sql-queries in

22 Historical information science

 the traditional database environment, and Xpath and Xquery expressions for texts
in xml-format. In addition it pertains to the user interface, for both the specifica-
tion of the query itself, and for the display of results (in the form of a simple list or
through in a more advanced visualized representation).
5 Analysis. Analyzing information means quite different things in historical re-
search. It varies from qualitative comparison and assessment of query results, to
advanced statistical analysis of data sets.
6 Presentation. Historical information is to be communicated in different circum-
stances through various forms of presentation. Although conceptually represented
here as the final stage, it will actually happen frequently in other stages as well.
Presentation of digital historical information may take quite different shapes, vary-
ing from electronic text editions, online databases, virtual exhibitions to small-scale
visualizations of research results within a single project.

In addition, three practical aspects have been grouped in the middle of the life cycle,
which are central to computing in the humanities and in different ways related to the
six aforementioned stages:
− durability, which has to guarantee the long term deployment of the thus produced
historical information;
− usability, which regards the ease of use, efficiency and satisfaction experienced by
the intended audience using the information produced;
and, finally,
− modelling in a broader sense than the data modelling or text modelling, mentioned
above. Here, modelling refers, amongst other things, to the more general model-
ling of research processes and historical information systems.

23 Historical information science

24
3 The past

3.1 The beginning of history and com­puting

The beginning of applying computer techniques to historical material is often as-
sociated with the well-known impressive literary and lexicographic projects from the
early days of humanities computing. Classic milestones are the production of the
Index Thomisticus by father Roberto Busa17 (finally resulting into 56 printed volumes
with analysis of the work of Thomas Aquinas), the extensive electronic editions of
Christian Latin texts by the cetedoc in Louvain from the 1970s onwards18, and the
founding of the widely accepted professional journal Computers and the Humanities
by Joseph Raben in 1966 (Gilmour-Bryson, 1987). One of the early larger undertak-
ings was The Philadelphia History Project (Hershberg, 1981). In art history, Laura
Corti directed attention to the relevance of computers (Corti, 1984a, 1984b; Corti and
Schmitt, 1984). All this may be regarded as the beginning of history and comput-
ing in a broader sense, but history and computing in a more narrow sense, explicitly
related to history as a discipline, took shape later. In this respect, the developments in
the United States differed from those in Western and Eastern Europe.
In the United States, computer-aided historical research took off in the 1960s,
mainly propelled by enthusiasm for social and economic history. Using methods and
techniques originating from the social and economic sciences, it focused, therefore,
on computation and quantitative aspects (Greenstein, 1997). Both Thernstrom’s
Other Bostonians (Thernstrom, 1973) and Fogel and Engerman’s Time on the Cross
(Fogel and Engerman, 1974) are considered milestones in this respect. It did not take
long, however, before quantitative history started to find opponents. Cliometrics was
severely criticised for placing too much trust in counting and in fragmentating the
historical image of the past.
In various countries in Western Europe, social history failed to promote broader
interest in the application of computing techniques. In France, for instance, the
detailed interest in the social and economic basis of life, so well expressed by the
Annales group, hardly impacted on historical computing. The first historians to use
computers for research purposes, were historical demographers. The Cambridge
Group for the study of population and social structure (established in 1964) centred
around Peter Laslett led the way.

17
Roberto Busa was born in Vicenza on November 28th 1913. He started his career as a Jesuit scholar and
became a full professor of Ontology, Theodicy and Scientific Methodology. In 1946 he planned the Index
Thomisticus. The work was mainly carried out in Gallarate and in Milan until 1967, in Pisa till 1969, in
Boulder (Colorado) till 1971 and, for the next nine years, in Venice, where, from 1974 till 1980, the pho-
tocomposition of the 70,000 pages forming the 56 encyclopaedic volumes of the Index Thomisticus was
accomplished using ibm computers: http://www.allc.org/refdocs/honmems.htm#Busa
18
cetedoc: CEntre de Traitement Électronique de documents, directed by P. Tombeur. For a list of publi-
cations refer to http://zeus.fltr.ucl.ac.be/recherche/publications/pub_source.html

25 The past
 It may very well be that the traditional character of the history curricula in the Eu-
ropean arts faculties did not foster close co-operation with the social sciences, as
had happened in the United States. Looking for guidance in computer applications,
scholars in the humanities had to rely on help mainly from computer linguists. A
great deal of the activity therefore centred around source editing, e.g. concordances
and source editions by the cetedoc in Belgium and the cnrs in France19, where,
amongst others, Lucie Fossier, Caroline Bourlet, and Jean-Philippe Genet from the
Institut de Recherche et d’ Histoire des Textes, have shaped computer-aided research
projects in medieval history, with the bulletin Le Médiéviste et l’ Ordinateur (founded
in 1979) as an important channel for scholarly communication.
In Western Europe, Germany was the exception to the rule. In 1975, German
historians and sociologists founded the Quantum-group in order to explore, in close
collaboration, possibilities and problems with the use of historical and process-pro-
duced data.20 It was driven by a feeling of uneasiness in empirical sciences with data
based on surveys only, and by the turn of historians away from ideographic and nar-
rative approaches. It was aimed at closing the gap between the German situation and
the upswing of quantitative history elsewhere. A few years later, its journal, Quantum
Information, changed its name into Historical Social Research (Historische Sozialforsc-
hung) and grew into a broader platform for publication of subjects concerning history
and computing, with its focus, however, on the computational aspects of historical
research.
At that time, in Eastern Europe, and especially in the former ussr, a remarkable
co-operation existed between historians and computer specialists with a background
in mathematics. The early experiences with historical computing were related to the
processing of statistical data in the field of social and economic history. In the 1970s
quantitative history gained a firm footing there, with special interests in problems of
historical simulation (Borodkin, 1996).

3.2 Take off: Manfred Thaller’s CLIO

In these early days, historians had to rely on general purpose software, available in
local computer centres, or had to develop – mostly with some technical assistance
– their own more specific tools, which usually boiled down to ad-hoc solutions. Some
historians experimented with standard database software: user-unfriendly hierarchi-
cal database management systems which required programming for navigation and
data look-up, or the not yet fully-grown relational database programs, already pop-
ping up sparsely on university mainframes. Once the problems of data storage and
retrieval had been overcome, data had to be reformatted for statistical analysis, which
required some additional programming.
The typically individual nature of most of these projects defied the potential
progress in historical computing that could have been made due to the widespread
awakening interest in the prospects of this new technology. This was well understood

19
cnrs: Centre National de la Recherche Scientifique, in this context particularly relevant its laboratory of
texts: L’ Institut de Recherche et d’ Histoire des Textes (irht).
20
Quantum Information 2 (1977) p. 1-2.

26 The past
 by the Max Planck Instut für Geschichte in Göttingen, where Manfred Thaller took
the initiative in producing dedicated historical software. In 1980, he announced the
birth of clio (later rechristened to κλειω), a genuine historical database manage-
ment system (Thaller, 1980). It stood out from other database management systems,
among other reasons, by its flexibility in data input formats which reflected the struc-
ture of the historical source rather than the requirements of the computer program,
a Latin-based command language, but above all by a clear vision on what historical
information processing should be. This vision was to be translated into the software’s
capability to support the process of historical research and the variety of historical
sources, without forcing data to be squeezed into standard formats. Moreover, it had
some built-in awareness of the fuzziness of historical data and the complex process
of historical interpretation, which should prevent hasty conclusions in the data entry
stage.
Although clio was primarily intended to support ongoing research on the Max
Planck Institute itself (which had some emphasis on social and economic history),
Thaller offered the clio package to research projects elsewhere and looked for co-op-
eration in further software development. Although this offer was necessarily limited
to a small number of projects, it heralded a new period of historical computing.

3.3 Getting organised

Apart from clio, which was set up to be of use to various historical research projects,
historians became aware that it would be worthwhile to share the computational
problems they encountered, and the solutions they reached, with their colleagues.
Organisations such as the Association for History and Computing were set up in
order to facilitate them. In the United States, for instance, the American Historical
Association created such a platform for quantitative historians (Greenstein, 1997).

3.3.1 The Association for History and Computing

European historians missed an independent organisation for sharing experiences
and establishing an understanding of historical computing. In 1983, an international
group of enthusiastic computer using historians gathered for a symposium at Hull.
Ambitious ideas circulated, like the launching of an ma in historical computation.
The most lasting outcome of this meeting was the Association for History and Comput-
ing (ahc), established three years later by an inaugural conference at Westfield Col-
lege (London). In the following decades, international conferences were organised
under the auspices of the ahc, from all over Europe to Moscow and Montreal, and
professional communication was sustained through its journal History and Comput-
ing.
The ahc was established at the same time as an enhanced awareness of the perva-
sive power of computer technology in society was developing. Governments across
Europe reacted to this through stimulation programs targeted at the acquisition of
basic computer skills by all students. The Dutch government, for instance, decided
to stimulate the use of computers particularly in the arts disciplines and the social
sciences. In 1985, perestroika favoured the start of a campaign for computerisation in

27 The past
 the ussr, which brought the requirement of it-certificates for university teachers and
obligatory computer courses for all students (Borodkin, 1996). However, for a long
time wide-scale progress was hindered by a lack of sufficient hardware.

3.3.2 As an example: alfa-informatica in the Netherlands

As a result of additional funding from the mid-1980s onwards Dutch universities
enlisted special computer literate staff to teach computer classes directed towards the
professional interests of arts students (the so-called alfa-informatica) and, to a certain
extent, to give support to fellow researchers. This lofty initiative was lacking in one
fundamental respect: it did not provide any research infrastructure for historical
computing itself. Although most of the staff members had some research opportuni-
ties alongside their teaching, research in technical aspects of historical computing
remained isolated. Within the framework of nwo (the Dutch agency for research
funding) applications for methodological and technical research projects tended
to fall between two stools. They were not sufficiently technical to arouse interest
amongst computer scientists, and tended to be outside of the intellectual sphere of
leading historians then unfamiliar with the use of computers.
Despite this there were further practical organisational changes in the field of
computer-aided historical research. In 1987 the Dutch-Flemish/Belgian counterpart
of the ahc was founded – the Vereniging voor Geschiedenis en Informatica and was soon
affiliated with the ahc, followed by the creation of the Dutch Historical Data Archive
(Nederlands Historisch Data Archief – nhda) in 1988 This ambiguous situation in the
Netherlands, a new domain of academic knowledge without an adequate research
infrastructure, reflected the diversity of views on historical computing held in gen-
eral. Even within the ahc different opinions on the relationship between ‘history’ and
‘computing’ managed to live side by side, while ‘history and computing’ as a particu-
lar form of methodological research was only a high priority for some individuals.

3.4 Ideas, claims and convictions

Reflection on the phenomenon of historical computing itself, i.e., the interplay of
leading ideas, claims and convictions about suitable methodology, software and the
relation of this field to history as a discipline, is far more relevant for an understand-
ing of its past than a reflection on the detailed historiography as unfolded through
numerous project papers in the proceedings of the several conferences, journals
and workshop volumes. These are, in any case already well documented by (Denley,
1994; McCrank, 2002; Speck, 1994; Woollard, 1999). Consequently, this histori-
cal overview will be curtailed and the remainder of this chapter will focus on more
philosophical aspects.
The prevailing ideas about historical computing can be reduced to two basic points
of view:
• Plain IT
This category of view is characterised by a high assessment of the native capabili-
ties of information technology for the historical discipline. Those who support this
point of view, tend to praise the possibilities of computer technology as it is (thus

28 The past
 ‘plain’ it). They will recommend computer usage and the mastering of necessary
computer skills. The underlying tacit assumption seems to be that it-as-available is
good enough and covers most, if not all historical requirements; it needs only to be
learned and to be applied. If technology fails in certain respects, its shortcomings
have to be accepted or to be circumvented. For various reasons, enhancing and
adapting the technology itself seem to be beyond their range.
• Enhanced IT
Views of this type tend to emphasise the special and complex nature of historical
data processing in contrast with computer applications in, for example, business
and hard sciences. They show less confidence in standard information technology and
pay more attention to dedicated software, to special tools, to the implementation
of additional knowledge layers, and to fine-tuned methodologies and techniques.
Here, the assumption is rather the opposite of Plain it: information technology as
it comes to us should be adapted and extended before it can meet the requirements
of historical research – and we should go for it!

Both points of view are not entirely contradictory, and there are scholars who take an
intermediate position between both extremes. There are of course situations where
standard solutions are sufficient, and there are situations where existing tools are
clearly lacking. Nevertheless, both approaches differ enormously with respect to the
direction of historical computing as a field of study. The first one directs attention
more to the practice of historical research, representing the computer as a ready-
made tool, like a car, a typewriter or a video camera and rates historical computing as
hard core historical research. The second calls for investments in dedicated software
and favours the development of historical computing into the direction of a historical
information science with a theoretical basis.
The ahc conferences in the late 1980s (in particular, Westfield i and ii, Glasgow
and Cologne) produced fruitful discussions on these subjects, but, unfortunately, did
not reached firm conclusions. The later conferences, after 1990, show a predomi-
nance of reports on the application of information technology with an emphasis on
historical aspects, and a decline in debates on philosophical issues.21

3.4.1 Plain IT

3.4.1.1 optimism about technological achievements

The first Westfield conference (1986) praised the progress in computer technology
and its salutary effects for the historical field. It was impressed by the ‘microcomput-
er revolution’ that had taken off. Ann Gilmour-Bryson remembered in her opening
paper:

‘Only six years ago, we were forced to learn dbase programming and to deal with its
truly dreadful manual if we wanted to manipulate historical databases on out tiny
16, 32, or 64 k cp/m based early micros. As I typed this article on my ibm pc at with

21
Exceptions were the Nijmegen (1994), Moscow (1996) and Tromsø (2003) conferences.

29 The past
 its 1.2 megabyte floppy disk, its 20 megabyte hard disk, making automatic backup
on its 20 megabyte tape backup system, and printed it on an 8 page per minute
printer, it was hard to believe that such progress had taken place in six years.
Readily available software with a type of artificial intelligence interface enables the
novice to build a database, query it, and report from it within a few hours at most.’22

She concluded that we could now easily test so many hypotheses by simply asking
questions of our databases and solve problems that would not have been solved ‘be-
fore the present day because of a lack of time and/or manpower’. The computer has
become the historian’s intelligent and high-speed slave.
In their introduction to the conference proceedings Deian Hopkin and Peter
Denley noted that historical computing had evolved considerably from the early days
of quantification and basic data processing and now provided a ‘common ground
between historians who otherwise inhabit segmented and secluded worlds’. Others
praised ‘the variety of historical uses to which the computer can be put’, unburdening
historians from tedious, repetitive tasks (Woods, 1987). What drew people together
was the enthusiasm of working with large volumes of complex data, now made pos-
sible by the blessing of cheap and generally available computer power (Harvey, 1990).
Although this idea of the ‘unifying power’ of the computer, bringing people togeth-
er to share their experiences, was a common feeling among the conference partici-
pants thanks to the warm hospitality of Westfield College, it was limited in practice
and a bit naïve. It did not make clear what, exactly, were the common assets, except,
perhaps, for co-operation in the acquisition of funding. Everybody who used the op-
portunities to learn more about historical computing, would discover very soon that
he had to made choices, particularly with regard to methodology and software to be
used, and was thus confronted with the different directions of thought.

3.4.1.2 Extending historical craftsmanship

Of course, from the beginning everybody agreed on the principle that users should
learn more about computers in order to use them productively. But this could mean
quite different things. Historians, for example, had to learn to be more explicit in
their research strategies because computers would require detailed instructions.
Some considered this as adding electronic aids to the critical intellectual skill histori-
ans had developed since the time of Herodotus and suggested that historians should
learn programming as well (Woods, 1987). From this perspective, historical comput-
ing had (or at least should have) become entirely part of the historical craft, which
had to be modernised due to advances in technology. As Adam Hodgkin expressed it
in 1986, speculating about the situation ten years later:

‘It is arguable that there will be no longer a need for a conference on history and
computing. By then the uses of computers in historical research will be so well un-
derstood and so much a part of the fabric of scholarship that it would be as unnec-
essary as having a conference on libraries and history. I predict that there will be a
22
(Gilmour-Bryson, 1987), p. 7.

30 The past
 history and computing conference in 1996, but I have some sympathy for the view
that there is nothing of importance, apart from historical content, that is unique
about historical research and computing. There can be very few computing techniques
which are solely of interest to historians.’23

The computer is a tool, which, like many other tools, has some general utility in the
study of history (Greenstein, 1989). The same idea was expressed in the introduction
to the papers of the Glasgow Conferences:

‘Computing does not have to be complicated; indeed there is a danger that those who
insist that it is so are losing focus of their initial historical enquiry, replacing it with
a technology dependent methodology.’24

At the end of that conference, Charles Harvey philosophised about the nature of
historical computing. Looking backward, he expressed ideas that proved to be wide-
spread among computing historians and which have not particularly favoured the
growth of historical information science as a methodological discipline. They mar-
shalled feelings and attitudes that justified a turn away from the technical aspects.
Pre-eminently, according to Harvey, historical computing must be concerned with
the creation of models of the past or representations of past realities. It cannot be
defined simply in terms of areas of application or applied information technology.
Database systems or expert systems might happen to be of tremendous interest, but
there is nothing specifically historical about such things. They are just general tools.
Historical computing can only be defined in terms of the distinctive contribution it
can make to historical research. As a subject, it exists on the methodological plane,
and none of its historical methods owes anything to computers as such: historical
computing can be done without computers. Computers merely make operational the
concepts and methods that are the product of historical computing. Historical com-
puting is a formal approach to research, that requires data and algorithms to be made
explicit, and, as such, it is part of scientific history.25

3.4.1.3 Relying on standards: the relational database

Others allowed more room for ‘foreign imports’ from computer science, like a thor-
ough requirements analysis, which would help to determine the feasibility of histori-
cal computer projects and could deliver its justification (Greenhalgh, 1987). From the
trust in the current state of technology followed naturally the reliance on universal
standards, like the relational model for data management.
The application of the relational database model to historical data collections has
been defended from the early beginning, e.g. (Greenhalgh, 1987). In the ahc com-
munity, this claim has managed to co-exist with arguments for a special historical
data model as that of κλειω. Although κλειω has kept a niche, about the end of this

23
(Hodgkin, 1987), p. 256.
24
(Mawdsley, Morgan, Richmond et al., 1990), p. xi.
25
Italics are all ours.

31 The past
 period the hegemony of the relational database was virtually unquestioned among
British historians and as scarcely less influential elsewhere (Denley, 1994b). Green-
stein looked upon the relational database as a tool particularly suitable for source-
oriented data processing. A source-oriented approach should allow for two basic
requirements: the same source is handled differently in various stages of historical
research and the uses of sources vary over time. A relational dbms catered very well
for the dialectic interpretative process with its resort to the original, because raw
source fragments could be copied to database records without any textual change and
be linked to standardised data afterwards, allowing thus efficient comparison and
analysis while the original text was kept as well (Greenstein, 1989).
At the end of this period Microsoft introduced its popular relational desktop dbms
Access, a wonderful big lie, with respect to the complexities of database design. It
was wonderful because of its user-friendly interface. It rapidly swept away its stub-
born predecessors like dbase and Paradox. If a historical data set was not too com-
plicated, database design and querying were easy. Finally, the computer seemed to
have reached the stage of development of the modern car: the mechanic with his
oilcan was no longer needed. Built-in ‘wizards’ compensated for lack of theoretical
knowledge and querying a database could be as simple as searching for words in a
text processor. One could even successfully complete certain tasks without knowing
exactly what had happened.
But when exactly does a database become so complicated that standard facilities are
no longer adequate? No red warning lights will flash! Software like Microsoft’s Ac-
cess had (and has) a tendency to mask real problems in database design, in particular
in historical database design. Many historians discovered far too late that the design
of their database did not meet their requirements and / or the inherent structure of
their source material, when the desired results failed to come out. Denley succinctly:

‘It has to be observed that the marriage of history and relational databases is one of
convenience (some would say inconvenience) rather than design.’26

The flexibility of the relational model in adding new data layers did not solve, of
course, all the typical problems of historical computing. Introducing clio, Thaller
had already pointed to the inherent fuzziness of historical data and the complex proc-
ess of historical interpretation. If the entire process of historical data retrieval was
left to the dbms, somehow historical knowledge had to be incorporated. This was not
easily done in the relational database environment itself.
The details of the relational model and related techniques of data modelling (like
the Entity Relationship Model) were made widely known through the work of Lou
Burnard and Charles Harvey, in several articles and in particular through Harvey’s
book on databases in historical research (Hartland and Harvey, 1989; Harvey, Green
and Corfield, 1996; Harvey and Press, 1993). Burnard also recognised the complexity
of representing historical reality in a computer. In designing a database, historians

26
(Denley, 1994a), p. 35.

32 The past
 should start with a sound ‘conceptual model’27), which comprised the real world ob-
jects, events and their relationships. Next, this model had to be mapped on to the sort
of data structures a computer can deal with. He admitted that the integration of the
different levels (from raw textual data to identified historical individuals and events)
was not easy with standard software, but, in spite of that, he considered the relational
model as the most viable solution. A lesser degree of refinement in automation
might be acceptable: some information can be stored separately and administered by
hand and the mapping from conceptual to physical model took place completely in
the historian’s head (Burnard, 1987, 1989, 1990).

3.4.2 Enhanced IT

3.4.2.1 Genuine historical software

From the beginning, the furthest-reaching position in the other line of thought has
been taken by Manfred Thaller. At the first Westfield conference he presented a
paper on methods and techniques of historical computation (Thaller, 1987) which
outlined his main ideas on this subject, though these were to be more fully formu-
lated in later publications:28

1. Historians deal with problems not appearing in other disciplines, which should
be controlled with a level of skill a historian can be expected to acquire without re-
focusing his main research interest. So, the enhanced-it view intends to make life
easier for common historians by providing expert tools.
2. Historical data is to be administered as pieces of text, without any assumption a
bout its meaning. Meaning depends on interpretation, which is a fruit of histori-
cal research. Therefore, data should be entered in a source-oriented way (keeping
together in a single file what appears in a single source document), rather than in a
program-oriented way. His definition of ‘source-oriented’ is, however, more inclu-
sive than those in the previous section:
‘Source-oriented data processing attempts to model the complete amount of in-
formation contained in an historical source on a computer: it tries to administer
such sources for the widest variety of purposes feasible. While providing tools for
different types of analysis, it does not force the historian at the time he or she cre-
ates a database, to decide already which methods shall be applied later.’ (Thaller,
1993b)
3. The typical historical database management system (like his clio/κλειω) would be
a hybrid between a classic structured dbms, a full-text retrieval system and a docu-
ment retrieval system (which sounds more familiar in a time of xml-databases than

27
A conceptual model is a map of the world whose data are administered in the database. It goes with an
early stage in methodological system design. It exists usually outside the dbms itself and is created with
the aid of a case-tool or a diagramming package (e.g. nowadays Visio). The dbms ‘does not know’ about
it. On basis of the conceptual model the database designer defines the physical model: the set of tables,
with fields of a specific length and data type. Only the physical model is actually used by the dbms in data
management operations.
28
For a detailed, more technical description, refer to (Thaller, 1993a).

33 The past
 twenty years ago), provided with some specific subject knowledge and inference
mechanisms in order to enable historically meaningful data retrieval (‘interpreta-
tion aware’).
4. Such a system must be able to overcome differences in spelling (e.g. in surnames)
and to handle data related to individuals in a careful way, allowing for research-
based strategies for linking source fragments containing names to historical indivi-
duals (so-called ‘nominal record linkage’). This would require the implementation
of some knowledge in form of ‘logical objects’, containing rules for interpretation
by the software.
5. Finally, it should take care of all other required transformations of data, for
example, for the benefit of statistical analysis.

Thaller’s view did not exclude the use of systems like dbase or Paradox for uncom-
plicated data processing and simple data storage. However, he suggested that one
should remain aware of the structural limitations of this kind software. His main
concern was not about the software in itself. At the second Westfield conference he
argued for a distinct theory of historical computing, a well-founded conceptual frame-
work which would allow professional discussions about the peculiarities in histori-
cal data processing, firmly stating his belief in the fundamental difference between
‘normal’ and historical data processing (Thaller, 1989).
The obvious question to ask is, why κλειω didn’ t sweep away its competitors in
world of historical research like Microsoft Access did on the desktop. As Peter Denley
noted in 1994 in his survey of the state of the art in historical database management,
the power of the software has taken its toll. There is an almost infinite number of data
structures possible; the tools to query, analyse, and manipulate the sources are pow-
erful and sophisticated. User-friendliness was not made a priority, simply because in
Thaller’s opinion historical computing was a demanding science and that historians
did themselves a great disservice if they made it look simpler. However, data prepa-
ration could be far less laborious with κλειω than with a relational system, and not
everybody needed to delve too deeply inside the tool set.In addition, source-oriented
data processing itself has attracted fundamental criticism. Many historians worry that
purists who wish to represent the source electronically in a form that is as close to the
original as possible, may be according a low priority to analysis, and may have a mis-
placed faith in the authority of text. Along with this line of reasoning the value of the
source itself can be put into perspective as a mediated representation of the historical
past (Denley, 1994a).

3.4.2.2 Historical tools on top of standard software

In addition to Thaller’s strict source-oriented philosophy with far-reaching impli-
cations of non-standard software, several computing historians felt a need for an
intermediate approach. They realised the limitations of commercial packages on the
one hand and the value of general standards and proven methodologies in compu-
ter science on the other hand. This position inspired the creation of add-ons and
reusable applications (or application frameworks) on top of standard software like
relational database systems. These solutions were designed to comply with special

34 The past
 requirements in historical data processing and, at the same time, would benefit from
rapid developments around middle-of-road information technology.
It is hard to distinguish in this category between ‘mere applications’ and ‘tools
with a wider scope’. The delicate point is the way the added value (specific histori-
cal knowledge, algorithms or data structures) is documented and made explicit and
available to a broader audience. Having a theoretical foundation together with some
philosophy about how to serve historical research is an essential prerequisite. How-
ever, the added value of such ‘tools with a wider scope’ was not easily recognized by
fellow historians, who tended to classify these attempts amid ordinary applications.
The list below is certainly not complete, but covers a few representative examples:
• One way of a realising this idea was creating a model application that clearly de-
monstrated how specific peculiarities of historical sources could be handled within,
for example, a relational database environment, as Boonstra did for event-history
(Boonstra, 1994a; Boonstra and Panhuysen, 1999; Boonstra, 1990). Gunnar
Lind compared and analysed different structures for prosopography, suggesting a
standard core database structure with the relational model in mind (Lind, 1994).
Morris explored how standard applications could be combined efficiently for ni-
neteenth century wills, hopping from one commercially available and well suppor-
ted program to another and exploiting each application in areas of functionality in
which it was strong and user-friendly (Morris, 1995).
• Welling studied intelligent data-entry strategies and user interfaces for highly
structured sources and implemented his ideas in Clipper using dbase-files for
storage (Welling, 1993; Welling, 1992). Reflecting on his work he stood up for a dis-
tinction between ‘history and computing’ and ‘historical computing’. The former
concerned the contributions of computation to history. The latter has to deal with
all the ‘grubby practicalities of hardware and software. It will have to deal more with
applying what information science has taught us.’ Implicitly he criticised κλειω: ‘If
we want to make software for historians, we must stop producing programs that
require attending several summer schools before you can work with them’ (Bos and
Welling, 1995).
• Jan Oldervoll shared Welling’s interest in historical tools with good interfaces, and
created CensSys, a historical software package for analysing census data. Although
CensSys was primarily designed as a fast and specialised tool for a specific kind
of source, it was also based on clear ideas about interfacing between programs.
Accepting its necessary limitations, it had provisions for delegating tasks to other
programs; including an interface to κλειω (Oldervoll, 1992, 1994).
• Breure created socrates, a framework on top of dbase, consisting of a program
library and tools (like program generators), that helped to build historical database
applications with this popular database package. socrates comprised not only
software, but also a few guides (‘grey publications’) about historical data modelling.
It particularly focused on problems with irregularities of source structures versus
the strict demands of the relational model, and on the handling of text embedded
factual data, like mentions of people, events and objects in wills and deeds (Breure,
1992, 1994a, 1994b).

35 The past
 3.5 Main topics in historical information science
Within the domain of historical information science, dozens of research themes have
attracted attention from hundreds of historians and information scientists over the
last 25 years. Some information problems were solved in the meantime; others have
come up, while some other problems are still being discussed. There are also a few
problems that have never attracted much attention, although they seem to fit very
well into the domain.
At this point, we could present a detailed historiography of history and computing,
as unfolded through numerous project papers in the proceedings of the several con-
ferences, journals and workshop volumes. However, this has been done already by
others: (Denley, 1994a; McCrank, 2002; Speck, 1994; Woollard, 1999). Nevertheless,
it is useful to get an idea of all issues that have been discussed, rejected and (could
have been) achieved within the domain of history and computing. The issues are
grouped according to the kind of data to which they are related: textual data, quantita-
tive data and visual data.

3.5.1 Databases and texts: From documents to knowledge

3.5.1.1 Databases and text in humanities research

Discussing databases and text in a single section may seem like a hasty attempt to
cover quite dissimilar kinds of computer application in an overview, without doing
justice to the complexities of each specific domain. The term ‘databases’ has strong
connotations with relational database systems, tables and relatively small, highly
structured chunks of data, while ‘texts’ is usually associated with full text and text
analysis software for producing concordances, word counts, metrics for style, and,
more recently, with text databases, supporting storage and publishing Web docu-
ments. ‘Structuring’ in classical databases means putting data in fields, while text is
structured by ‘marking up’ information through the insertion of tags or other codes.
For years these concepts have corresponded to different strains in the evolution of
software, fostering an archipelago of database and text islands, where bridges were
hard to build.
Recent technological developments, especially the advance of xml, have blurred
this distinction. xml is widely used in business applications for the exchange of
database data, which can be merged with running text, if required. xml-data may be
stored in relational database management systems, or in dedicated object database
systems. The landscape is changing now, and the islands are growing slowly together.
Historical information, once liberated from its paper substratum, may subsequently
‘reincarnate’ in quite different shapes. So, the current state of technology is the first
good reason for a combined approach.
Roughly speaking, ‘databases’ and ‘texts’ have also stood for different areas of
humanities computing. We may safely say that database themes, in some form or
another, dominate the majority of publications on historical computing, while text
applications and text analysis abound in the field of literary computational studies.
Of course, this image is stereotypical, but it still holds true and it would indeed justify

36 The past
 two separate chapters. The second reason for combining is an attempt to gain a broad
overall view of the entire field of storing and creating digital representations of his-
torical and cultural documents. Facing the abundance of literature dealing with the
peculiarities of database and text problems in the unique context of specific sources,
it seems worthwhile take a few steps backwards, and, from a greater distance, to look
for similarities and comparable methodological problems and strategies at a little
higher level of abstraction. Because historical and literary studies have had their own
publication channels, we should alternate between both fields if we want to discover
parallel developments.
A few words of caution. In spite of this wide angle view, the reader should be
warned that this section will not provide a balanced encyclopaedic description (for
that purpose, one may, for example, refer to (McCrank, 2002)). Our main questions
will be: What has been done so far? Where has it got stuck? What has to be done in
the near future to ensure scientific progress? The underlying methodological ques-
tion is, how historical data processing on this basic level of storage and transforma-
tion can be further streamlined, taking advantage of current developments in other
disciplines, in particular information science and computer science. In search of
answers, we will be selective, outlining primarily the developments at the historical
side and looking for matching parallels in the literary field. On the other hand, some
parts of this section may look too detailed and even trivial to humanities scholars.
Because the text is not intended for this audience alone, a discussion of distinctive
computational techniques will only make sense to people outside our field, if they are
accompanied by a few introductory comments on characteristics of humanities stud-
ies in general.

3.5.1.2 Historical studies versus literary studies

Although both historical and literary studies use textual material as an object of
study, methodologies and techniques are quite different in many respects, which
has likewise resulted in differences in computational approaches. Literary research
has been focused on critical editing and analysis of the texts themselves: the recon-
struction of the lost common ancestor of surviving texts through the examination of
manuscripts, the comparison of one text with another to discover textual variation,
and the emendation of the text reconstructed, thereby removing errors. The analytical
tradition has created a wide array of techniques and tools for textual analysis, varying
from lemmatised word lists and concordances, to programs for stylistic and thematic
analysis.
History is mainly based on written sources, which have, strictly speaking, also the
form of texts. However, in contrast with literary studies, historical research prob-
lems pertain less to the texts themselves, but are more related to the historical reality
beyond the documents handed down. Traditionally, historical questions will direct
the attention to the factual elements in the texts, to person names, dates of birth and
death, to belongings or to cargo carried in ships or to the departure and destination
of voyages. Texts perceived in this way tend to lose their textual qualities and will be
regarded as lists of data.

37 The past
 Where administrative documents are concerned, this point of view is mostly cor-
rect and efficient; however, the domain of historical sources is fuzzy and full of excep-
tions. Text features are indispensable links in the chain of historical interpretation
and may be still be relevant in a later stage of research. They may reveal characteris-
tics of unknown authors and of the history of the manuscripts themselves. A rare,
but very good example is the use of cluster analysis, applied to strokes of letter forms
occurring in undated manuscripts, written by the same scribe, in order to establish
the probable date of completion (Friedman, 1992). Some highly relevant notes can be
found in margins, a quite regular source may suddenly lose its structure at a certain
point, or interesting information may be appended to regular data in an unexpected
way, for example these cases in a British census list:

‘The ability to include apparently insignificant and microscopic detail from the
census within the κλειω dbms has important macroscopic implications. For ex-
ample, the refusal of a young woman to reply to the question on occupation in the
1881 census for Winchester, coupled with the comments of the enumerator, whose
definition of her occupation as ‘on the town’ (implying prostitution) provides an
important glimpse behind the curtain of the surviving sources – the enumerators’
book, and towards an understanding of the process through which the original cen-
sus schedules (which have not survived) were transformed into the documents we
have today. Conversely, a two-line entry in the census which reads ‘Assistant Classi-
cal Master/ba Trinity College Dublin’ which is reduced by the editorial pen to ‘Prof’
helps the researcher to grasp some of the smoothing out process of categorisation
which went to contribute to census statistics overall.’ (Burt and James, 1996)

However, traces of human life have not always been recorded in the form of lists.
Charters, probate inventories, notarial deeds and wills form a category of sources
which is not easily positioned on the sliding scale from structured data to running
text. Both factual data and the surrounding text may be of interest and should be
stored, therefore. In a subsequent stage, the former type of data will be used for
quantitative analysis, while the latter kind of information may prove to be valuable for
correct interpretation of individual cases. In 1995, History and Computing dedicated a
special issue29 to probate inventories and wills, which shows the struggle with these
ambivalent data structure requirements. Some researchers preferred to record the
entire text structure, while others chose a middle path, entering data verbatim from
the source, but without preserving the grammatical structure of the text. Software
varied from special packages (table-based), to Paradox for Windows and at that time
popular text database systems such as AskSam (Litzenberger, 1995; Morris, 1995;
Overton, 1995; Schuurman and Pastoor, 1995; Webb and Hemingway, 1995).
In addition to documents that have been produced by administrative systems in
the past, a substantial part of historical research is based on narrative sources, like
chronicles, biographies, diaries, journey accounts, treatises, political pamphlets, and
literary works. Here, historical, literary and linguistic scholars share a considerable
amount of material, however, with distinctive intents, which has important conse-

38 The past
 quences for information modelling and data processing.
The well-known diaries of Samuel Pepys (1633-1703)30, a highly personal account
of seventeenth-century life in London, are first of all of historical interest, but have
also given rise to linguistic studies, and his work appears in literature courses, for
example, in the broader context of studying the emerging modern expression of self-
identity in the literature of his age. Historians will isolate and label historical events
in the digital text, with mark-up for persons, places and dates, and prefer storage
in a database for sorting and easy look-up, preferably with links to the original text.
Alternatively, the voluminous text may be scanned first for themes of interest in order
to locate relevant passages (Louwerse and Peer, 2002). Techniques like text mining,
Topic Detection and Tracking, and Text Tiling, developed in information retrieval,
could be helpful.
A study of Pepys’ linguistic usage itself will require counting of words, phrases
and linguistic constructions – Pepys sometimes used a kind of private code involv-
ing words from Spanish, French and Italian, obviously for reasons of concealment,
hiding text from the casual browser. Stylometric analysis could help to cluster certain
parts of these diaries in order to test hypotheses about the author’s distinct character-
istics in certain periods. A complex phenomenon as self-expression can be studied
in a quantitative manner by applying content analysis techniques, searching for and
counting key words and phrases related to this subject.
This example demonstrates that decisions about text encoding are far from easy, if
a text is to be reused for various purposes. An important part of the discussion from
both sides concerns what level of digital representation is adequate for historical and
cultural source texts. In the historical field, this discussion has centred around the
dichotomy ‘source-oriented versus model-oriented’. In the domain of literary studies
its counterpart is to be found in the debate on the nature of the critical text edition.

3.5.1.3 The critical edition in the digital age

background
Both history and literary studies share a reliance on high quality editions of textual
sources. From the nineteenth century onwards history as a discipline has been based
on an important tradition of printed critical source editions, including an extensive
apparatus of footnotes, which explain and comment on the main text. The applica-
tion of information technology has led to reflection on the nature of scholarly source
editions. Some works have got a digital companion in addition to the printed book
(e.g. as pdf-file or cd-rom)31 In other cases, an on-line database will be an obvious
solution, as with the material concerning the Dutch-Asiatic trade of the Verenigde

29
History and Computing 7:3 (1995), p. iv-xi, 126-155.
30
See, for example, http://www.pepys.info
31
For example, Kroniek van Peter van Os. Geschiedenis van ’s-Hertogenbosch en Brabant van Adam tot 1523,
A.M. van Lith-Droogleever Fortuijn, J.G.M. Sanders & G.A.M. van Synghel ed. [Instituut voor Nederlandse
Geschiedenis], Den Haag (1997).
32
Instituut voor Nederlandse Geschiedenis: http://www.inghist.nl/Onderzoek/Projecten/DAS
/EnglishIntro

39 The past
 Oostindische Compagnie (voc)32 Voluminous printed editions, being the editor’s
lifework, are difficult to uphold. Moreover, information technology has liberated the
critical source edition from the constraints of the printed book.
‘Comprehensiveness’ is an important goal, but cannot always be attained by time-
and paper- consuming full-text editions. That is why the Institute of Netherlands His-
tory (Instituut voor Nederlandse Geschiedenis) decided to publish the correspondence
of William of Orange, comprising approximately 11,000 letters, in the form of short
summaries, carrying metadata and linked to digital images of the original sources.
Another example is the edition of the Resolutions of the Dutch States General. An
electronic edition with full-text search facilities would have been attractive (search-
able on-line as with the Papers of George Washington, president from 1789 to 179733,
and with the scientific archive of Samuel Hartlib, c. 1600-166234, which has been
published as a cd-rom edition of text images with transcriptions), but bare text-re-
trieval software doesn’ t handle these kind of seventeenth century texts very well,
because they contain many spelling variants and terminology quite different from the
concepts modern historian are looking for. Therefore, for the time being, one of these
projects has settled for a compilation of summaries in modern Dutch (Haks, 1999).
More systematically, a useful classification is that into ‘digital facsimiles’, ‘digital edi-
tions’, and ‘digital archives’ (Thaller, 1996).

• Digital facsimiles
A digital facsimile provides access to an individual hand-written or printed source
text, by means of scanned images, a complete transcription, and a linked database
of persons, locations and concepts (specifically ‘time’) mentioned in the text. Optio-
nal extras for historical sources are other tools, which document former calendar
systems, currencies, and specific terminology. Within literary studies, the term
‘image-based computing’ has a special meaning. It may be said to descend from
the so-called social or materialist theories of textual production advanced by such
scholars as McKenzie and McGann in the early eighties, coupled with the means to
create (relatively) high-quality and (relatively) low-cost digital facsimiles of docu-
ments. A wide array of special applications is grouped around this concept, which
Kirschenbaum in his introduction of a special issue of Computers and the Humani-
ties has called ‘venue for representation’ (Kirschenbaum, 2002).
A nice example is the work of William Blake, an eighteenth century poet, who il-
lustrated his own work in watercolours, later printed from a copper plate. In a certain
sense, Blake created an eighteenth-century multimedia presentation: he used print-
ing as a mode of production rather than reproduction, etching himself and combin-
ing text and illustrations on the plate for the first time rather than reproducing a
pre-existent page design. Editorial problems arise with numerous, later impressions,
often of an inferior quality, all more or less different. It may be obvious, that a mere
transcription does not do any justice to the work’s artistic qualities and that the great

33
American Memory Project: http://memory.loc.gov/ammem/mgwquery.html
34
The Hartlib Project: http://www.shef.ac.uk/hri/projects/projectpages/hartlib.html

40 The past
 number of variants are not easily reduced to a single edition. A natural solution was
a digital facsimile edition, with trustworthy reproductions of the illustrated text and
full transcription in sgml. An advanced image manipulation tool enabled resizing
to actual size at any monitor resolution. The images could be examined like ordinary
colour reproductions, but could also be displayed alongside the texts, enlarged, com-
puter enhanced, juxtaposed in numerous combinations, and otherwise manipulated
to investigate features (such as the etched basis of the designs and texts) that were
previously imperceptible without close examination of the original works (Viscomi,
2002).

Figure 3.1 The Blake Project: ‘The Tyger’, Songs of Innocence and of Experience copy C (Vis-
comi).

• Digital editions
A digital edition goes a step further, providing access to different versions of the
text, realising the aims of the critical edition in digital form. A good example, from
the literary domain, is The Canterbury Tales Project (De Montfort University) which
aims to:
− Publish transcriptions of all the manuscripts and early printed books of Chaucer’s
Canterbury Tales into computer-readable form (eighty-four manuscripts and four
printed editions survive from before 1500!).
− Compare all the manuscripts, creating a record of their agreements and disagree-
ments with a computer collation program.
− Use computer-based methods to help reconstruct the history of the text from this
record of agreements and disagreements.

41 The past
 − Publish all the materials, the results of the analysis, and the tools which were used
by the researchers (materials are available both on-line and on cd-rom). 35

Figure 3.2 The Canterbury Tales Project: Facsimile with transcription of the second Caxton
edition. The search box with Find-button will produce a kwic-index for a given key word.

Figure 3.3 The Canterbury Tales Project: Word by word collation of whole text of both
Caxton editions, with all differences highlighted in red.

35
Project: http://www.canterburytalesproject.org/

42 The past
 • Digital archives and digital libraries
A digital archive (or virtual archive) is characterised by making a large collection
digitally available through a variety of retrieval mechanisms, linking different
databases behind a uniform user interface, with additional tools for analysing in-
formation (e.g., mapping software) and offering some personalization. Because of
the larger scale the granularity of disclosure will vary greatly. The term has strong
connotations with administrative documents, which shows to full advantage in pro-
jects like the computerisation of the Archivo General de Indias (González, 1998)36,
which holds all documents concerning the Spanish administration in the Americas
and the Philippines, the Duderstadt Project (Aumann, Ebeling, Fricke et al., 1999),
which is developing a computerised version of the files of the municipal archive37
and the digitalisation of the medieval archive of Regensburg (the Fontes Civitatis
Ratisponensis – fcr).38
However, the term is also used for different kinds of collections, e.g., the Thomas
Jefferson Digital Archive (containing mainly letters), the Codices Electronici Ecclesiae
Coloniensis (ceec – digitized manuscripts of the church of Cologne) and Prometheus
(a digital image archive), and the World of Dante, a hypermedia visualization of the
work of the famous poet (Parker, 2001).39

There is only a vague borderline between digital archives and digital libraries such as
Perseus. Organizations like the Council on Library and Information Resources (clir)40
and Digital Library Federation (dlf)41 do care for both, subsuming them under the
comprehensive notion of networked digital repositories.

electronic textuality and the new philology

In the literary domain, the advances of information technology have given rise to the
concept ‘electronic textuality’, grouping together issues about digitized material and
the potential of hypertext techniques for electronic text editions (Abram, 2002). A
great deal of current electronic texts has been encoded according the guidelines of
the Text Encoding Initiative (tei). The tei guidelines uses descriptive mark-up, e.g.
<title>, <chapter>, <paragraph>, in contrast with mark-up that specifies formatting
procedures or presentation through fonts and colours. The tei convention has been
implicitly based on an underlying text model, which defines a text as ‘a hierarchy of
content objects’ : as composed of nesting objects such as chapters, sections, para-
graphs, extracts, lists, and so on. Although practical and appropriate in many cases, it
has provoked extensive discussions, in particular with regard to overlapping hierar-
chies (depending on multiple points of view) and texts that lacked a clear hierarchical
structure, such as poems (Ide, 1995; Renear, Mylonas and Durand, 1993).
36
Archivo General de Indias: http://www.clir.org/pubs/reports/gonzalez/contents.html
37
Duderstadt Project: http://www.archive.geschichte.mpg.de/duderstadt/dud.htm
38
Fontes Civitatis Ratisponensis: www.fcr-online.com
39
Thomas Jefferson Digital Archive: http://etext.lib.virginia.edu/jefferson/
ceec: http://www.ceec.uni-koeln.de/
Prometheus: http://www.prometheus-bildarchiv.de/
The World of Dante: http://www3.iath.virginia.edu/dante/
40
clir: http://www.clir.org/
41
dlf: http://www.diglib.org/

43 The past
 By the mid-1980s, however, a more fundamental criticism came from textual schol-
ars like McGann and Shillingsburg, who viewed a text primarily as a product of social
interaction between a number of agents: author, editor, publisher, composer, scribe
and translator (Schreibman, 2002). This has started a debate about the form and
function of the critical text edition. The so-called New Philology has questioned the
role of the editor in favour of the position of the reader. It fits into the post-modern
thinking against all forms of authority and pays more attention to the historical situa-
tions of texts, their function in time and place, and to the interaction with their social
context (the ‘textual turn’). It no longer sees different versions of a text as witnesses
of a lost original, which has to be reconstructed from variants, found in extant copies.
Not a reconstructed text, but a diplomatic transcription of texts handed down, has to
be the basis of an edition (Kobialka, 2002; Ott, 2002; Robinson, Gabler and Walter,
2000).
Of the scholars working in this area McGann and Landow have been especially
influential. McGann’s ideas about ‘social editing’ and ‘hyperediting’ emphasise the
value of hypertext and hypermedia in relation to the social aspects of literary texts
(McGann, 1991, 1992, 1995, 2002). Landow argued that ‘the dispersed text of hy-
pertext has much in common with the way contemporary, individual readers of, say,
Chaucer or Dante, read texts that differed from one another in various ways’ (Landow,
1996?). Editing becomes an ongoing process by means of collaboratories, where
readers can play an important role in on-line annotation.
The consequences of this new paradigm for editorial practices and tools have been
clearly expressed in the Electronic Variorum Edition of Don Quixote, which has as
primarily goal ‘to develop a replicable program that permits the creation of online
critical editions as hypertextual archives, using the Quixote as test bed.’strictly speak-
ing, it is no longer an edition, but a dynamic, hypertextual archive composed of a
series of databases with special tools, such as a text collator, a viewer for displaying
digital images of the text and transcriptions side by side, annotation and update facili-
ties for texts and stored information objects (Urbina, Furuta, Goenka et al., 2002).
Of course, the liberal publishing of documents and empowering the reader will
create new problems: digitisation should not become a substitute for scholarship,
and the new means of cheap distribution poses the question how to select documents
and where to stop. Moreover, as Prescott has remarked in the context of the Electronic
Beowulf, the impression has always been that digital images will be free or at least
very cheap, thanks to governmental grants. The free ride will come to an end, when
digitisation projects have to recover their costs (Prescott, 1997). This uneasiness
has led to new methodological solutions, as formulated, for example, by Vanhoutte
(the Streuvels Project), who made a distinction between the archival function and the
museum function.

44 The past
 $IGITALåARCHIVE

%JHJUBM
GBDTJNJMFT %DITION

#JCMJPHSBQIJD
EPDVNFOUBUJPO

%JQMPNBUJD
USBOTDSJQUJPO

3'-,4%)

Figure 3.4 The Archive/Museum model by Vanhoutte.

The archival function is responsible for the preservation of the literary artefact in
its historical form and for documenting the historical-critical research of a literary
work. The museum function pertains to the presentation in a documentary and
biographical context, intended for a specific public and published in a specific form
(Vanhoutte, 1999).
It is, of course, beyond the scope of this report, to draw up a balance. As for text
projects it appears that hypertextual techniques have captured a core position in
the editorial field. New forms of source editing have been established, varying from
digital facsimiles to digital archives and digital libraries, all equipped with an array
of dedicated tools, changing the traditional role of the editor, and empowering the
reader, but without making the editorial rigor obsolete.
For historians, the critical edition is the most complete way of making material
available, in comparison with other forms of digitisation discussed below, and, there-
fore, it has been presented first. Application of computer technology creates new
capabilities in disclosing information, however, automation should be used also to
make the process of disclosure less time-consuming (e.g., the previously mentioned
letters of William of Orange and the Resolutions of the Dutch States General). There-
fore, progress with this kind of publication is closely linked with solving problems of
digitising larger quantities of historical data for analysis purposes.

3.5.1.4 Digitising historical data for analysis

Unfortunately, the majority of historical research material will never be available in
the form of a critical edition. Although libraries and archives on the one hand and

45 The past
 researchers on the other hand may co-operate more in digitising, a considerable ef-
fort in this respect will be left to historians themselves. In this category project aims
and constraints are usually quite different from those with critical editions discussed
above. Nobody wants to put more effort in data entry and data processing than neces-
sary, however, the complexity of historical research, which sometimes comes close
to the work of a detective, may make it difficult to determine what ‘necessary’ exactly
means. Thaller has summarised neatly the nature of problems with representing
historical sources in a discussion on digital manuscripts:

‘Speaking on the most general level, we consider a text to be ‘historical’, when it

describes a situation, where we neither know for sure, what the situation has been
‘in reality’, nor according to which rules it has been converted into a written report
about reality. On an intuitive level this is exemplified by cases, where two people
with the same graphic representation of their names are mentioned in a set of
documents, which possibly could be two cases of the same ‘real’ individual being
caught acting, which, however could also be homographic symbols for two com-
pletely different biological entities.

At a more sublime level, a change in the colour of the ink a given person uses in
an official correspondence of the nineteenth century could be an indication of the
original supply of ink having dried up; or of a considerable rise of the author within
the bureaucratic ranks.

Let us just emphasize for non-historians, that the second example is all but artifi-
cial: indeed the different colours of comments to drafts for diplomatic documents
are in the nineteenth century quite often the only identifying mark of which diplo-
matic agent added which opinion.’ (Thaller, 1996)

A clear formulation of the research problem and limitations in time and money will
usually dictate practical solutions with regard to entering ‘what’ and ‘how’, frequently
overruling more theoretical considerations of potential reuse. Although these prob-
lems with data representation are generally well known in the historical community,
we shall briefly review the implications surrounding these problems:

• The orthography of data

The spelling of names of persons and places, the denomination of currencies and
the amounts expressed in old weights and measures may require further standardi-
sation in order to be analysed statistically. An important decision is at which stage
this standardisation has to be carried out and how the coupling with the original in-
formation is preserved. The main problem behind it is the linking of source data to
the historical entities as we know them: for example, to which geographic location
does a certain placename refer, and which variant names correspond with the same
historical person? Precisely because such answers may require further research,
name standardisation should be avoided at data entry. Preferably, it should be pos-
sible to repeat the standardisation process automatically with different rules after-

46 The past
 wards. A closely related problem is the reading of historical data (i.e., the particular
way of understanding what is written). Sometimes individual data are difficult to
separate from their context. A qualification as ‘smith’ may indicate an occupation or
a surname: what should be stored where?

• The data model

A major point of discussion has been, whether the design of historical databases
should reflect the model of the historical situation as envisaged by the researcher,
or whether it should reflect the structure of the sources themselves. A historical
source is a complex network of information, usually well understood by the histo-
rian, but difficult to represent in conventional storage systems. Although relational
database systems allow the implementation of hierarchical data models (typical
for texts), several kinds of historical material do not fit into this scheme easily. In
addition, data structures may vary over a group of sources of the same kind, or even
worse, within a single source. This makes conventional data modelling42 a precari-
ous matter in those cases (for examples: (Burt and James, 1996)).

However, it becomes much less of a problem, when the research model, rather than
the structure of the source, is mapped onto the database design, e.g., an economic
market with suppliers, goods and consumers, or employment in an occupational
group (Bradley, 1994). The particular form of choice made depends largely on the
nature of the project and the refinement in computer assistance. The tension be-
tween the complexity and the irregularity of the data structure of historical sources on
the one hand, and the rigid nature of the relational database model in combination
with analysis purposes on the other has fuelled the discussion about source-oriented
software in the 1990s decade (see below).

• Linking data with the original source text

In historical discourse historians are used to refer painstakingly to the informa-
tion’s place of origin in the underlying sources. This approach has also affected
the architecture of historical computer applications that are used for research
purposes. Although source data have to be converted to a form suitable for specific
processing (e.g., statistical analysis – see below) and will appear at different levels
of aggregation, many projects have demonstrated the need to preserve links to the
deepest level, i.e., to the original source text. This can be realised by simply linking
scans of the source document to the respective records in the database, or by ad-
ding the source text in transcription. The latter may be realised through the exten-
sive use of text fields, or with the help of text look-up mechanisms and other built
in applications (Breure, 1995a, 1995b).

source-oriented versus model-oriented

As described in Section 3.4, in the 1990s a part of the historical community had
become aware of the typical problems inherent to historical data processing. Harvey,
Press, Greenstein and Burnard have done a lot to make historians acquainted with
42
Data modelling: designing the structure of the database.

47 The past
 the theoretical principles underlying relational database systems (rdbms) (Burnard,
1989, 1990; Greenstein, 1989; Harvey and Press, 1992, 1993, 1996). The discussion
about ‘whether or not to use an rdbms’ tended towards a distinction between differ-
ent kinds of projects. Some rules of thumb were formulated to help fellow historians
make the right choice.
Denley did so in his balanced article on historical database design in the 1990s
(Denley, 1994a). The model-oriented approach is for researchers with specific ques-
tions, using regular sources, accepting some arbitrary decisions about data and with
quantitative analysis at the forefront of his/her intentions. A tight schedule and
mainstream tools may be other arguments to choose this path. On the contrary, the
source-oriented approach is more appropriate when the historian places high priority
on maintaining the integrity of the source, wants to treat his material both as text and
structured data, aims at a database that comes close to a source edition and has more
time to spend on complex tools.
About the same time this point of view was demonstrated in practice by Bradley in
a reconstruction of the British medical profession. He chose deliberately a relational
database, accepting that the model would be a simplification of the historical reality,
and nothing more than an attempt to replicate the structure of employment (Bradley,
1994). An rdbms is fine, if a project aims at data analysis, rather than at source analy-
sis, the data are open to aggregation and therefore to be used in statistical analysis
and relationships between the database objects can be described as ‘one-to-many’.
Two years later Burt and James praised the superiority of κλειω, its fluidity and flex-
ibility at data entry and emphasised the macroscopic relevance of microscopic details:

‘Thus it is argued here that source-oriented models set a benchmark for historical
studies in computing. This benchmark is above the level attainable by the rigid
and exclusive technique of the relational database. There may indeed be a different
mindset in operation in the source-led and source-oriented approach of historians
when compared to the mindset of certain practitioners of computing and data
mining in business and computer science. The microscopic detail in the historical
source can prove to be of key significance in macroscopic results.’ (Burt and James,
1996)

For the remainder of this period both streams have run in parallel. It is beyond the
scope of this report to quantify the market share of each.43 Choice for the one or the
other has largely depended on the particular interests and preferences of the re-
searchers as described above.
The situation in about 2000 was summarised well by a Swiss PhD-student,
Christian Folini. Having a good overview of current computer practice in the his-
torical field, he discussed the needs of historical (graduate) students on basis of a
small e-mail survey (Folini, 2000). He found a preference for relational databases,
particularly for Microsoft Access. His rules of the thumb for selecting a strategy form

43
This does not mean, however, that relational desktop databases as FoxPro, FilemakerPro and Microsoft
Access have been opposed to κλειω alone. A minor position in the source-oriented camp has been hold by
full text systems such as TACT, Idealist, Wordcruncher, and Atlas.ti.

48 The past
 a correlate to the scheme as described by Denley. He also found a group of research-
ers who used relational database and worked simultaneously with full-text systems.
His own research was about female mystics in southern German convent of the
thirteenth and fourteenth centuries, encompassing also full-text material. Finally, he
based his solution on Access, with text excerpts entered in text fields of the relational
database. His complaints were about difficulties in estimating the time required for
applying information technology, the lack of technical support, the rapid succession
of software releases, and the technical limits of the basic versions of software, usu-
ally installed at universities, and concluded with hope on the unifying role of xml.
His account is interesting, because his approach was open-minded and it demon-
strates a serious lack of reliable and dedicated tools, especially for a younger group of
researchers who are willing to apply computer techniques, but who have to cope with
severe constraints in time and lack of straightforward methodology.

multiple options: what is the problem?

How should we assess this situation? Can we conclude that the historical commu-
nity has found appropriate practical solutions for natural difficulties inherent in the
historical craft, and that there is not any serious problem at all? The answer is both
yes and no. ‘Yes’, because, in daily practice – as Folini himself demonstrated – one
can manage to get on with the various options regarding source-oriented and model-
oriented. ‘No’, from a methodological point of view, because the gap between model-
oriented and source-oriented requires far-reaching decisions in an early stage of a
project. A versatile approach would be more attractive, leaving open various options
of database processing, statistical analysis and text analysis, without enforcing redun-
dancy in the machine-readable source material, which easily happens when the same
material must be encoded in different ways for different packages. In addition, our
research agenda and related assumptions with regard to encoding and storage should
be reversible within reasonable limits. This will be feasible only if the effort in digitis-
ing and data entry can be considerably reduced by applying in this stage computer
techniques on a larger scale. Processing data requires structure. This may be created
manually, but it is a challenge to discover to what extent information technology can
be deployed for that purpose. There is an additional reason why ‘the gap’ is to be con-
sidered harmful: accepting the current situation without any further action will also
widen an other gap, the one between historical information science and computer
science, thus leaving ample ground for many projects to keep reinventing the wheel.
These ideas are far from new and return us to the very beginning of the discipline.
In 1980 Thaller outlined already the ideal trajectory when he introduced clio (the
original name of κλειω): historical data processing should move through different
stages, starting with a data representation as close as possible to the original source,
to more practical formats suitable for analysis and retrieval (Thaller, 1980). κλειω
comes with a so-called ‘logical environment’, comprising several rule-based algo-
rithms for pre-processing raw source material.44

44
See for example the Introduction of the on-line tutorial: http://wwwuser.gwdg.de/~mthalle2/manual/
tutorial/intro.htm

49 The past
 The fact that after almost twenty years since its introduction not every historian is
working with κλειω cannot be explained by theoretical inadequacies with regard to
the practice of historical research. Neither has the system been criticised for its lack
of power. However, user unfriendliness, a steep learning curve, the feeling of a black
box deploying a technology far away from generally accepted standards and main-
stream computing have kept many potential users away, especially in a community
that has not been particularly fond of computers at all (Denley, 1994a; Everett, 1995).
It must be admitted that this criticism is not (fully) justified. Much has been based on
blunt misunderstanding and lack of interest. In the meantime, κλειω has evolved as
well: it is now web-enabled and has learned to speak xml. κλειω uses a very general
data model, related to the semantic network, which can represent xml data structures
as a subset (a capacity which is not immediately clear from the outdated documenta-
tion available on the web site).
Irrespective of whether one wants to use κλειω or not, the problem does not seem
to be that we have no idea how to bridge gap. Greenstein and Burnard pleaded for
joining the two approaches in what they called the ‘textual trinity’ of (1) printing,
publishing, and word-processing; (2) linguistic analysis and engineering; (3) data
storage, retrieval, and analysis. They showed how tei solutions could be used to cre-
ate machine-readable transcriptions comprehensible for different applications. They
demonstrated the power of tei in encoding multiple and even competing interpreta-
tions of text (Greenstein and Burnard, 1995). The main drawback of this solution is
the effort required: in most cases, there are simply more sources to be used than we
are able to encode (remember: we are not envisaging here critical editions, but data-
sets for a particular research project).
Looking backward and tying up loose ends following from the variety of arguments
above, the core question seems to be: how can we create structure in historical mate-
rial in a way that is:
1. appropriate to the source’s complexity,
2. modular, i.e., split into discrete and transparent steps, each preferably well-docu-
mented and clearly modelled, which
3. adheres to standards and
4. allows a critical choice of the best tools / techniques available (where ‘best’ implies,
among others things, ‘appropriate’, ‘well documented’ and ‘user-friendly’),
5. without spending an unwarranted amount of time either to manual encoding or to
developing complex technological solutions.
Commercial database systems have satisfied criteria 2-4, but cannot easily represent
the complexity of structure (criterion 1), and will therefore only be a good solution
in those cases where structure is relatively simple and regular or when it has been
defined as such within the project’s framework. Scanning and manual xml encod-
ing of full-text is unrealistic for many mainstream historical projects, due to point 5
This last requirement has been precisely the major motivator for developing κλειω,
although this system seems to have suffered from an image of being bound to a non-
standard, monolithic solution, thus failing on criteria (2), 3-4.

50 The past
 3.5.1.5 Automatic structuring
Creating structure and linking the structured text to semantic models (e.g., authority
lists of persons, places etc.) are essential for historical data processing, however, it is
desirable to automate this process to a large extent. Recently, technological solutions
for transforming raw text into encoded information by means of (semi-)automatic
techniques have made good progress. A few long-lasting, large-scale projects seem
to be well on the way to satisfy all of the criteria mentioned above to great extent.
They draw upon a combination of techniques from different domains, like natural
language processing, text mining, speech recognition and corpus linguistics. These
solutions are mostly not (yet) disseminated in a form of ready-made tools that can be
easily used by others, like concordance programs and statistical packages. The variety
of original methodological contexts makes it far from easy to decide ‘what’ to use
‘when’ and ‘how’ in specific historical research. Although they certainly fall outside
this category of ‘digitising for practical purposes’, they indicate a promising direction
for future methodological and interdisciplinary methodological research.

perseus
The Perseus Digital Library45 is not only interesting as an on-line source of a wealth of
information, but also because of methodological aspects, in particular the transfer of
methods and techniques from one domain (antiquity) to another (modern time, the
history of the city of London), covering different languages (Greek, Latin, English,
Italian, Arabic), and its intention to formulate a more widely applicable strategy of
digital collection building, together with a generalised toolset. Perseus has a phi-
losophy of starting with simple mark-up (concerning morphology, or the tagging of
proper names of places and persons on the basis of different authority lists), succes-
sively taking advantage from each information layer, without immediately striving
for a perfectly encoded text. Crane has well documented this strategy (Crane, Smith
and Wulfman, 2001):

‘Automatic tagging takes place in two steps, of which only the first has been fully
implemented. In the first step, we look for proper names but make no attempt to
resolve ambiguities. We tag ‘Oliver Cromwell’ as a personal name but do not try to
determine which Oliver Cromwell is meant, nor do we look for instances such as
‘the Oliver Cromwell’ (which might refer to a building or institution named after
the historical figure).

Once possible proper names have been tagged, there are various strategies to
analyse the context and rank the different possible disambiguations. Our energy at
this stage has focused on acquiring and, where necessary, structuring the data that
we have had to enter ourselves.… The human editor could also enter at this stage,
going through the automatically tagged text. Ideally, the editor would find most
features properly identified and would have to intervene in only a small percentage
of cases.
45
Perseus: http://www.perseus.tufts.edu/

51 The past
 But even without disambiguation or hand-editing, we have been surprised at how
useful the subsequent electronic environment has proven. We consider this to be
an important finding in itself because the performance of a system without clever
disambiguation schemes or expensive hand editing provides the baseline against
which subsequent improvements can be measured. Our experiences suggest that
both clever disambiguation and hand editing will add substantial value to docu-
ments in a digital library, but, even failing those two functions, the automatically-
generated tags can be employed by useful visualization and knowledge discovery
tools.’ (Crane, 2000)

From the beginning Perseus has adhered to standards (sgml, xml, tei, relational da-
tabases). Originally it did not spend much effort in programming (one programmer
until 1994) and developed only one important piece of software: a rule-based system
to analyse the morphology of inflected Greek words (later extended to Latin and Ital-
ian). Gradually, more tools have been created and made publicly available on-line.46
By around 2000 the toolset comprised sophisticated full-text searching facilities, the
creation of links among documents in the system, extraction of toponyms and the au-
tomatic generation of maps, discovery of dates and the dynamic display of timelines,
the automatic implicit searching and discovery of word co-occurrence patterns, and
linkages to morphological analysis (Rydberg-Cox, Chavez, Smith et al., 2002; Smith,
Rydberg-Cox and Crane, 2000).
These tools enable the generation of a knowledge layer, consisting of repertoria,
large collections of metadata and comprehensive display schemes. An interesting
example in this context is the inference step from text to knowledge through colloca-
tion analysis. For this purpose the technique of Topic Detection and Tracking (tdt),
developed in information science under guidance of darpa, was tested and adapted
to historical needs.47
tdt aims at developing techniques for discovering and threading together topically
related material from streams of data such as newswire and broadcast news. tdt
systems will aggregate stories over a span of several days into single event topics. The
most significant problem in adapting tdt methods to historical texts is the difficulty
of handling long-running topics. Many historical documents discuss long-running
events, and many users will wish to browse digital libraries at a scale larger than
events of a few days’ length. Moreover, historical texts tend to be discursive, not bro-
ken into discrete date units, and digressive. Even if there is a main linear narrative,
a historian will often digress about events from before or after the main period, or
taking place in another region. These digressions, of course, may themselves provide
information about other events. Last but not least, date extraction is far from easy,
amongst others because dating schemes other than the modern, Western Gregorian
calendar. As a solution place-date contingencies were calculated and several meas-

46
Tools available through the tool page of Perseus itself, and through the Stoa consortium:
http://www.stoa.org/
47
Topic Detection and Tracking: http://www.nist.gov/speech/tests/tdt/

52 The past
 ures of statistical association were tested, to find the best ranking of events in the
presentation of query results (Smith, 2002).

the integrated dutch language database (inl)

The Instituut voor Nederlands Lexicologie (inl – Institute for Dutch Lexicology) in
Leiden48 administers the Dutch linguistic heritage by automatically encoding text and
storing millions of words in databases: the Integrated Language Database covering
the eighth to the twenty-first centuries, and a subcorpus of the international parole
database (a mixed corpus of newspapers text, newsreel footage and periodicals). It
goes without saying that processing texts on such a large scale requires extensive
automation. Raw texts are tagged as much as possible automatically using tei, using
PoS-taggers (PoS: Part-of-Speech) for word classification49 and rule based programs
for sentence splitting. This technology has been developed in natural language
processing and comprises different strategies (e.g., rule-based tagging, memory-
based learning and Markov-models) (Depuydt and Dutilh-Ruitenberg, 2002; Does
and Voort van der Kleij, 2002; Dutilh and Kruyt, 2002; Raaijmakers, 1999).
Also of interest in this area is the testing of this technology on a wide variety of
historical material. General algorithms are tested for appropriateness for domain
specific purposes, thus leading to more refined solutions. This is mostly related to
characteristics of the Dutch language, and has resulted, for example, in the develop-
ment of the PoS-tagger and testing of different underlying models. We see the same
strategy here as in the Perseus project: more general technologies are systematically
tested and transformed into suitable tools.

other examples
There are several converging, more detailed research lines in creating structure auto-
matically. As mentioned before, from the beginning κλειω has had a ‘logical environ-
ment’ with rules and procedures to convert historical data sets in this way. Currently,
it is used in several large-scale digital archive projects like ceec, Prometheus, Duders-
tadt and Regensburg (see above). The application of automatic procedures is reported
to transform the raw source text into an encoded data set by means of rule-based
editing and semantic parsing (Kropac, 1997).
Text-image coupling is essential for facsimile edition. The linking should be
precise. Lecolinet et al. reported progress in automatic line segmentation of scanned
hand-written manuscripts. They developed a semi-automatic approach that lets
the user validate or correct interactively transcription hypotheses that are obtained
from automatic document analysis. This capability facilitates interactive coupling by
pre-segmenting manuscript images into potential line (or word) areas. As hand-writ-
ten documents often have quite a complex structure, it is generally not possible to
process them in a fully automatic way. Consequently, user validation and correction
is needed with most documents (and especially with modern hand-written manu-
scripts) (Lecolinet, Robert and Role, 2002) .

48
Instituut voor Nederlandse Lexicologie: http://www.inl.nl/
49
For an overview, refer to http://www-nlp.stanford.edu/links/statnlp.html

53 The past
 Figure 3.5 Automatic line segmentation in scanned manuscripts (Lecolinet, Robert and
Role, 2002)

From 1997 to 2001 the Institute of Netherlands History (ing) and the niwi have
co-operated in retro-digitising the printed volumes of the Repertorium van boeken en
tijdschriftartikelen betreffende de geschiedenis van Nederland (Repertorium of publica-
tions on Dutch history), covering the years 1940-1991 Being scanned, the biblio-
graphic information had to be corrected, completed and split into database fields
(the repertorium is now a database in Pica-format, which is used by a substantial
number of ministry libraries, mainly in the Netherlands and Germany). In particular
the older volumes caused all kinds of problems due to irregularities in typography
as well as typical ‘book conventions’, such as cross-references to other sections, well
understood by readers but very impractical for database storage. The large quantity of
pages to be processed made the use of automated procedures paramount. The entire
project has comprised several experiments in (semi-)automatic structuring, among
other things, by implementing sub-processes in Perl and using regular expressions
in more advanced text editors as TextPad.50

ocr conversion to text enrichment (pattern matching) text to database

printed bibliography ocr output simple text layout sgml/xml structure databases (pica)

Figure 3.6 Repertorium Project (niwi): Conversion path

50
Information from D. Stiebral. For project information: http://wwwoud.niwi.knaw.nl//nl/dd_nhda/
projects/proj_rep.htm

54 The past
 3.5.1.6 Nominal Record Linkage
Any discussion on texts and databases is incomplete without a discussion of a par-
ticular technique of distilling knowledge from data: nominal record linkage, which
links occurrences of names in different source texts to historical individuals.
The obvious first step in constructing historical knowledge is identifying the
individuals, who left their traces in historical sources. The extensive literature about
this subject, dating back to the early 1970s (Winchester, 1970; Wrigley, 1973), deals
with questions as how spelling variants in sources are to be standardised, to what
degree linkage can be automated, and what level of relative confidence is acceptable
with automated procedures. A final consensus has not yet been reached, in spite of
thorough debates, experiments and testing. For a great deal this may be explained
by the variety of sources and historical problems involved, covering different times
and cultures, thus creating an endless range of peculiarities: poll books, census and
tax registers, baptism and death records, used for a diversity of research purposes as
studying political behaviour, land holding, family reconstruction, life course analysis,
regional demography, and prosopography (i.e. using individual data to describe the
history of a group51) (Adman, 1997; Adman, Baskerville and Beedham, 1992; Davies,
1992; Harvey and Green, 1994; Harvey, Green and Corfield, 1996; King, 1992, 1994;
Ruusalepp, 2000; Tilley and French, 1997; Vetter, Gonzalez and Gutmann, 1992;
Winchester, 1970).
Harvey and Green, studying the political behaviour of eighteenth century inhabit-
ants of the City of Westminster on the basis of poll books, have given a few examples
of the problems one may come across:

‘Some voters moved house during the period; others changed their jobs; surname
spelling was inconsistent; data drawn from printed or manuscript copies of the
original poll books contain transcription errors; and some data have been lost. Each
of these may lead to a failure to link records which relate to a person. Moreover, the
practice of naming sons after fathers, and of those sons inheriting their fathers’
estates, gives rise to the possibility of establishing false links between distinct vot-
ers. This possibility may also arise by personation, as well as by the coincidence of
two distinct voters in successive elections sharing common names, addresses, and
occupations.’ (Harvey and Green, 1994)

Although nominal record linkage is mainly discussed with regard to administrative

sources, the problem is not restricted to this category alone, and has its counterpart
in literary texts as well. In a prosopographical study about the struggles between
Norwegian factions in the period of civil wars (1130-1240) Opheim has constructed a
database from the kings’sagas, encoded in xml. Recording the participants in the civil
wars requires the identification of personal names, place names and group names. A
person may be referred to by his given name and a patronymic (e.g. Dag Eilivsson),
or with some signifiers of status or office (king, archbishop), or individual character-

51
For the different shades of meaning of prosopography, refer to the introduction to the special issue about
this theme in History and Computing 12:1(2000).

55 The past
 istics (Guthormr Grey-Beard), but any combination may occur. As a narrative genre,
sagas provide complications by omitting sometimes a name at all, even if a person is
essential in a story (Opheim, 2000).
The Association for History and Computing dedicated two special issues of its journal
to this theme (in 1992 and 1994). The use of a broad range of software has been
reported, varying from regular database management systems, κλειω, to special pack-
ages as carl (Adman, Baskerville and Beedham, 1992), Famlink (Vetter, Gonzalez
and Gutmann, 1992) and Genesis (Bloothooft, 1995), and differing strategies have
been proposed: rule-based, probabilistic, and learning systems with various degrees
of human interaction.
A great deal of the recent discussion has been centred around the so-called ‘multi
pass algorithms’ in automatic record linkage. These are better referred to as subse-
quently relaxing strategies in comparing nominal records on varying combinations
of data elements (e.g., standardised forename and surname, in combination with oc-
cupation and year of birth – or Soundexed surname plus shortened forename, com-
bined with occupation, etc.) and the relative confidence attached to these tests. This
approach has been introduced by Harvey and Green in identifying voters in West-
minster (Harvey and Green, 1994; Harvey, Green and Corfield, 1996), and has been
criticised, both on methodological grounds (Adman, 1997) and on basis of manually
created true links in nineteenth century census lists (Tilley and French, 1997).
The scope of this section precludes a detailed overview of the rather detailed techni-
cal issues involved. More interesting is the overall gain in methodological knowledge.
Three aspects are worth mentioning:
• Name standardisation has grown considerably more sophisticated since the early
days of the Soundex and Gloria Guth algorithms, as, for example, Bloothooft
demonstrated in applying techniques from computer linguistics (Bloothooft, 1994,
1995, 1998).
• Identifying people in a historical context requires an appropriate data structure,
which separates original source data from the linkage information. A layered archi-
tecture is desirable, where (i) original source data are transformed into (ii) a format
suitable for analysis, and to which (iii) the results of the linking process are added
as additional knowledge (Bloothooft, 1995; Boonstra and Panhuysen, 1999; Keats-
Rohan, 1999; King, 1992).
• This is well illustrated by the coel system, a digital archive of about 5,000 docu-
ments and records, most notably elements from the Doomsday Book, pertaining to
the acquisition of English land by the Norman conquerors of the century following
1066 The system comprises three levels. Level one contains all source files, for the
most part the text of primary sources, which are given in full. Level two is a data-
base of person names, retaining original Latin forms, and level three represents the
interpretative stage, where nominal record linkage has taken place. Here, the user
can look for individuals and families, together with commentaries attached. Whate-
ver position in the database the researcher is in, the user is only a double-click away
from returning to the original primary source (Keats-Rohan, 1999).52
52
coel (Continental Origin of English Landholders 1066-1166): http://ahds.ac.uk/creating/case-studies/
coel/

56 The past
 • Different strategies have been tested and compared. However, clear guidelines sta-
ting which strategy has to be used with specific kinds of historical data are still mis-
sing. That is hardly surprising, because of the variety of conditions that will have
influence upon a data set (see above), but would have been in line with the intent of
finding generalised linking strategies. Tilley and French, for example, rejected au-
tomatic linking by means of multi pass techniques for nineteenth-century census
records, but failed to explain in a generalising manner, which source characteristics
are most relevant in this respect (Tilley and French, 1997).

3.5.1.7 Stylometry, Content Analysis and Thematics

Knowledge discovery in narrative sources is mainly based on word counting and
pattern matching, coupled with advanced statistics. In stylometry53 and content
analysis54 much work has been hypothesis driven, e.g., about disputed works of
Shakespeare, or the authorship of the Imitatio Christi, now undisputedly attributed
to Thomas a Kempis. There are a few examples of stylometric research in history
(Holmes, Gordon and Wilson, 1999; Martindale and McKenzie, 1995), but most
important, in the literary field, is the monumental work by Burrows (Burrows, 2003;
Burrows, 1987).
A somewhat related, multifaceted subject is thematics, an interdisciplinary field
of study focusing on textual themes. The precise definition depends on the scholarly
domain. As Louwerse and Van Peer pointed out in a recent publication, an important
feature is that themes allow the ‘grouping of meanings into manageable chunks’.
Therefore, related information technologies based on thematic analysis can be seen
as a form of knowledge management:

‘Themes render the multitude of information meaningful by streamlining indi-

vidual pieces of information into a meaningful whole which can then be processed
more efficiently and linked to ongoing cultural concerns.’ (Louwerse and Peer,
2002).

Thematic analysis covers many elements in the communicative process, varying

from the role of a theme during the process of a text’s creation, to the complemen-
tary aspect of analysing and modelling how texts are read and understood (cognitive
psychology, reception theory, etc.) (Kintsch, 2003; Meister, 2003). The computer
program presented by Louwerse is based on the concept of coherence in text; various
types of coherence can used in the retrieval of thematic texts (Louwerse, 2003).

3.5.1.8 Conclusions
Both the textual nature of historical and literary sources and the current state of
technology are good reasons for discussing text and databases together. In spite of a
difference in research objects, historical and literary studies have much in common

53
For a short introduction refer to (Rudman, Holmes, Tweedie et al.) �����������������������������������
http://www.cs.queensu.ca/achallc97/
papers/s004.html
54
A concise overview of content analysis is available at the web site of Colorado State University: http://
writing.colostate.edu/guides/research/content/

57 The past
 with regard to computing methods, techniques and tools, e.g., the digitised versions
of critical editions, concordances, retrieval facilities and statistical data processing.
Both go beyond the text itself, in literary reception studies, or in mainstream his-
torical studies that use digitised data without aiming at a critical edition. Particularly
here, the tension between the abundance of source texts and limited resources of
time and money appears, which has given rise to the dichotomy of ‘source-oriented’
versus ‘model-oriented’ data processing. Although during the last decade this issue
has been amply debated in the community of computer using historians, it seems to
be more a practical matter, rather than a dichotomy with a firm methodological basis.
The extreme cases will be always clear: when digitising the complete text is an im-
perative, or when the computer is used to implement a data-driven model. The ‘grey’
area in the middle is the most interesting and forms a methodological challenge.
The main requirement for any higher level processing is data and text having an
appropriate semantic structure. For the time being, creating this structure will re-
quire human intervention to a certain degree. However, the challenge is precisely the
search for automation. Both large scale historical projects and current developments
in other disciplines, like computer science and computer linguistics, show converg-
ing lines into that direction. If scanning is feasible, a wide variety of potentially appli-
cable techniques for further (semi-)automatic structuring do exist. The main problem
is not the lack of knowledge, but rather a disparate spread of techniques and expertise
over different disciplines, from social sciences, computer linguistics to knowledge
engineering and statistics, together with a gap between the theoretical solutions and
practical implementations. This is most clearly experienced in the lack of appropriate
historical tools.

3.5.2 Statistical methods in historical research

Statistical tools have been applied in historical research for quite a long time. Its use
has changed over the years, however. In the 1970s and 1980s, during the heyday of
cliometrics, statistics were used predominantly for testing hypotheses, analogous
to the way statistics were used in social and economic sciences. Today, statistics
are valued much more as a descriptive or even an exploratory tool than as an induc-
tive method, i.e., as a tool that either summarises or helps to find patterns in large
historical datasets. The reason for this change first of all has been the concern of
many historians that most statistical methods assumed their data to be sampled and
distributed statistically in a way to which historical data fell short by definition. A sec-
ond reason was that historians did not wish to generalise the statistical results they
achieved to a larger population; they were satisfied with a mere statistical description
of their data, and therefore did not need to test for significant deviations from a null-
hypothesis. This does not mean to say that inductive statistics are out of the scope
of the historian – on the contrary, a few new inductive statistics have received atten-
tion from quantitative history or would be worthwhile for quantitative historians to
investigate.
In the following paragraphs a few new statistical tools are described which seem to
hold great promise for future historical research.

58 The past
 3.5.2.1 Descriptive and inductive statistics

logistic regression
Multivariate analysis is of key importance to historians who want to explain variation
in a dependent variable by a series of independent variables. Traditional multiple
regression techniques are based on a number of assumptions that cannot be met in
historical research very easily. Traditional cross tabulation techniques fall short when
the number of independent variables is larger than two and interaction effects start
to blur the results. Therefore, techniques that are based on fewer assumptions are
favoured. At this moment, logistic regression analysis, which has a dichotomous vari-
able as dependent variable, seems to overcome most of the restrictive assumptions
traditional (ols) regression techniques have. In the first place, it does not assume
a linear relationship between dependent and independent variables. Secondly, the
dependent variable does not need to be normally distributed. Furthermore, normally
distributed error terms are not assumed and it does not require the independent vari-
able to be measured at an interval level. Normally, the number of cases in both cells
of the dichotomous dependent variable needs to be fairly large. King and Zeng, how-
ever, have expanded the use of logistic regression to situations in which the number
of cases in one cell of the dependent variable is much less than in the other (King and
Zeng, 2001). In doing so, it is possible to start multivariate analysis into the occur-
rence of rare historical events.
Thus far, the use of logistic regression in historical research has been limited. Only
in historical demography and in political historical research analysis, some examples
of its use can be traced. In historical demography (Lynch and Greenhouse, 1994)
studied the impact of various social, environmental, and demographic factors on
infant mortality in nineteenth century Sweden with help of logistic regression; (Reid,
2001) carried out similar methodological work on neonatal mortality and stillbirths
in early-twentieth-century England. Derosas used logistic regression to test which
variables influenced mobility within the city limits of Venice, 1850-1869 (Derosas,
1999).
In political historical research, logistic regression has been applied to analyse vot-
ing behaviour. Schonhardt-Bailey studied voting behaviour in the German Reichstag
to test the coalition of landed aristocracy and heavy industry around a policy of tariff
protection (Schonhardt-Bailey, 1998); (Cowley and Garry, 1998) tested seven hy-
potheses of voting behaviour on the Conservative leadership contest of 1990.Finally,
(Henderson, 2000) tried to analyse the extent to which political, economic, and cul-
tural factors are associated with civil wars in sub-Saharan African states, 1950-1992
Results indicated that previous colonial experience was a significant predictor to the
likelihood of civil wars. It was also found that economic development reduced the
probability of civil war, while militarisation increased it.

multilevel regression
If one wants to carry out a statistical analysis of historical data, the number of data
may be abundant, but the number of variables one can use is rather limited. Espe-

59 The past
 cially in micro-level research, where attention is focused on individuals and the rela-
tionships they have with their next of kin, the number of variables is small. It would
be wonderful if additional data that have been preserved only on other, aggregated,
levels could be included into an analysis as well, without losing the statistical tools for
testing hypotheses, as is the case in standard regression analysis.
In social sciences, a new technique has become available that indeed is able to cope
with different levels of analysis without losing its testing capabilities. This technique,
called multilevel regression analysis, was introduced by (Bryk and Raudenbush,
1992). Other, less mathematical, introductions have been published since then (Hox,
2002; Kreft and Leeuw, 1998).
In recent years, multilevel regression has attracted some attention from historians,
in particular from a group of historical demographers and economic historians at
Lund University in Sweden. For instance, (Bengtsson and Dribbe, 2002) studied the
effects of short-term economic stress on fertility in four Swedish parishes, 1766-
1865.

event history analysis

In event history analysis, the behaviour of individuals is studied as a dynamic proc-
ess of events which did – or did not – take place in the life of the persons under study.
The dependent variable in event history analysis expresses the risk or hazard of expe-
riencing the event. Besides that, event history analysis does not focus on the specific
moment in time at which a certain event occurred in the past, but on the time period
before it occurred. It does so by looking at the way in which one or more independent
variables may have contributed to the relative risk of the event to happen. It is im-
portant to realise that, because event history analysis measures periods of time and
not points in time, data can be appended to the analysis which are only available for a
limited time span. Even if we do not know what happened to a person before or after
such a time span, the information that is contained in this particular time span still
can be put to use. It is this inclusion of so-called ‘censored’ data, which makes event
history analysis a very powerful tool for analysing dynamic historical data.
Good introductions to event history are (Allison, 1984) and (Yamaguchi, 1991). A
good manual which can serve both as a student textbook and as a reference guide for
the use of event history analysis is (Blossfeld and Rohwer, 2002). There is a software
program called tda (Transitional Data Analysis), which is written exclusively for
event history analysis.
Although originating in biomedical research (where it is called survival data analy-
sis), event history analysis has been welcomed by historians in a pretty early stage,
for instance by (Raffalovich and Knoke, 1983). Especially in historical demography,
the technique has become a standard procedure (for an overview see (Alter, 1998).
Already in 1984, the first applications in historical demography have been published
(Egger and Willigan, 1984). In recent years, historical demographic studies have
been done on child mortality by (Bengtsson, 1999), on mortality by (Campbell and
Lee, 1996) and (Bengtsson and Lindström, 2003), on child bearing by (South, 1999),
and on marriage by (Gutmann and Alter, 1993) and (Cartwright, 2000).

60 The past
 Social historical research based on event history techniques have been done on social
issues like migration by (Schor, 1996), (Kok, 1997), (Campbell and Lee, 2001), on
legislation by (McCammon, 1999) and on inheritance by (Diekmann and Engelhardt,
1999).
Finally, event history analysis has been applied to the field of economic history,
notably on employment issues (Alter and Gutmann, 1999; Drobnic, Blossfeld and
Rohwer, 1999).
Although event history analysis is widely used, there are some unresolved prob-
lems attached to it. Some have to do with the assumptions that accompany various
different event history models, while other problems have to do with the impact
censored data still have on the results, or with the interpretation of results when an
‘event’ has more than one meaning. In any case, event history analysis has proven to
give new insight in historical processes.

ecological inference
Because of the lack of data on the individual level, historians often use aggregated
data in order to gain more insight into individual’s behaviour. In doing so, there is
always the possibility of falling into the trap of ‘ecological fallacy’, i.e., the problem
that arises from attempts to predict individual behaviour based on aggregate data of
group behaviour.
In the past a few attempts have been made to solve this so-called ‘ecological infer-
ence’ problem. Goodman proposed a solution in 1959, called ecological regression
(Goodman, 1959). Although this regression technique was able to overcome some
of the ecological inference problems, it created another: the results of an ecological
regression were hard to interpret. For instance, standardised regression coefficients
well over the maximum limit of 1 often appeared as a result.
In 1997, Gary King presented a different approach to the problem. ‘A Solution to
The Ecological Inference Problem: Reconstructing Individual Behaviour From Ag-
gregate Data’ was a book that introduced a new statistical tool; EI and EzI were the
accompanying freeware computer software programs (King, 1997).
Historians responded quickly to this new approach. In 2001, Historical Methods
published two volumes in which King’s solution was introduced and evaluated (start-
ing with an introduction by (Kousser, 2001)). This was done by replicating historical
research with known individual data on an aggregate level with King’s (and Good-
man’s) method to see whether the results would coincide. On the whole, the conclu-
sion was that King’s method might be a solution, but definitively not the solution.
Nevertheless, King’s method has been embraced by other sciences, especially social
and political sciences, and it is to be expected that it will be used in historical research
as well.

time series analysis

An important category of techniques for historical research is time series analysis.
Time series models can be applied to all sources for which diachronic series exist, not
only economic sources such as trade statistics or shipping movements, but also to

61 The past
 demographic sources such as population data. Therefore, time series analysis can be
applied to a great variety of historical research topics. However, applications outside
the domain of economic history are remarkably scarce (Doorn and Lindblad, 1990).
In population studies, three types of time-series effects are distinguished, having
to do with age, period, and cohort. Age effects are effects related to ageing or the
life cycle. For instance, individuals often tend to become more conservative as they
age. Period effects are effects affecting all cohorts in a given historical period. For
instance, individuals who experienced the Great Depression became more likely to
support social welfare policies. Cohort effects are effects which reflect the unique
reaction of a cohort to an historical event, or which were experienced uniquely by the
cohort. For instance, the post-wwii cohort, reaching draft age during the Vietnam
War, experienced unique issues that seem to be associated with increased alienation
from government. Disentangling these three types of effects for one set of time-series
data is a major challenge of time series analysis.
Some examples can be found in historical demography (for instance on mortality
by (Bengtsson and Broström, 1997) and on illegitimate fertility decline in England,
1850-1911 by (Schellekens, 1995). Other examples come from the field of political
history, where (Pacek and Radcliff, 2003) tested whether Left-wing parties were the
primary beneficiaries of higher rates of voter turnout at the poll-box. Time series have
been applied in historical climate research, where Schableger analysed time series
for the daily air temperature in Vienna between 1874 and 1993 Climate research was
also part of a spectacular interdisciplinary time series analysis by (Scott, Duncan
and Duncan, 1998) on four centuries of various grain prices in England (1450-1812).
Their analysis revealed cyclic effects of changes in weather conditions.
Finally, there are some examples of applying time series in social history. Although
a nice introduction of time series to social history has been published by (Stier,
1989), only very few social history articles have been published thus far. The papers
by (Isaac, Christiansen, Miller et al., 1998) on the relationship between civil rights
movement street tactics and labour movement militancy in the post-war United
States, and (Velden and Doorn, 2001) on strikes in the Netherlands are an exception
to the rule.In economic history, things are different. In this domain, time series anal-
ysis has remained alive and well over the past decades, and applications have been
diverse. A few special applications stand out. First of all, time series analysis was used
for international comparisons. Raffalovich investigated the impact of industrialisa-
tion, economic growth, and the unemployment rate on property-income shares in
a sample of 21 nations in Europe, Asia, North America, and the Pacific area during
1960-90 (Raffalovich, 1999). Li and Reuveny analysed the effect of globalisation on
national democratic governance 1970-1996 for 127 countries in a pooled time-series,
cross-sectional statistical model (Li and Reuveny, 2003). Finally, (Greasley and Oxley,
1998) reviewed three approaches to explain the timing, extent and nature of shifts in
British and American economic leadership since 1860 They concluded that better
educational opportunities played an important role for the usa to gain economic
leadership in the twentieth century.

62 The past
 The relationship between public expenditure on education and economic growth
has also been studied by (Diebolt and Litago, 1997) for Germany and by (Ljungberg,
2002) for Sweden.
Finally, there has been methodological interest in time series analysis as well.
There was a plea for time series as a descriptive tool instead of a tool to test hypothe-
ses by (Metz, 1988a). New tools for time series analysis were introduced: for instance,
methods for the analysis of long-term cycles like the Kondratieff cycle (Diebolt and
Guiraud, 2000; Metz, 1988b, 1993). Another example is the application of new
methods of filtering in order either to discern various cycles from one another (Muel-
ler-Benedict, 2000) or to analyse non-stationary stochastic processes (Darné and
Diebolt, 2000).

3.5.2.2 Exploratory data analysis and data mining

Data mining is a general term for a variety of techniques that are meant to gain
insight into a data set. Most often, its goal is to uncover the underlying structure of
the data set, but it may also have the purpose to find out what variables are the most
important ones, or what cases are connected to each other or what cases are the out-
liers in the dataset.
If the techniques employed are based on statistical formulas, data mining is called
‘exploratory data analysis’. It is these techniques that are presented here. The way
to present results from exploratory data analysis can take many shapes, but often a
graphical representation is favoured. Most of such visual methods are described in
the next paragraph on visual data analysis. In this paragraph only two methods of
exploratory data analysis are presented, cluster analysis and simulation.

clustering techniques
Cluster analysis seeks to identify homogenous subgroups of cases in a population.
That is, cluster analysis seeks to identify a set of groups in which within-group
variation is minimised and between-group variation is maximised. There are many
different clustering techniques, many of which are hierarchical. In hierarchical analy-
sis, the first step is the establishment of a similarity or distance matrix. This matrix
is a table in which both the rows and columns are the units of analysis and the cell
entries are a measure of similarity or distance for any pair of cases. The second step
is the selection of a procedure for determining how many clusters are to be created,
and how the calculations are done. The third step is to select a way to present visually
the results of a cluster analysis. Normally, a (cluster-) dendrogram is the method that
is used most often, but historians also use geographic information systems to plot
the results. Finally, statistical analysis on the results of cluster analysis can help to
interpret the outcome.

63 The past
 60 70 80 90 100

wbg8287
wbg8289
wbg8324
wbg8375
wbg9007
wbg9229
wbg1576
be8
wbg8397
bb225
bb270

Figure 3.7 A dendrogram, showing, among other things, the similarity between wbg8287
and wbg8289, and the dissimilarity between the first six cases and the remaining five. From
(Price, O’ Brien, Shelton et al., 1999). NHS
SBJH

NHS
HBCBC
CPTT

TUF

HQS

PES
PES TUF
TSE
GSJ[
TSB
WJQS
TSC
MBUS
TSF
TSH PB
DBS
WO

PMG

TSP

QF

IU IB

64 The past
 Figure
Cluster3.8
1 A cluster-dendrogram, showing four groupsCluster
of cases.
2 From (Graul and Sadée).

Figure 3.9. Geographical representation of two clusters of municipalities, The Netherlands

1899. From (Boonstra, 2001).

Around 1990, a few introductions to cluster analysis in historical research were

published (Bacher, 1989; Boonstra, Doorn and Hendrickx, 1990). Since then, cluster
analysis has been used by a variety of historians for exploratory purposes. Cluster
analyses, in which geographical information systems have been used to visualise re-
sults, have recently been done on the basis of nations (Obinger and Wagschal, 2001),
provinces (Delger, 2003), districts (Debuisson, 2001), municipalities (Boonstra,
2001), and parishes (Song, 2002). Cluster analysis on households has been done by
(Galt, 1986) and (Spree, 1997).

simulation
Although the idea of ‘simulation’ is simple enough, the statistical basis of its scientif-
ic use is very diverse and often very complex. Computer simulations are designed to
evaluate behaviour within a specifically defined system, establish models and criteria,
develop and analyse strategy, predict outcome, determine probability, and identify
relational patterns.
A model is always the starting point of simulation, and a correct operationalisation
of all variables and links between variables in the model is a prerequisite for good
analysis. When the model has been established, a series of simulation runs can be
done. During these runs, various techniques can be applied. One way of simulating
is by changing variable parameters, so that the impact of these changes on the model
can be measured. When the simulation process consists of a finite number of states,
in which the future behaviour of the system depends only on the current state and
not on any of the previous states, this is called a ‘Markov chain’ model.

65 The past
 Another way of simulating is by repeating the same model with the same parameters
over and over again in order to find out how stable the model is through all of these
runs. Repeated simulation can also be used to create a hypothetical dataset, which
can then be tested with various statistical tools. Simulations of this kind are called
‘Monte Carlo’ experiments.
Simulation techniques were widely established in the social and economic sciences
in the 1970s. In historical research with its wonderful possibility to check histori-
cal simulation results with the outcome of real historical events and developments,
applications have been rather scarce, however. The use of simulation techniques,
especially in political history has been disappointingly low, although its importance
has been stressed (Mielants and Mielants, 1997). A very early exception to the rule
has been the semi-computerized attempt to simulate the outbreak of Word War i
((Hermann and Hermann, 1967). A more recent example is the research done by
(Artzrouni and Komlos, 1996), who devised a very simple simulation model in order
to find out why Western European states achieved stable boundaries much earlier
than Eastern European states. The model does not take many factors, such as eco-
nomic differences, state policy, or military effectiveness, into account. Nevertheless,
it demonstrates that geographical constraints have played an important role in deter-
mining the map of modern Europe. A final political example is from (Deng, 1997),
who made a simulation model of the 1989 Chinese student movement. From the
model it becomes clear that the Chinese government, by concealing its preferences,
caused the death of many demonstrators who believed that the army would never
harm the Chinese people. The model also makes clear that an information gap can
lead to unintended and undesirable outcomes, even when actors behave rationally.
Most applications can be found in economic history and historical demography. In
historical demography, an important impetus to the use of simulation was given in
Reher and Schofield’s book on new methods in historical demography (Reher and
Schofield, 1993). In this book, a special part was reserved for research in which his-
torical demographic processes were simulated. More recent studies, in which simula-
tion techniques are used for historical demographic research, are for instance (Brod,
1998), who used Monte Carlo simulation for the analysis of marriage seasonality.
Zhao investigated the relationship between demographic conditions and family or
kin support systems in Victorian England with help of a simulation model in which
kinship patterns change during the life course (Zhao, 1996). Hayami and Kuroso
used a similar approach for their research into the relationships between demograph-
ic and family patters in pre-industrial Japan (Hayami and Kurosu, 2001). Okun used
simulation techniques in order to distinguish between stopping and spacing behav-
iour in historical populations (Okun, 1995). In doing so, she was able to determine
what the principle method of regulating family size was during the demographic
fertility transition of the nineteenth century. McIntosh used a simulation model in
trying to solve the question why populations of small towns in southern Germany
stagnated following the Thirty Years War (McIntosh, 2001).
Simulation models in economic history have been used to study macro-economic
effects in nations like Germany (Ritschl, 1998) and Russia (Allen, 1998), urban

66 The past
 systems (Guérain-Pace and Lesage, 2001), or in periods of time like the Industrial
Revolution (Komlos and Artzrouni, 1994). At the micro level, the work of the Com-
puting & History Group at Moscow State University must be mentioned. Andreev,
Borodkin and Levandovskii, all members of the group, used simulation models for
an explanation of worker’s strikes in Russia at the beginning of the twentieth century
(Andreev, Borodkin and Levandovskii, 1997). One of these models, using chaos theo-
ry as a starting point, even pointed towards a ‘spike’ in strike dynamics for 1905, even
though there was no input about events relating to the Revolution of 1905 Borodkin
and Svishchev used a simulation model based on Markov chains to find out more
about the social mobility of private owners under the New Economic Policy (nep) in
the Soviet Union during the 1920s (Borodkin and Svishchev, 1992).
Finally, simulation techniques have also been applied in historical environmental
research. Nibbering and DeGraaff developed a watershed model for an area on the
island of Java, using historical data on land use in order to simulate past hydrologi-
cal conditions and erosion (Nibbering and DeGraaff, 1998). Allen and Keay analysed
various simulation models to find out what caused the bowhead whale to be hunted
almost to the point of extinction by 1828 (Allen and Keay, 2001).

3.5.2.3 Texts and statistics

As mentioned above, stylometry, authorship attribution and content analysis are
important techniques for knowledge discovery. This does not mean to say that there
is communis opinio about tools for authorship attribution. On the contrary, recently
quite a few new methods have been introduced. In all introductions, the Federalist
Papers serve as a benchmark for testing various tools. The 85 Federalist Papers were
written by someone called ‘Publius’ in 1787 and 1788 to persuade the citizens of New
York State to adopt the nascent Constitution of the United States. But ‘Publius’ was
not a single person; of the 85 texts, 52 were written by Alexander Hamilton, 14 by
James Madison, 4 by John Jay, and 3 by Hamilton and Madison jointly. The author-
ship of the remaining twelve texts is disputed.
The Federalist Papers were used for authorship attribution for the first time by
(Mosteller and Wallace, 1964). Thirty years later (Holmes and Forsyth, 1995) came
up with a few new solutions, which triggered off new research by many others.
Tweedy et al. applied a neural network analysis (Tweedie, Singh and Holmes, 1996),
(Baayen, van Halteren and Tweedie, 1996) proved that attribution results could be
improved by annotating the texts with syntactic tags, whereas Khmelev (Khmelev,
2000; Khmelev and Tweedie, 2001) showed that the success rate would be even bet-
ter using Markov chain analysis on letters.
Against impressive results there are also principal objections in post-modern dis-
cussions about authorship. A much-voiced criticism boils down to the argument that
stylometry has indeed speed and power, but that it doesn’ t know what it is counting
(Merriam, 2002).

67 The past
 Very Positive (58) Hear (100)
Positive (25) Reason (100)
Neutral (8) Talk (83)
Negative (8) Is evaluated Think (100)
Very Negative (3) by author Does action Walk (58)

Non-Metallic Humanoid (8) Fear (58)

Metallic Humanoid (67) Has characteristic Friend (58)
Metallic Non-Humanoid (25) Is of this type Robot attributed by Loyalty (50)
characters Pride (50)
Brain (100) Trust (58)
Circuits (83)
Consciousness(83)
Eyes (75) Has feature Has emotion Fear (58)
Face (83) Friendship (58)
Hear (100) Loyalty (50)
Intelligence (100) Pride (50)
Memory (100) Trust (58)
Senses (100)
Sight (75)
Voice (83)

Figure 3.10 Content analysis: A map representing relationships among concepts in robot
descriptions. From (Palmquist, Carley and Dale, 1997)

Content analysis dates also back to the 1960s, when the famous Harvard program
The General Inquirer was deployed in automatic analysis of political documents, folk
tales and private correspondence, revealing secrets that could not be caught by the
naked eye. A more recent example is a study about the changing depiction of robots
in writing over more than a century. Content analysis has been deeply rooted in the
social sciences, and has not received the attention it deserves in historical research.55
In some cases stylometric research comes close to content analysis, as, for example,
with the multivariate analysis of two texts of the American novelist Charles Brock-
den Brown. Analysing the novels Wieland and Carwin Stewart was able to show that
Brockden has succeeded in creating a narrator with a distinctive voice, thus providing
evidence which could not be obtained through normal reading and aesthetic interpre-
tation (Stewart, 2003).
Both, stylometry and content analysis provide additional structure by adding a
connection between text parts and a conceptual level (e.g. a list of topics, events or
authors). Once established, these links can be utilised in search and retrieval opera-
tions.

55
For example some projects at the Zentrum für historische Sozialforschung (Cologne): http://www.zhsf.
uni-koeln.de/, the analysis of the historical newspaper corpus at Rostock (http://www.tu-chemnitz.de/
phil/english/chairs/linguist/real/independent/llc/Conference1998/Papers/Schneider.htm) and (Breure,
1999).

68 The past
 3.5.2.4 Conclusions
Recently, quite a few new statistical techniques have been developed that hold a great
promise for historical research. Although varied in the statistical results aimed at, as
well as in the underlying mathematical formulas, they are promising, because they
possess at least one of the two characteristics that have been described above: they
have a much better fit with historical data, and they are much more in line with the
traditional methodology of historical science. Logistic regression, multilevel regres-
sion, event history analysis and ecological inference are examples of techniques with
such a better fit; the various techniques for exploratory data analysis are examples of a
better fit with the traditional research methodology of historical science.
What will be most interesting, will be the development of new statistical tech-
niques that possess both characteristics. Traditionally, there has been already such a
technique: cluster analysis. Its use, however, has been hampered by the fact that in
social science, from which most statistical tools are derived, researchers disapprove
of cluster analysis because of its inability to test hypotheses. Therefore, it is to be ex-
pected that new methods for exploratory data analysis will not be developed by social
scientists. It will be information scientists who will do the job, especially within the
framework of data mining research. Furthermore, it will be information science that
will come up with new methods to present visual results from data mining. It is this
combination, in which visual presentation tools are added to techniques for extract-
ing information from large datasets that will set the agenda for research into statisti-
cal methods for historical science in the near future.

3.5.3 Images: from creation to visualisation

Until very recently, serious historians thought images to be of limited value to his-
torical research. Of course, in popular historical writing, images always have played
a pivotal role; very often, text in such histories was of minor importance. It was the
images, and their captions, that drew the reader’s attention first. In serious, scientific,
historical writings, it was the other way around: images were only valued when they
could illustrate a historian’s text – they played their part when words fell short.
At the moment, this situation is changing rapidly. First of all, the cost of displaying
images in print, which was a serious impediment to using images more extensively,
has diminished. Especially in the new forms of presentation that have come about,
like Internet publishing, costs have stopped being a problem at all. But the reduc-
tion in costs is only a small part of the explanation for the growing use of images in
historical research. Another, much more important reason is the developing inter-
est in the ‘culture of images’ and in the image as a historical source (Burke, 2001;
Zeldenrust, 2003). There are even attempts to ‘write’ a history by showing pictures
only (Staley, 1998, 2003). For a long time, research on images has been the exclusive
domain of cultural history, but nowadays interest has awakened by social and eco-
nomic historians as well.56

See for instance the international symposium on History and New Media, which was held in Berlin, 9-11
56

April 2003. See http://www.clio-online.de//Portals/_Rainbow/documents/hist03_programm.pdf.

69 The past
 Additionally, the amount of historical images available for research has grown
tremendously over the past few years. More and more images have been digitised
and put into visual archives, opening up a source for research by allowing historians
much easier access to thousands of images than ever before. ‘Images’ have become
‘information’, and fit into the life cycle of historical information in the same way as
textual or numerical data. As a consequence, the possibilities for historical analysis
on the basis of images have grown as rapidly, and, as a consequence, the methodo-
logical implications for using images in historical research have become an issue.
In this section we will deal with these methodological issues at the various stages
in the lifecycle of digitised historical images, from the phase of creation, via enrich-
ment, retrieval and analysis to the phase of presentation. To do so, we pay attention
mainly to digitised historical photographs. But there are other kinds of digital images
than historical photographs, e.g., movies and graphics, etc. which all have different
characteristics for storage, analysis or presentation. Whenever appropriate, these
other types will receive special attention in this section as well. The section ends with
two special paragraphs: one devoted to visualisation of textual data, and one devoted
to historical geographic information systems, and the maps these systems can pro-
duce.

3.5.3.1 Creation
If one wants to use thousands of images in order to do research into a specific histori-
cal period of time or into a specific theme, one cannot proceed without digitising
the images and putting them into a database. At the moment, already hundreds of da-
tabases of this kind have been set up, many of them accessible through the Internet.
They vary greatly in size, from hundreds into millions of images.
There are a few introductions to digitising photographs. A book by (Frey and Reilly,
1999) contains detailed discussions of technical and quality issues with lots of il-
lustrations. Ostrow is one of the few who published a guide for digitising historical
photographs for the internet (Ostrow, 1998).
This does not mean to say that there is a standard way of creating digitised images.
Already this first step in the lifecycle of digitised images shows a wide range of solu-
tions. Some databases contain only low-resolution or compressed images, in order to
reduce costs or to overcome copyright problems, while other databases contain differ-
ent copies of one image, each with a different resolution. Some databases contain im-
ages that have been enhanced with special software in order to improve the pictorial
quality of the image; others have explicitly decided not to do so.
Finally, image databases also differ in the way the images are stored and preserved.
Preservation techniques are object of numerous studies, for instance within the
framework of the European sepia project.57 Klijn and De Lusenet have presented
an overview of various datasets that have recently been set up in the eu (Klijn and
Lusenet, 2002). The report shows that there is no standard in the technical process
of digitising, no standard in the quality of the digitised images, no standard in the

57
sepia project: Safeguarding European Photographic Images for Access. http://www.knaw.nl/ecpa/
sepia/

70 The past
 preservation techniques employed and no standard in the use of metadata to describe
an image.

3.5.3.2 Enrichment
A very important issue is the enrichment of digitised photographs in order to be able
to retrieve them systematically and comprehensively. There are two major approach-
es to image information retrieval: content-based and metadata-based image retrieval.
In the metadata-based approach, image retrieval is based on descriptions of images.
The problems surrounding what to describe and what not are numerous. Until
recently, databases of photographic collections were set up mainly by art historians.
The metadata they used mirrored their interests, describing for instance in detail the
kind of paper the original photograph was printed on, leaving a description of what
was on the picture aside.
But making a description of what can be seen on a picture is easier said than done.
Contrary to other kinds of data, a description about what is on a picture can take
many forms, and will be different for almost every single viewer. The picture in Fig-
ure 3.11, for instance, can serve as an example (Klijn and Sesink, 2003).

Figure 3.11 Information about a picture’s content can be different for various viewers: does
the picture show the knaw main building, a Volvo 234 turbo fuel injection, or autumn on a
canal in Amsterdam?

Nevertheless, at the moment, a lot of effort is being made towards a standardised

description of photographic images. From the European sepia project, a proposal
has been made to have a picture described with no less than 21 fields of information
(Klijn and Sesink, 2003).

71 The past
 Main reference code Geographical location
Name of institute Access restrictions / copyright
Acquisition code Relationships
Location Status
Description Technical identification Dimensions
Title Photographic type
Creator File format
Descriptors/ subject headings/classification References
Names Origins of collection / grouping
Date Contents of the collection / grouping / acqui-
sition

Figure 3.12 sepiades’ 21 core elements to describe a photograph

The metadata-based enrichment of video images poses a special problem, because

the image that needs to be described changes constantly. A video therefore needs to
be divided into segments in order to describe them separately. To do this efficiently,
automated methods for segmenting videos are called for. At the moment, some effort
is being put into such methods within information science: a few parsers have been
developed which divide a video stream into constituent building blocks (Koprinska
and Carrato, 2001). These efforts primarily have the intention to compress video data
files, but can be used for classification purposes as well.

3.5.3.3 Retrieval
As has been stated before, there are two kinds of retrieval systems: content-based and
metadata-based retrieval systems. In content-based image retrieval the images are
retrieved based on the characteristics of the digitised image, such as colour, texture,
shape, etc. Its advantages are obvious: there is no need for adding metadata to the
database, keeping the cost of a digitised collection of images very low. Content-based
image retrieval is a hot issue in information science at the moment. At the moment,
work is done on dozens of content-based image retrieval systems, an overview of
which is given by (Gevers and Smeulders, 2004). Some of the systems use histori-
cal photographs to test the quality of the system.58 Although fascinating the results
suggest that the prospects of content-based image retrieval systems for historical
research are not good.
Therefore, historians and archivists have pinned their hopes on metadata-based
retrieval systems. To retrieve images from a database, the queries normally are based
on keywords. Because of the lack of standardisation at the enrichment of images, the
precision and recall of such search methods is often disappointingly low. It is there-
fore better to use ontology-based annotations and information retrieval (Schreiber,
Dubbeldam, Wielemaker et al., 2001).
A solution to the problems related with content-based and metadata-based retrieval
systems could very well be a combination of both. A prototype of such a system,

58
For instance, in Leiden, where a content based image retrieval system is being tested on a database of
21,094 studio portraits, 1860-1914. http://nies.liacs.nl:1860

72 The past
 called ‘Ontogator’ has been put to use with a photo repository of the Helsinki Univer-
sity Museum (Hyvönen, Saarela and Viljanen, 2003).
A second issue of interest to historians and information scientists, is the creation of
ontologies for various image collections, or systems to query various data sets. An
interesting example, especially for historians, is the German Prometheus project
(Hartmann), which is based on the κλειω database software. Image databases are lo-
cated at various servers; a central server only works as a broker between these image
databases and the end-user, making the impression to the end-user that it makes his
queries and views the results from one single database. Prometheus also includes a
number of new devices to retrieve and analyse digitised images.

3.5.3.4 Analysis: visualisation

Visualisation, or visual data analysis, refers to computerised methods in which data
are synthesised and summarised in a visual way, comparable to the way statistics re-
capitulate data in a formal numerical way. In this sense, visualisation is not meant to
make a presentation of known (statistical) results as is done with traditional graphics
and maps, but is meant to explore data in order to generate ideas and explore alterna-
tives (Grinstein and Ward, 2002; Gröller, 2001). In so doing, visualisation offers a
method for seeing the unseen.
The interest in visual data analysis has grown immensely during the last couple of
years in many sciences, especially information science. In historical research, exam-
ples of its use are hard to find, except for the use of cluster analysis (see Section 3.5.2).
Like cluster analysis, visual data analysis falls within the framework of explorative
data analysis, as a means of gaining understanding and insight into the data in a
visual way. The difference between explorative visual data analysis and visual presen-
tation is small, however. Similar tools can be used in both ways: if a special technique
allows researchers to explore and interpret their data well, it will serve as a means to
present their data efficiently as well.
Visual exploration of data can take many forms, for instance (Gershon and Eick,
1997; McCormick, Defanti and Brown, 1987; Reinders, 2001). For historical re-
search, the visual analysis of time-varying data will be most interesting. A number of
applications have already been developed over the years, even in the pre-computer era
(Tufte, 1983).59 A few examples of new visualisation tools are given below.

• Lifelines
Among the first computerised visual data analysis applications in history has been
the display of the dynamics of household composition, originally created by the
Swedish Demographic Database at Umeå in de 1970s (Janssens, 1989). Time runs
from left to right, starting with the moment of household formation. Children
enter the graph when they are born and leave the graph when they die or migrate
from the household. The same procedure is kept for parents, grandparents, other
relatives and lodgers. In this visual way, a comparison can be made between the
dynamics of various households.

59
A nice overview of classic contributions to visualization can be found at http://www.csiss.org/classics/

73 The past
 Figure 3.13 Lifeline of a household, 1880-1920 (Janssens, 1989). Explanation of symbols
used: S start of observation, E end of observation, B birth, D death, O migration out,
I migration in, M marriage, N entry.

A very similar approached has been use by (Plaisant, Milash, Rose et al., 1996) in
order to visualise personal histories.

• The Lexis pencil

There has been little development of graphical methods to visualise an event his-
tory dataset. This is partly because of the complexity of such datasets; it is easy to be
overwhelmed with the number of variables and different dimensions of a typical
study.
In order to display event history data, a Lexis ‘pencil’ (introduced by (Keiding, 1990))
can be made for each case history. Three variables can be selected for three co-ordi-
nate axes. Normally the Y-axis at least will be assigned a variable that is some measure
of time. Any of the remaining variables may be used to ‘paint’ the faces of the Lexis
pencils with colours according to the value of the variable. There will normally be one
variable that is an index of the case histories. The resulting ‘geometry’ is displayed in
a window together with a set of axes and annotation.

74 The past
 Figure 3.14 The life of a married couple displayed as a Lexis pencil. Time runs from left to
right, starting at date of marriage and finishing at the survey date. Each face of the pencil
represents a different variable. The top face represents the employment history of the wife,
the middle face that of the husband, and the bottom face the age of the youngest child in
the household. From (Francis and Pritchard, 1998).

• Calendar view
Van Wijk and Van Selow use time series data to summarise them in a calendar-
like fashion in order to find changes and irregularities in standard patterns (Wijk
and Selow, 1999). A similar approach, called ‘the agenda’, has been proposed by
(Daassi, Dumas, Fauvet et al., 2000).

75 The past
 Figure 3.15 ‘The Agenda’. Visualisation of the monthly productions of an assembly line us-
ing the calendar technique (Daassi, Dumas, Fauvet et al., 2000).

• Concentric circles
Concentric rings represent different variables. The colour, shading and width of
the rings as they are traversed in a clockwise direction represent changes of the
variables over time (see (Barry, Walby and Francis, 1990); see also (Daassi, Dumas,
Fauvet et al., 2000))

Graphical representation of a work history

1990 1920

Part-time working ND
000 Work history member 183
Not employed Production
Service
Professional
Full-time working Clerical
Manual
Missing
Occupation ring

Time axis
Industry ring
Event history boundary
Age marker
(set at age 25)

Figure 3.16 Example of the concentric circles technique, illustrating a lifetime work history.
Colour is used to represent movement in and out of different industrial sectors and social
class; the width of the rings represents the number of hours worked. From (Barry, Walby
and Francis, 1990).

76 The past
 A number of more or less similar visual techniques have been developed in order to
find irregularities in time series (Keogh, Lonardi and Chiu, 2002).

3.5.3.5 Visual analysis of texts

A novel application of visualisation techniques is in the exploration of texts, especially
in large textual datasets, where visual interfaces to digital libraries apply powerful
data analysis and information visualisation techniques to generate visualisations of
large document sets (Börner and Chen, 2002). ‘Semantic timelines’ may serve as an
interesting example for historians, because it includes time as a classifying principle
for historical texts. Since much historical information involves commentary on previ-
ous data, different time periods are vertically stacked, so that a period of comment or
consequences can sit directly on top of a period of earlier activity. In addition, seman-
tic timelines are zoomable, can illustrate time granularity and indeterminacy, and
can include hierarchical nesting structures for drill-down. Together, these techniques
allow a vast amount of historical information to be viewed within a rich semantic
context, at various scales simultaneously (Jensen, 2003).

Figure 3.17 Four time periods are vertically stacked, showing references among publica-
tions dealing with republican theory. Arrows indicate support, opposition, comment and
familiarity with the issue at stake (Jensen, 2003).

A second example is ‘AuthorLink’, a project by Jan Buzydlowski. AuthorLink explores

author relationships through co-citation patterns. The assumption is that if two
authors are often cited together by many other authors, these two authors likely have

77 The past
 a common intellectual interest in their research and writing. When many related
authors’ pair-wise co-citation patterns are explored, we will have a map of a subject
domain where authors on the map represent ideas or subtopics as well as their rela-
tionships (Buzydlowski, White and Lin, 2002).

Figure 3.18 Example of the use of AuthorLink. The text base used by AuthorLink is the Arts
& Humanities Citation Index, 1988-1997, comprising a total of about 1.26 million records.
A search was done on Plato as the main author. From (Buzydlowski, White and Lin, 2002).

In the Netherlands, an approach similar to AuthorLink has been put into practice
within the historical domain by the Digitaal Erfgoed Nederland consortium. They use a
commercial product called ‘Aquabrowser’ to find co-occurrences of words in various
web sites.60
But textual databases of lesser size can benefit from visual data analysis as well.
Monroy et al., for instance (Monroy, Kochumman, Furuta et al., 2002a; Monroy,
Kochumman, Furuta et al., 2002b), developed a way to study differences in various
early text editions of Cervantes’ Don Quixote with visualisation tools.

60
More about the aquabrowser: http://www.medialab.nl/

78 The past
 Figure 3.19 A timeline viewer, depicting variants among six early editions of Cervantes’ Don
Quixote. (Monroy, Kochumman, Furuta et al., 2002b).

Finally, Lecolinet et al., who introduced tools for automatic line segmentation of
scanned hand-written manuscripts (see also Section 3.5.1.5.), also showed innovative
methods for visualising literary corpus which contains many versions or variants of
the same manuscript pages (Lecolinet, Robert and Role, 2002).

Figure 3.20 Perspective Wall model adapted for visualising manuscript collections. From
(Lecolinet, Robert and Role, 2002).

79 The past
 It is to be expected that visualisation of historical data will become a tool to take off in
the near future in historical science. However, it will take a while before application
will be wide spread. Before that, standards need to be developed: standards on how
to visualise certain kinds of data, and, especially, standards on how to interpret visual
representations (Chen and Börner., 2002).

3.5.3.5 Historical GIS

‘It is easy to predict that when we recollect the development of history at the end
of the twentieth and the beginning of the twenty-first century, the introduction of
gis to research and teaching about the past will be one of the signs of the success-
ful continuation, and reinvigoration, of that tradition [of innovation in historical
research]’ (Guttman, 2002)

If we define geography as the study of spatial differentiation, and history as the study
of temporal differentiation, historical gis can be defined as the study spatial patterns
of change over time (Knowles, 2002).
‘gis’ refers to ‘Geographic Information System’, an integrated system in which geo-
graphic co-ordinates and research data are stored, as well as tools retrieve and analyse
information.
Until recently, historians have made only limited use of gis tools. There are a
number of reasons why this has been the case. First of all, historians for a long time
have thought of maps only as a means for presenting data instead of analysing them.
For such a limited scope, gis software was too expensive to use for a long time. Next
to that, the data structure of gis software was very exotic, while at the same time the
software was not very capable of importing data into its system. And in the third
place, standard gis software seemed poorly suited to handle geographic changes over
time.
Nevertheless, an overview in 1994 revealed that there were a number projects
being carried out in various countries spread across Europe and the United States
(Goerke, 1994). But that number is negligible if look at the use of historical gis
nowadays. And rightly so: at the moment, gis software is widely available and capable
of importing data in various formats. Moreover, the visual quality of recent histori-
cal gis applications is stunning.61 There is even good software freely available, and,
maybe most important, the problem of geographic changes over time has been noted
by historians, geographers and information scientists alike.
In a recent introduction to the use of gis in historical research, Ian Gregory (2002)
cites Peuquet who stated that a fully temporal gis must be able to answer three types
of queries:
Changes to a spatial object over time, such as ‘has the object moved in the last two
years?’, ‘where was the object two years ago?’ or ‘how has the object changed over the
past five years?’

61
A good sample of recent historical gis projects is in (Knowles, 2002).

80 The past
 Changes in the object’s spatial distribution over time, such as ‘what areas of agricul-
tural land-use in 1/1/1980 had changed to urban by 31/12/1989?’, ‘did any land-use
changes occur in this drainage basin between 1/1/1980 and 31/12/1989?’, and ‘what
was the distribution of commercial land-use 15 years ago?’
Changes in the temporal relationships among multiple geographical phenom-
ena, such as ‘which areas experienced a landslide within one week of a major storm
event?’, ‘which areas lying within half a mile of the new bypass have changed from
agricultural land use since the bypass was completed?’
Gregory comes to the conclusion that at the moment there is no gis system that
can cope with these three queries. But this does not mean to say that no progress has
been made at all in this respect. For instance, regarding the problem of boundary
changes between spatial objects over time, a number of solutions did have been de-
vised. Leaving aside the very easy solutions – a new map is drawn every time a bound-
ary change takes place – which is very time-consuming, takes a lot of disk space and
does not facilitate easy comparisons of spatial changes over time, there are two ways
to solve this problem: by using a date-stamping approach, or by using a space-time
composite approach. The data-stamping approach, which for instance has been used
by (Gregory and Southall, 2000) and (Boonstra, 1994b), defines time as an attribute
to a spatial points and spatial objects. An administrative unit x is a spatial point, with
a specific starting date and a specific end date as its time attributes, as well as a set of
lines, also with specific starting and end dates. When drawing a map of x at a specific
moment in time t, only those lines are selected that have start and end dates on either
side of t. See Figure 3.21 Boonstra used a similar approach, using polygon attributes
instead of line attributes.

Label points attributes

Name Start date End date
a Anarea 0/0/0000 0/0/5000
1 b Elsewhere 0/0/0000 0/0/5000
2
a Line attributes
b
3 Start date End date
1 0/0/0000 0/0/5000
2 0/0/0000 1/9/1894
4 3 1/9/1894 0/0/5000
4 0/0/0000 0/0/5000

Figure 3.21 Example of the data-stamping approach, showing how a boundary change
between Anarea and Elsewhere on 1 September 1894 can be handled. Source: (Gregory and
Southall, 2000).

81 The past
 The space-time composite approach defines administrative units as a set of smaller
polygons that do not change over time. Each polygon has at least one attribute: the
administrative unit to which it belongs. If a polygon changes from one administra-
tive unit to another, the attribute data changes as well. These smaller polygons are
referred to as the Least Common Geometry (lcg). This can consist of ‘real’ low-level
administrative units that are known to be stable over time, as in the Swedish system
that uses parishes to create districts, municipalities, and counties (Kristiansson,
2000), but it can also consist of ‘virtual’ polygons that were created as a result of
boundary changes. Such a solution has been proposed and tested by (Ott and Swiac-
zny, 2001) and put to use in hisgis, the web-based Belgian Interactive Geographic
Information System for Historical Statistics.62 In both cases, a dissolve operation is
needed to re-aggregate the polygons in the lcg to form the units in existence at the
required time. See Figure 3.22.

Label points attributes

Time 1 Time 2
a Anarea Anarea
b Anarea Elsewhere
c Elsewhere Elsewhere

a c
b

Figure 3.22 Example of the data-space-time composite approach, showing how a boundary
change between Anarea and Elsewhere is handled: during Time 1, Anarea is an aggregate of
polygons a and b; during Time 2, Anarea consists of only polygon a. Source: (Gregory and
Southall, 2000).

62
More information on the Belgian Historical gis project at http://www.flwi.ugent.be/hisgis/start_en.htm

82 The past
 3.5.3.6 Conclusions
Internet and low-cost computer storage facilities have triggered much interest in the
use of images as a source for historical research. The availability of hundreds of large
collections of digitised images all over the world does not mean that no work needs
to be done anymore before serious research can make use of such collections. On the
contrary, the possibilities to retrieve images with a sufficient degree of precision and
recall are still small.
The inclusion of specific metadata, which deal with the historical meaning and
context of the image, in order to cater for the variety in which an image can be
interpreted, poses a serious problem for information science to deal with. Trying to
develop ways for content-based retrieval of historical images instead of metadata-
based retrieval seems to be a spectacular way to circumvent this problem, but it is to
be expected that in the near future metadata-based retrieval will generate results that
are more interesting for historical research.
Most of the various types of visual data analysis fall within the framework of ex-
plorative data analysis, as a means of gaining understanding and insight into the data
in a visual way. Although quite a few tools have been developed, there is still much
to do in finding ways to present time-varying data visually. Next to that, there is also
need for research in which visualisation techniques are developed in combination
with tools for exploratory analysis of historical data.

3.6 Final remarks

The structure of this chapter has been based on trends in humanities computing,
which has some drawbacks. It leaves us with blind spots, simply because certain
techniques or approaches are not explicitly discussed in publications. Multimedia
is slowly appearing in humanities research, ‘information retrieval’ is not labelled as
such and mostly addressed implicitly, knowledge in a technical sense (as in ‘knowl-
edge engineering’ and ‘knowledge technology’) does not belong to the humanities
vocabulary.
The comments made in Section 3.4 on the ambiguity within the ranks of the Asso-
ciation of History and Computing on how to use information technology in historical
research, had the intention to make clear that historical information science cannot
do without a structural scientific footing, a footing which the ahc has not been able
to give. Both content and structure demand conditions that need to be fulfilled to
ensure a future for historical information science.
The overview in Section 3.5 was intended to focus on the complexity of computa-
tional problems and to indicate handles for applying achievements from, amongst
others, computer science, while taking advantage of work that has been done already
in the humanities research itself. Text mining and knowledge technology, for exam-
ple, may open quite new perspectives, but such expertise should be applied on the
correct level of abstraction, fitting in with (or at least being aware of) conventional
techniques, as record linkage, statistical analysis and source criticism; otherwise new
solutions will not be meaningful to the practitioners in the field.

83 The past
84
4 The present

4.1 Conclusions from the past

‘And yet, and yet, while there is much to celebrate about the last decade, the fact
remains that the profession is still divided between the small minority of histori-
ans who uses computers as tools for analysing historical data and the vast majority
who, while they might use a pc for wordprocessing, remain unconvinced of the
case that it can become a methodological asset.’ (Speck, 1994)

These words were written by Speck in 1994 At that time, ten years after the Hull
symposium, the position of computing historians had been consolidated; they had
become organised and had established their own communication channels. However
they had not reached the majority of the profession, which remained resistant to this
new branch of methodology.
Moreover, the community was divided in itself. Although scientific discussions
would be fruitful and foster intellectual progress in a field, the Association for His-
tory and Computing (ahc) had not succeeded in coaxing clear scientific conclusions
out of the scholars it addressed. The debate had stopped at the level of being aware of
different points of view.
Historical computing had different meanings for different categories of compu-
ter using historians. Within the ranks of the ahc a majority supported the idea that
information technology was something that simply had to be applied to historical
research. They were impressed by the blessings of the ‘mighty micro’ and convinced
of the capabilities of the new technology. For them, historical computing referred
mainly to strategies for obtaining historical results with the aid of computers. The
precise nature of the procedures to be followed and the limitations of the technol-
ogy-as-provided were far less important than the results themselves. The methods
and techniques deserved only attention as far as they had to be acquired. In this way,
this category stayed relatively close to the majority of historians who rejected histori-
cal computing as such. From their point of view, historical computing was above all
history.
A relatively small number of computer using historians pleaded for historical
computing in the sense of a historical information science, and did so with good ar-
guments. Their weakest point, perhaps, was their noble motivation of helping fellow
historians to deal with computational problems at an acceptable level of skills with-
out losing focus of their proper historical enquiry. They kept trying to convince their
unresponsive colleagues, and unfolded missionary activities in historical projects
that could benefit from their expertise. With hindsight, one cannot avoid the conclu-
sion that convincing did not work at all. It must be admitted that in many cases this
noble attitude may have been nourished by the understanding that any funding for

85 The present
 research could only be obtained through co-operation with regular historical projects
– and vice-versa. This problem was difficult to solve, as we hinted at already in Chap-
ter 3, in the context of the Dutch situation. In addition, historical computing requires
a continuous testing of software and techniques under realistic conditions, based on
the practice of historical research. Nevertheless, in spite of several promising at-
tempts, this part of the ahc community failed to maintain a clear common focus and
to establish broad co-operation.
As a consequence, research into computerised methods and tools for historians to
use remained limited. To a large degree, research was either done by outsiders, who
did not take part in ahc conferences and did not report to the ‘history and computing’
community, or by individuals who did not find a sounding board for the results they
had achieved. In sum, the Association for History and Computing did not live up to
the promise of being a platform for researchers in historical information science, nor
in disseminating the tools they created to a wider audience of professional historians.
The result of it all is threefold. Firstly, researchers from within the domain of his-
tory discussed not all relevant topics and computing - much work was done outside
the field. Secondly, a proper research infrastructure for historical information science
was not set up. And thirdly, a link between historical information science and general
information science failed to become established. These three issues will be dis-
cussed in more detail in the next three sections.

4.2 The lost topics

If we look at the themes that have been researched over the last couple of years, it is
surprising to see the expanse of issues that has been investigated. If we put all issues
mentioned in Chapter 3 along the life cycle of historical information, it is positive to
see that all aspects have been touched upon, from creation (creating visual databases
for instance), through enrichment and retrieval to analysis and presentation. It is
nevertheless evident that within the framework of ‘history and computing’, i.e., the
ahc and its journal, some issues have received more attention than others. It is also
clear that some issues almost have exclusively been discussed and researched outside
the traditional domain of history and computing. Defining such issues as ‘lost top-
ics’, a list can be made of issues that did not get much attention from the domain of
history and computing.

86 The present
 Discussed within Discussed outside
Enrichment Editing
'history and computing' 'history and computing'

Creation
Retrieval source oriented data xml modelling; creating
Creation
modelling; optical char- visual databases; time-
acter recognition of old varying historical gis;
prints and manuscripts; textual databases;
Presentation Analysis

Enrichment
Enrichment Editing
metadata for historical xml standards for
sources; adding historical
metadata; linking source
Creation Retrieval fragments;

Presentation Analysis

Enrichment Editing
Editing
record linkage; family source-critical editions;
reconstruction;
Creation Retrieval

Presentation Analysis

Enrichment Editing

Retrieval
content based image ontologies for histori-
Creation Retrieval retrieval; cal research; history &
the semantic web;
metadata-based image
Presentation Analysis retrieval;

Enrichment Editing Analysis

multiple regression exploratory data analy-
analysis; event history sis; visual data analysis;
analysis; simulation; visual text analysis;
Creation Retrieval statistics on fuzzy data; historical gis; author-
ship attribution; content
analysis;
Presentation Analysis

Enrichment Editing
Presentation
historical gis; visual data analysis;
visual text analysis;
Retrieval timelines;
Creation

Presentation Analysis

87 The present
 4.3 A failing infrastructure
As has been stated above, ‘history and computing’ did not succeed in creating an in-
ternational platform for historical information science. At national levels, attempts to
build an infrastructure failed as well. In the Netherlands, for instance, only a very few
history departments set up centres for methodological research and development.
Consequently, not much thought was given to the formulation of research problems
and their solutions, and even less so to the formulation of it related problems and
solutions. In the Netherlands, all national historical research centres like the Huij-
gens Institute and the N.W. Posthumus Institute focus exclusively on thematical issues.
There is no room for research into methodological issues.
Secondly, there was the short-lived success of History & Computing centres.
Twenty years ago, history departments encouraged such centres, in which historians
with some knowledge of information technology tried to help their colleagues to cope
with the information problems they had. The introduction of easy-to-use Windows-
based software resolved a few of the problems specific to history (like for instance the
use of relational database systems to help keep track of changes in records over time),
causing management officials to think that there was no need for further support
any more. University cutbacks have erased History & Computing rapidly and almost
completely.
Thirdly, historical research has remained an individually based kind of research.
Solutions for historical information problems are therefore always linked to one par-
ticular research project, showing a lack of awareness of generalising results.
And finally, when dissemination of it results in history did take place, it was not
among historians, but among the specialists that were working within the History &
Computing centres, leaving ‘normal’ historians unaware of relevant contributions of
it to historical research.

4.4 The failing relation between history and computing and information science
If we look at the list of ‘lost topics’, it becomes obvious that those who adhere to the
field of history and computing have not kept themselves abreast of the developments
in it research. For instance, there has hardly been any discussion about the way his-
torical data – with its typical characteristics – could be modelled; no standards have
been developed about the way metadata should be added to historical digital data;
hardly any research is done into possible new tools for analysis, and so on.
On the one hand, the reason for this is that people working in the field of history
and computing do not succeed in disseminating their it-based solutions to histori-
cal problems to the traditional field of historical science. On the other hand, another
reason is that, because ‘history and computing’ did not succeed in creating an inter-
national platform for historical information science, there is only a poor relationship
between historical information science and information science.It is not entirely due
to the problems encountered within the field of history and computing that there
is no relationship between historical information science and information science.
It cannot be denied that information scientists rarely give attention to information
problems that are typical of historical research. On the one hand, this is because they

88 The present
 are not aware that such problems exist, but on the other hand, the idea of solving
such problems is not within the realm of problems that are tackled in informa-
tion science research projects.Be that as it may, as a consequence, historians and it
specialists have not established a fruitful communication, and therefore cannot at
present exchange views, problems and solutions.

89 The present
90
5 The future

5.1 Conclusions from the present: A paradox

Up to the present, the field of history and computing has been rather productive in
generating publications on the application of computer techniques to historical re-
search. The World Wide Web has changed the way historical sources are published,
at least to some extent. Printed editions have met competition from an increasing
quantity of digitised sources made available on the Internet, although so far these are
only rarely integrated and provided with additional tools as digital archives and digital
libraries do.
However, considered as a field of scientific study, the position of history and com-
puting is far less favourable. With a few exceptions, methodological and technical
problems have not been studied in a generic way, aiming at solutions that could be
used in the entire field for a specific class of problems. If this will be realised, the cur-
rent ‘best practice’ will pass into a historical information science (Dutch: historische
informatiekunde), or more precisely, into a specialised branch of information science
concerned with the historical domain. As elsewhere in this report ‘historical’ has to
be understood in the broadest sense of the word: regarding ‘historical information’,
not restricted to history as a discipline, but also being the corner-stone of other fields
of study in the humanities. At present, it is unlikely that this will happen soon with-
out any change of policies.
The historical community is divided on the question regarding how much em-
phasis this research direction should get. As long as the phenomenon ‘history and
computing’ is discussed, one paradox has turned up again and again: for historians,
the computer will provide usable tools, but somehow these tools need to be devel-
oped. ‘The computer’ will never help one to write better history automatically, but
the analysis of historical problems can be carried out in a more sophisticated and
innovative way within the framework of information technology. This is the histori-
an’s task, not that of a computer scientist or an information analyst. It is the duty of
these ‘technicians’ to adapt the information technology framework in such a way that
historians can do their work: to find new answers to old questions, or even to be able
to pose questions that were not thinkable or answerable without the application of
information technology. Although this may sound obvious to some of us from our
distant point of view in this report, it may be intuitively hard to grasp in the practice
of historical studies. Therefore, some elaboration of this paradox may be helpful.
Historians want to write history, and for that purpose they may use a computer to
manage and to analyse their data. The computer makes data available, reduces the
amount of time spent to searching, and links dispersed chunks of information, allow-
ing questions that require such an extensive quantitative or qualitative analysis that
would have otherwise been unfeasible. This leads easily to the conclusion that the
activities in historical computing should be directed to practical matters. Along this

91 The future
 line of reasoning the value of the solution is demonstrated through the computer-
aided historical study itself. Reflection on these activities is considered as useful, but
mainly as a form of sharing experiences. Moreover, historians want to stay historians,
and do not want to delve into the intricacies of information technology – and they are
perfectly right.
The paradox is that enabling the historical community to be practical computer
users in such a way, somehow and somewhere a substantial amount of energy has
to go to more fundamental methodological and technical research with respect to
computing in history. In the early days precisely this has been Thaller’s claim, when
he developed a special database management system for historical research. It does
certainly not mean that the every historian has to be turned into a computer expert.
There is a parallel with the expertise in editing sources: for example, a historian, who
makes use of published medieval charters, can accept very well that some of them
are fabrications and deploy this knowledge without being an expert in the technical
assessment procedure. However, he should be sufficiently aware of the underlying
reasoning and be able to pose adequate questions.
Those historians, who are reluctant to pass the threshold of computer-related
methodology and techniques, may be inclined to adapt their questions and research
strategies to means readily available. This will result in the use of common commer-
cial software with all its limitations and inconveniences, which, however, will not be
necessarily detrimental to the quality of the historical discourse itself. However, even
if at first sight standard software seems to suffice, the gap may widen between the so-
lutions thus imposed and the current standards in information technology. But who
will tell and who will acknowledge that opportunities are under-utilised? The current
tendency to a more narrative history will not easily yield an incentive in this respect.
This leaves us with the thorny questions whether further insistence on methodo-
logical and technical research in this field will make any sense, and why the historical
information scientist in particular should do so, if the historical community itself is
divided and humanities scholars seem to be also happy without more advanced tech-
nology. A report on the past, present and future of history and computing inevitably
poses the question of scientific progress. Scientific innovation has rarely been based
on the consensus of a majority. In the foregoing chapters the development of histori-
cal computing has been analysed from a panoramic perspective, highlighting meth-
odological and technical issues in relation to gains in historical knowledge. They
open new vistas, showing promising recent developments in computer science and
information science and describe successful experiments in larger projects, mostly
related to digital heritage and digital libraries or in neighbouring fields, like compu-
ter linguistics.
Much of what is blossoming requires further elaboration and has to be translated
into more widely applicable and usable tools. Better infrastructure is needed in order
to guarantee a transfer of results from the methodological and technical level to the
daily practice of historical research. On the contrary, denying these challenges and
opportunities will, in the long run, segregate the study of history from the technical
capabilities currently being developed in the information society and will turn ‘the
computer’ into an awkward tool with limited use and usability for historians.

92 The future
 5.2 Relevant research lines
In summarising the more detailed information on databases, texts, statistical meth-
ods and images from Chapter 3 a few areas of further research stand out (which is not
to say that the following list is complete):
1 Modelling sources and user behaviour; standardisation. More extensive model-
ling, of both, the data structure of historical sources and the way sources are used,
will greatly aid the interoperability between applications and will make tools more
usable. Data modelling applies to the overall structure of sources as well as to data
patterns on a micro-level (like references to persons, locations, money, time etc.).
In addition, the transformation processes between one data structure and another,
as required for specific research purposes, are to be documented through model-
ling too. All this should be aimed at more uniformity in the data structures and pro-
cedures used. The next step will be standardisation, to at least a de facto standard.
2 Supporting editorial processes. At present most historical and literary text editions
use xml. The traditional distinction between typical database data and full text data
becomes rather blurred due to novel xml database software. The current genera-
tion of xml editors possesses some sophisticated features, but is still rather primi-
tive regarding the editorial process of historical information. Their rigid schema-
driven nature may support the encoding of business data which can be completely
modelled beforehand very well, but they are less helpful in editing heterogeneous
historical material, full of exceptions, which have to be handled on an ad hoc basis.
Additionally, more analytical views on the text being edited would be welcome. Fi-
nally, a modern edition is no longer bound to the deadlines of printing. Editing can
be organised as an ongoing process, realised by means of collaborative software.
An edition may become available in instalments, which are reviewed and annotated
online by experts all over the world. Organising this process requires new metho-
dological insights, based on additional research.
3 Discovering structures and patterns. Apart from critical source editions, which will
require a great deal of manual editing, historians will have to cope with lots of raw
text. The application of intelligent computer techniques to unstructured texts may
be promising, in order to structure texts by creating an elementary form of tagging
(semi-)automatically. Next to this is the discovery of patterns and the generation of
knowledge through text mining, semantic parsing, content analysis and techniques
now summarised under the heading of thematics. A related field of increasing
importance is the analysis of images through pattern analysis, possibly in combina-
tion with metadata.
4 Tuning statistical techniques to historical research. Examples of upcoming statis-
tical techniques more suitable to historical problems have been mentioned in the
section on statistical methods above: logistic regression, multilevel regression,
event history analysis and ecological inference and new methods for exploratory
data analysis with an interactive and visual display of results.
5 Tuning information retrieval to historical requirements. Although rarely addressed
as such in historical publications information retrieval forms the core of historical
information processing. Information retrieval is a well-studied field in computer
science; however the complex semantics of historical data and the dimensions of
time and space make special demands. Information retrieval is closely connected

93 The future
 with authoring, in particular with the addition of metadata. The use of metadata
has had a special focus in the research of digital libraries and digital heritage
institutions, but it is still far from clear how these mechanisms can be applied in
smaller historical research projects.
6 Multimedia, reconstruction and simulation. A large and fascinating conglomerate
of different technologies is growing around multimedia: geographical information
systems (gis) as applied to historical data, imaging techniques in the reproduction
and analysis of source texts, 3d-reconstructions of historical buildings and locati-
ons, providing a special sense of presence and allowing explorations not feasible
in 2d-representations. With the exception of historical gis these technologies are
almost virginal fields of study for historians.
7 Publishing historical discourse. The new digital counterparts of printed publica-
tions tend to adhere to old conventions, often longer than strictly necessary, thus
under-utilising the capabilities of the new media. New standards have to be for-
mulated, for example in online journals, for integrating articles with the related
historical data and other resources, which might be published together. Museums
and libraries are creating more and more exhibitions online, but the majority have
difficulties in going beyond the traditional catalogue comprising pictures with a
description. Although the Web invites the use of more explorative structures, these
are still rare. More effort has to be put in cheaper engineering of high quality histo-
rical content.
The realisation of any of these research areas will require organised co-operation
with other information-oriented disciplines, like computer science and information
science. The latter domain is difficult to define. It varies from university to university,
and covers not only the more practical application of information technology in dif-
ferent fields of society, but it is also affiliated with some parts of the social sciences,
like cognitive psychology. In the previous chapters we have used the information
lifecycle (well known in information science) as an organising principle. Here, we
want to summarise these research lines in a slightly different way: as a conceptual
framework, comprising several layers, which are closely interlinked (and therefore
difficult to represent in a 2d-diagram).

94 The future
 Publication

Personalization Time
Presentation

Filtering Space
Selection

Advanced Content Intelligent

Data Mining Parsing Semantics
Statistics Analysis Search

Intelligence
Digital
Production & Editing Sources
Content creation

Figure 5.1 Conceptual research framework for a future historical information science.

The blocks in the middle represent major topics in computer science and informa-
tion science. These are also relevant to historical computing, if the typical factors
– characteristic of historical information problems (on the right side) – are taken
into account: time, space and semantics.63 ‘Content creation’ pertains to the scholarly
preparation and editing of digitised historical source material. The intelligence layer
is a variegated set of intelligent information techniques, aimed at adding structure
and classification to, and deriving knowledge from, the historical content. ‘Selection’
refers to a broad field of information retrieval techniques, which will operate on the
already enriched and structured content. Finally, the presentation layer addresses
the delivery of information in accordance with the user’s level of interest and prefer-
ences.
Some of the research themes mentioned are positioned on a single layer (like the
tuning of statistical techniques), while others will mainly belong to a certain layer,
but will need also expertise from another (like the support of the editorial process,
which is to be primarily classified as content creation, but has aspects of selection and
presentation as well).

5.3 A future infrastructure for historical information science:

Resolving the paradox

5.3.1 Stakeholder communities

If we want to resolve the aforementioned paradox, then it is necessary to create a situ-
ation where this may happen. This pertains directly to mutual relationship between
the different disciplines and institutions involved and the infrastructure of scientific
collaboration. In the Netherlands the following stakeholder communities could play

Here, ‘semantics’ refers to the problem that in historical sources it is often unclear what a source frag-
63

ment means, see for details Chapter 3.

95 The future
 a vital role in a new information technology offensive in the humanities:
• Cultural heritage institutions, e.g., the Koninklijke Bibliotheek (Royal Dutch Libra-
ry), the Nationaal Archief (National Archive), Instituut voor Nederlandse Geschiedenis
(Institute of Netherlands History), Internationaal Instituut voor Sociale Geschiedenis
(International Institute of Social History), major museums, etc.
• Computer-aided projects in humanities, in particular in history, literary studies,
archaeology, etc.
• Computer scientists with interest in cultural applications
• (Historical) information scientists

computer scientists cultural heritage

institutions

(historical) scholars in the

information scientists humanities

Figure 5.2 Stakeholder communities in humanities computing: relatively isolated and thus
under-utilising the field’s capacity.

5.3.2 Patterns of co-operation

In the past several smaller and larger projects have existed, in which some of the
groups mentioned above have managed to collaborate successfully.
What we would like to see, is a broader span of collaboration (Figure 5.3), which
includes all stakeholder communities mentioned above, and leaves room for the
participation of those scholars in the humanities working on interesting informa-
tion problems and wanting to see fundamental solutions applicable beyond their
own projects. If at the same time computer scientists and information scientists join
forces, the envisaged future will come close to reality and historical information will
stand a real chance of succeeding.

96 The future
 computer scientists cultural heritage
institutions

(historical) scholars in the

information scientists humanities

Figure 5.3 Fuller co-operation between the stakeholder communities.

5.3.3 Project models

Collaboration has to be organised. The system of scientific funding and the way
research proposals are assessed is highly relevant and may yield a powerful incen-
tive towards a new direction. Furthermore the organisational structure of projects
will be a critical success factor. Figure 5.4 shows a frequently used model: techni-
cal experts are dropped into a computer-aided project in historical research or in a
cultural heritage institution. This model may work very well, as long as ‘technicians’
restrict themselves to offering services, for example, in the form of system analysis
and programming. Tensions may arise, when the ‘technicians’ want to do their own
research, using the data and user groups in the heritage sector, where the first prior-
ity is to finish a practical project in time and to deliver digital content. Methodological
and technical research activities seldom keep the same pace as the content-driven ac-
tivities. Moreover, both parties may disagree about the direction in which the project
should evolve. At this stage the project’s goal is split. The experiences of the past,
where historical computing merely played an auxiliary role, point out that the combi-
nation of technical and substantial historical research in one single research project
restricts methodological progress. This is therefore not recommended.

97 The future
 ◄ computer-aided projects in
programs

▲
the Humanities

data
enriched
computer scientists drop-in
information scientists publications

◄
Projects
Digital Heritage Institutions

◄
◄
◄
Solutions tend to remain confined
Digital Heritage Data
to the projects in which they have
been developed

Figure 5.4 A familiar collaboration model, not preferred in this context.

A better alternative is shown in Figure 5.5, where two interrelated projects are envis-
aged: one project, run by computer scientists and (historical) information scientists
having specific methodological-technical goals, closely connected with a computer-
aided project in history, literature or any cultural heritage institution.

scientific Projects
publications
◄ Computer science & solutions:
Information science

▲▲
◄
demonstrators
problems

themes

Interaction requires
data

special attention! ▲
practical
computer-aided projects in reports
◄

▲
the Humanities
data
enriched

Feedback
▲ loop

Projects
Digital Heritage Institutions ◄
◄
◄
◄

Digital Heritage Data

Figure 5.5 Recommended dual-project model for collaboration. This leaves room for paral-
lel research tracks, linked, but without impeding each other.

98 The future
 The success of the second model will depend on a few critical factors. These should
be explicitly mentioned.

Effective communication. Good communications between each project is required

relating to the research themes and the problems under scrutiny. The technical
project should propose themes that are of interest for the people involved within
that element of the project, but these themes should not be carried forward without
a careful taking considerable note of the real problems in the partner project. This
process will be necessarily iterative and will require a substantial amount of time.
Organising it will be a major management task in itself. An important precondition
is a mature state of the historical project: the people there should have explored their
problems sufficiently and know already where they are heading. Furthermore, the
computer-aided project should be adequately represented in the project team (or
steering committee) of the technical project, to guide and to watch over the direction
and effectiveness of the problem solving.

Scientific publications outside the humanities. The technical project will have pro-
duce regular scientific publications, which may not be of primary interest for schol-
ars in the humanities. They have to do this in order to continue their own scientific
activities, or, in the case of historical information science, to establish this activity as a
scientific field.

Feedback to the humanities. If this collaboration is to mean anything, the technical

project will have to make its results available to the partner project in a usable form.
The scientific publications mentioned above will not be sufficient. They will likely to
be too far away from the daily research practice in the humanities. Therefore, im-
portant additional deliverables are demonstrators and technical reports that discuss
implementation in detail. A demonstrator in this context should be a working system
and not simply a mock-up. However, the right balance will have to be found and the
technical project must not slide towards programming on demand. The deliverables
meant here are not to be confused with the technical and programming assistance
that is supplied by supporting staff in arts faculties and cultural heritage institutions.
The results may be still beyond the horizon of many less computer-literate scholars
in historical and literary studies. Only in this way historical information science can
bring methodological progress for the field as a whole.

5.4 Is that all there is?

This report started with a rather gloomy view on the scientific status of history and
computing, and this evaluation has been repeated several times in this overview. On
one hand, we have painted the rich and varied landscape of computer applications
in history and in the humanities at large and how this field has evolved over the past
two decades. On the other hand, the methodological progress was less energetic
than it might have been. From the mid 1990s onwards, graphical user interfaces for
operating systems and sophisticated ‘office’software have provided the illusion that

99 The future
 computing could be easy for a large community of historians and could help them
to conquer their basic resistance to technology. The Internet has brought scholars
and their data together and made distributed, large- scale projects feasible. Digital
archives, digital libraries and rich web sites of all kinds of cultural heritage institu-
tions became the nodes in a large information web. These focal points have particu-
larly contributed to the more optimistic side of historical computing in the recent
past. Large-scale projects have been most fruitful in developing techniques and tools.
However, this has often been because the large amount of data to be processed has
required automation, and thus justified the development of tools, and the creation of
a representative test bed.
Now we are at the edge of a new development, which may unite those who are
interested in information problems (the computer scientists and information sci-
entists) and those who are the treasurers of a wealth of information problems (the
scholars in the humanities). As McCarty pointed out, interdisciplinary activity is not a
matter of simply importing or exporting ideas and methods, but it is constituted by a
unifying perspective at the intersection of two or more fields (McCarty, 2001). Taking
two important disciplines in this field as an example, history and computer science,
the unifying computing perspective lies in reaching an intermediate level of abstrac-
tion with regard to formulating problems and solutions. Historians are inclined to
overestimate the uniqueness of their problems, while computer scientists live with
the beauty of universal solutions for rather abstract problems. Problems will grow
more complex (but also more meaningful) when defined closer to practice. Finding
the right balance will require an organisational context where scientists of different
denomination meet and work together. We hope that we have outlined such a context
convincingly.
One final comment must be added. Any structural change in mentality will require
a substantial effort in education. If the envisaged new infrastructure can be realised,
a structural co-operation with both undergraduate teaching in universities and with
postgraduate research schools must be set up to consolidate this new approach and to
connect historical information science with the life-style of a new generation which
has grown up with sophisticated information technology.

100 The future

Appendix
Possible themes for historical
information research
In Section 5.2, an outline was sketched for a new research program in historical in-
formation science. Below a number of themes are presented which fit well within the
range of research topics such a research program could address. The themes are ar-
ranged according to the life cycle of historical information. It is important to note that
they do not comprise a full-grown research program; they need to be thought of as
examples, nothing more and nothing less. But they give some idea about the issues
that could be handled, as a field of scientific research in which historians cooperate
with information scientists, on a level of abstraction which makes it possible to make
generalisations and tools that can be disseminated to a large audience of historians.

1 Creation

Enrichment Editing

Creation Retrieval

Presentation Analysis

1.1 Generic modelling of historical cultural sources

Historians have long had access to source commentaries describing the characteris-
tics of certain types of historical sources, like parish registers of baptisms, weddings
and funerals, legal deeds and estate inventories, to name just a few. As an extension
of this, it must be possible to compile generic data models for sources – in the same
way as tei has designed generic dtds for literary texts. The term ‘generic’ refers to
the data structure shared by different variants of the same main source type, and to
a basic functionality which characterises the source type in question. If such generic
data models were available to us, data models for specific research exercises would be
quicker to produce, and the resulting data collections easier to share.
Because the technologies used for processing texts and those for processing tra-
ditional database information have ceased to be so clearly distinguishable since the
arrival of xml, it is advisable that these models be defined at a high conceptual level
(with uml class diagrams, for example), not directly related to a particular technology.
There are also more fundamental questions. The practicability of tei standards
is a matter of debate, not only due to their prescriptive character but also because of
questions about the nature of text (hierarchical structure versus network structure).
Moreover, one can ask how deep and how detailed tagging should be, and what

101 Appendix
 should be done by the editor and what by the user-researcher. Dynamic modelling
should also be possible, based upon the researchers’ interactions with the text. The
good thing about this is that xml is used to express the model, but the model itself
is becoming much more flexible, is able to adapt to different interpretations and no
longer overburdens the researchers.

question : How can historical cultural source material be modelled generically in such a
way that physical data models are produced more quickly and uniformly? How can we
actually create and publish a series of generic models?

sub-question:
The modelling of multimedia sources – such as illustrations and music notation
– combined with textual material. tei provides various ways of describing the re-
lationships between text and images in ‘bimedia’sources, but more specific guide-
lines are desirable. This issue is also important to historical research programmes
like the visual culture programme of the Dutch Institute for Netherlands History.

associated themes : 4.6

1.2 Ontologies for historical cultural research

Research in the humanities generates new knowledge, which can also be stored in in-
formation systems. To do this, however, a system of unambiguous terms – an ‘ontol-
ogy’64 – is needed. There are currently several projects under way which add specific
information, in the form of metadata, to existing digitised sources. Much of this work
is done at the heritage institutions which hold the original source material.
Following on from this, and building upon it, researchers in the humanities should
be able to provide their own products with a terminology that makes knowledge
accessible through search engines and enables a more effective execution of search
queries. The key to designing new ontologies and expanding existing ones is finding
domain-specific applications of accepted standards. The actual usability of the design
by the target group requires particular attention.

question : How can existing techniques for making knowledge accessible, and in particu-
lar those related to the design of ontologies, be applied to the practice of historical cultural
research?

associated themes : 2.1, 4.2, 4.7

64
Ontology: ‘An explicit formal specification of how to represent the objects, concepts and other entities
that are assumed to exist in some area of interest and the relationships that hold among them. It implies
the hierarchical structuring of knowledge about things by subcategorising them according to their essen-
tial (or at least relevant and/or cognitive) qualities. A set of agents that share the same ontology will be able
to communicate about a domain of discourse without necessarily operating on a globally shared theory.’
(Based upon Hyperdictionary).

102 Appendix
 1.3 Data models for metadata
Research into the usability of rdf and Topic Maps for the storage of metadata. Use
of these data models in combination with other – xml and non-xml – formats and
models. For example, mets. The interoperability of different systems, such as gis and
thesauri, is associated with this.

question : How can metadata be modelled?

2 Enrichment

Enrichment Editing

Creation Retrieval

Presentation Analysis

2.1 The embedding of historical cultural knowledge and insight so as to enrich

information systems
The meaning of objects and terms from the past can only be understood in a his-
torical cultural context. Therefore, that context must be included when enriching
information systems that contain or describe cultural heritage digitally. Research is
needed into the question how this can best be done.
If it proves possible to ‘store’ the historical cultural context as metadata, then a fur-
ther problem soon arises: the addition of metadata can be accelerated considerably if
the extraction of terms and the attribution of meanings to metadata is automated. In
the case of historical cultural data, this can only be done when the historical cultural
context of a source is included in the process of extraction and labelling.

question : How can historical data be provided with metadata pertaining to its historical
context? If metadata can be provided with metadata pertaining to its historical context,
is it also possible to extract and label the metadata automatically?

associated themes : 1.2, 5.4, 5.5

2.2 Grammatical help

When developing thesauri or other types of metadata, the meaning of a term can be
defined more precisely if its grammatical position is included. So, when adding meta-
data in order to index information systems, it is advisable to investigate how the ad-
dition of such information on the grammatical position of terms can be achieved and
what improvements in quality might result from it. This is part of a broader topic: the
parsing of a text and the use of the information thus obtained to make that text more
useable. For example, there is a particular need for a parser of Middle Dutch and for a
diachronous parser that can deal with Dutch from different periods.

103 Appendix
 question : How can the grammatical position of terms be used to refine their meanings,
thus resulting in greater precision in their indexing and easier access to the text?

associated themes : 5.4

2.3 Changes in the meaning of information over time

For historians, time is a very relevant context. But much of the historical informa-
tion does not take this into account or does so only implicitly. A trade can shift in
relevance or status, a location can change in size or geographical context. How do you
include such time-sensitive context, which is not known in advance, in an informa-
tion system?

question :How can changes in the meaning of terms over time be taken into consideration
when designing a historical information system?

associated themes : 4.2, 4.5

3 Editing

Enrichment Editing

Creation Retrieval

Presentation Analysis

3.1 From historical source to XML text

The current technique for recording textual sources uses xml. Editing an xml text is
a fairly technical task, requiring considerable knowledge of the mark-up language
when standard xml editors are used. Those who have to do this work because they
possess the necessary subject knowledge may well not have the required technical
expertise. Research into user-friendly procedures for converting source material to
xml form therefore seems desirable.
The existing technology does make it possible to create a modified xml editing
environment, but this requires considerable initial technical effort. In addition, it
presupposes a more-or-less fixed data model. But historical data tends to be inconsist-
ent and full of exceptions. One possible solution might be found in a modified form
of editing, whereby it is easy to change the data model as one works, with the editor
itself providing assistance, thanks to its own in-built intelligence.

question : Is it possible to design an XML editor tailored to the functional requirements of

editing historical cultural sources?

associated themes : 5.4, 6.1

104 Appendix
 3.2 Architecture of digital publication
The formulation of standards for reliable digital publication. What should such a
publication look like, which data structure is desirable and which functions should be
available to the end user?

question : Which requirements should be met by a source or text edition in digital form?

3.3 Publication of sources as an ongoing online process

The prominence of the Internet and the creation of broadband networks make it a re-
alistic proposition to work jointly on the publication of sources at different locations
and to regard editing as an ongoing process – with interim results being made avail-
able to users immediately. The continual publishing of illustrative material requires
particular consideration in this context. In other disciplines, online co-operation has
already become commonplace in the form of so-called ‘collaborations’. In particular,
the joint online annotation of illustrative material should be investigated further. A
related topic is the availability of tools for virtual communities.

question :Is joint online work on texts for publication feasible and desirable, and which
functionality is required to achieve it?

4 Retrieval
Enrichment Editing

Creation Retrieval

Presentation Analysis

4.1 Historical search strategies

How do historians search for their information? What strategies do they use? Are
these strategies based upon traditional handbooks such as Zoeken en Schrijven
(Search and Write), or have new ones been adopted, using the new forms of informa-
tion gathering offered by the Internet? If the latter were the case, would it be wise to
analyse these new search strategies and to evaluate their effectiveness? Moreover,
there is also a new and growing group of interested lay people.

question :Which search strategies are used by historians, and which of these are most ef-
fective? What implications does this have for search systems and their user interfaces?

associated themes : 4.3.

105 Appendix
 4.2 Semantic analysis of search queries
When developing search procedures for the historical cultural domain, changes
of meaning and relevance over time must be taken into consideration. A possible
way of doing this is to apply a semantic analysis of the search query. This technique
translates the original query into a new one, which conveys the user’s intention to
the underlying information system more accurately. The use of ontologies in search
queries is worth considering.

question :Is it possible to consider changes of meaning and relevance over time when
formulating search queries on history? Could semantic analysis be a solution?

associated themes : 1.2, 2.3

4.3 Linguistic search assistance

Simple searches using strings and ‘wild cards’ are often inadequate. More generically
applicable solutions for the reduction of search terms to word stems, searches using
stems and retrograde searches – that is, based upon word endings – are desirable.
Another option to consider is the use of pattern matching. Such solutions are avail-
able to some extent in commercial packages. As yet, however, there are no package-
independent solutions tailored specifically to historical literature research.

question :How can the search process, and textualisation in particular, be improved from
a linguistic point of view?

4.4 Text mining, automatic tagging and content analysis

There is far more interesting source material available than can be marked up by
hand. This fact forces us to consider the idea of information retrieval from ‘raw’ text
– using text mining, for example. Experience with techniques of this kind is mainly
confined to modern texts. Very little research has yet been done on their application
in historical investigation. In line with this is the question to what extent information
can be marked automatically in ‘raw’ texts. The progress with part-of-speech taggers
is encouraging. Can the step up to the semantic level be made? In this respect, we
should also look at recent research into content analysis and thematics.

question :
To what extent can information be gleaned from texts which have not been
marked up manually?

4.5 Question-answering techniques

When answering factual questions, it is better to produce a factual answer than
to simply generate a list of documents in which the answer may – or may not – be
found. Question-answering techniques offer this potential. But in the field of histori-
cal cultural studies, a factual answer may sometimes be quite correct but only half
the story. Is it possible to determine, automatically or otherwise, when a question-an-
swering technique is appropriate and when a more wide-ranging answer is required?

106 Appendix
 In addition, question-answering techniques will have to be able to include changes in
meaning and relevance over time in answers to questions of a historical nature.

question : Is it possible to reveal when question-answering techniques should be used to

answer search queries, and when they should not? Is it possible to take changes of mean-
ing and relevance over time into consideration when using question-answering tech-
niques?

associated themes : 2.3, 4.1.

4.6 Searching multiple information systems

In order to be able to search multiple information systems with different structures
and enrichment methods, it is necessary to develop systems that enable interoper-
ability. This is no easy task, as historical sources are retrieved from information
systems in many different ways. Moreover, those systems can contain various types
of data: running text, tables, images, audio or any combination thereof. To be able
to search the sources effectively, procedures are required which can cope with the
variety of retrieval systems used and the many possible combinations of text, images
and audio.

question : Can procedures be developed for the retrieval of information from multiple in-
formation systems? If so, can these be made independent of the type of media upon which
those systems are based?

4.7 User profiles during searches

Historical information systems will be used by a wide range of people for a variety of
reasons. Based upon the search queries submitted, a system can compile an ‘ad-hoc’
user profile that enables the answers to the questions posed to be tailored to that
user’s knowledge domain (‘adaptive hypertext’). But, in addition to this ‘ad-hoc’ one,
it is also possible to have a system – either on its own or in consultation with the user
– gradually build up a query-independent user profile. Users can submit this profile
at a later time or when visiting other search systems, so that the answers to their
questions are tailored to their knowledge.

question : Is it possible, based upon the search queries submitted, to compile an ad-hoc
user profile which enables those queries to be answered more precisely? And is it possible to
compile a permanent, query-independent user profile?

associated themes : 1.2.

107 Appendix
 5 Analysis

Enrichment Editing

Creation Retrieval

Presentation Analysis

5.1 Analysis of numerical data: multi-level statistical techniques

Multi-level techniques make it possible to incorporate variables into a statistical anal-
ysis at a higher aggregated level and to produce good estimates of their effects. In this
way aggregated data about, for example, a research subject’s place of residence – its
population, geographical location and economic circumstances – can be combined
with individual variables such as gender and year of birth. As of yet, this technique is
not used very much in historical study, even though the data required for it is gener-
ally available. It is therefore worth investigating the technique’s potential in historical
research. Such a study should incorporate trials of methods for linking micro-level
and aggregate-level data.
Analysis of the opposite type is also possible: research at the aggregate level in-
corporating information about individual people. Again, such ‘ecological inference’
research is still rare in the study of history.

question : What potential is there for the application of multi-level record linkage and
multi-level regression analysis in quantitative historical research?

5.2 Analysis of numerical data: event history analysis

Event history analysis is a technique for calculating the chances of a particular event
occurring, based upon circumstances and events experienced by the research sub-
jects during their lives. This appears to be a good technique for quantitative research
into life histories. But the method of analysis does involve certain assumptions which
might be at odds with historical reality. If so, what are the consequences and can they
is overcome?

question : What potential is there for the application of event history analysis in quantita-
tive historical research?

5.3 Analysis of numerical data: simulation techniques

One of the positive aspects of historical research is that the outcome of the proc-
ess being studied is already known. This opens up particular opportunities for the
evaluation of simulations, and therefore for the way in which this technique can be
applied.

108 Appendix
 question : What potential is there for the application of simulation techniques in histori-
cal research?

5.4 Analysis of textual data: text-analysis techniques in historical research

Perhaps one of the most promising text-analysis techniques in the study of history is
content analysis – providing that the historical context is incorporated into the analy-
sis. But there are also other forms of text analysis which have seldom, if at all, been
applied to historical research. Many of these are based upon large written corpora,
the grammatical structure of which is already known. With the appearance of histori-
cal text files, improved methods of revealing the grammatical structure of texts and
the use of xml to record this information as metadata, it is becoming possible to con-
duct more qualitative textual analysis. This type of research is now attracting many
historians. In this respect, so-called ‘textual statistics’ merit particular consideration.

question :What potential is there for the analysis of ‘enriched’ texts in historical research?

associated themes : 2.1, 3.1.

5.5 Image analysis: content-based image-research techniques

Image analysis uses techniques designed to extract content-based ‘meaning’ from
one or more images by means of digital analysis. For example, ‘this image contains a
church/a tree’. But can this technique also be applied to historical images – to state,
for example, ‘this image contains William of Orange/Muiderslot Castle’.
The technique just described uses only whole or partial images for its search and
analytical procedures. This application seems rather limited, though. On the other
hand, its combination with textual information opens up many additional possibili-
ties – particularly in the analysis of historical images. So there is every reason to
investigate how such contextual information could be used in analysing images.

question : Is it possible to use content-based image-research techniques for the analysis of

historical images? What potential is there for image analysis using contextual informa-
tion?

associated themes : 2.1.

6 Presentation

Enrichment Editing

Creation Retrieval

Presentation Analysis

109 Appendix
 6.1 Dynamic generation and presentation of historical information
Texts structured using xml can be broken down into sections. These so-called ‘com-
ponents’ can then be presented in different forms and context, in accordance with the
user’s wishes. This opens up a range of opportunities. The system can imperceptibly
record the user’s choices and modify the presentation accordingly (‘adaptive hyper-
text’). Alternatively, the user can specify what form the presentation takes (‘dynamic
content’). If the information is to be presented on the web, the location of this process
can also vary: either the provider’s server or the user’s pc.
In order to reach this point, thorough research is required into the structure of
the textual material to be used – perhaps involving such things as genre theories or
Rhetorical Structure Theory – so as to create models from which the computer can
synthesise presentations.

question :How can presentations tailored to the needs of a specific audience, or to a large
extent configurable by the users themselves, be generated from well-structured historical
cultural material?

sub-question
The above can be expanded to include the generation of ‘virtual exhibitions’ of
multimedia material. The rare examples of successful web projects show that such
an exhibition can be much more than ‘pictures with text’. Exploration, animation
and elements of play can be incorporated into the ‘story’, producing a far more
intensive experience. Although staging a virtual exhibition is an artistic activity,
which has little direct relationship with algorithmisation, the computerisation of its
components certainly seems a subject worthy of research.

associated themes : 3.1.

6.2 Visualisation of historical research subjects and results

By applying information technology, it is possible to recreate a visual representation
of historical data, which has come down to us in written form but actually refers to a
past three-dimensional reality.
− Historic buildings can be reconstructed on screen using 3d techniques, giving the
user a much more intensive experience than could be obtained from studying the
sources.
− Animations can be used to enhance the study of works of art (on-screen rotation,
panoramic views, exploded views et cetera).
− Co-ordination of historical data in time and space:
− Geographical information systems (giss) have been in use for the historical study
of cities, towns and regions for some time now;
− Timelines – which never really came into their own in printed historical atlases
because of the medium’s limitations – can be shown on screen in a rich variety of
forms.
− Visual representation of textual structures.
− Different ‘views’ of the same text, each with its own structure and including or

110 Appendix
 omitting certain information in order to highlight certain aspects, visualise the
frequency of textual properties or show the structure of a manuscript. This theme
encroaches upon a relatively new field, much of which is yet to be mapped out. Not
only is there still a lot of interesting detailed research to be carried out, but a clear
overview is also lacking.

question : Which main themes in visualising historical material, in the broad sense of the
term, are still to be revealed? Can historical cultural functionality be formulated generi-
cally? With which aspects of information technology do those themes correspond? How
efficient is the available software?

7 Central themes

Editing
Enrichment

Creation Durability
Usability Retrieval
Modelling

Presentation Analysis

7.1 Digital durability of research products

Digital durability is currently the subject of considerable attention, particularly with
respect to the archiving of computerised administrations. Another aspect is the
durability of digital research products such as databases and digital sources with em-
bedded scripts. The proper preservation of these for future use must be guaranteed,
which requires effective archiving of digital historical data.

7.2 Usability of software tools

As a rule, researchers working in the humanities use off-the-shelf software products.
There are only a very few specialist programmes developed for this field of study. An
inventory needs to be compiled covering both categories, particularly with regard to
their range of applications. A web site on historical software would be particularly
useful! Research is also needed into the question how useful particular tools are for

111 Appendix
 the tasks they are actually carrying out. By considering usability more systematically,
the need for specifically-developed software tools will become clearer.

7.3 Modelling of historical cultural information systems

The methods and techniques used to design information systems are still largely
based upon situations in the commercial world. There, data structures and desired
functionality are usually determined within the organisation itself. The design fol-
lows the structure of the work process – which we can reasonably assume is fairly
stable, at least for a certain period.
Information systems for historical cultural research, on the other hand, are char-
acterised by the fact that whilst – apart from exceptions caused by the heterogeneous
nature of the source material – the overall structure of the material is clear, functional
requirements tend to evolve relatively rapidly. After all, as the understanding of the
subject increases it prompts new questions and processes. The tool must be able to
evolve at the same pace, a factor which needs to built into the design and incorpo-
rated into the project plan.
It would be useful to formulate a framework of criteria, using which it-related
historical cultural research projects can be described and evaluated.

sub-question:
Content management in historical cultural research: Content management sys-
tems are used by organisations to store, administer, process and publish all kinds
of data. A wide of range of such software is available commercially. What all the
packages have in common is that they support the information lifecycle. Once we
succeed in modelling the lifecycle of historical cultural research – at a certain level
of abstraction – then it will become possible to better harmonise and integrate
different software tools. A content management system for this academic sector
should be virtual in nature, perhaps consisting of a series of separate but co-ordi-
nated and standards-based programs.

112 Appendix
 References

E. Aarseth (1998). ‘From Humanities Computing to Humanistic Informatics: Creating a Field of

Our Own’, Paper presented at The Future of the Humanities in the Digital Age, Bergen, Norway.
K. Abram (2002). ‘Electronic Textuality. A bibliographic essay’. Derived from the World Wide Web:
http://www.mantex.co.uk/ou/resource/elec_txt.htm
P. Adman (1997). ‘Record Linkage Theory and Practice: A Matter of Confidence’, History and Com-
puting 9 (1-3): 150-155.
P. Adman, S.W. Baskerville and K.F. Beedham (1992). ‘Computer-Assisted Record Linkage: or How
Best to Optimize Links Without Generating Errors’, History and Computing 4 (1): 2-15.
R.C. Allen (1998). ‘Capital accumulation, the Soft Budget Constraint and Soviet Industrialisation’,
European Review of Economic History 2 (1): 1-24.
R.C. Allen and I. Keay (2001). ‘The first Great Whale Extinction: the End of the Bowhead Whale in
the Eastern Arctic’, Explorations in Economic History 38 (4): 448-477.
P.D. Allison (1984). Event History Analysis. Regression for Longitudinal Event Data. Beverly Hills /
New York: Sage.
G. Alter (1998). ‘L’ Event History Analysis en Démographie Historique. Difficultés et Perspectives’,
Annales de Demographie Historique (2): 25-50.
G. Alter and M.P. Gutmann (1999). ‘Casting Spells Database Concepts for Event-History Analysis’,
Historical Methods 32 (4): 165-176.
A. Andreev, L. Borodkin and M. Levandovskii (1997). ‘Using Methods of Non-Linear Dynamics in
Historical Social Research: Applicaiton of Chaos Theory in the Analysis of the Worker’s Move-
ment in Pre Revolutionary Russia’, Historical Social Research 22 (3-4): 64-83.
M. Artzrouni and J. Komlos (1996). ‘The Formation of the European State System a Spatial ‘Preda-
tory‘ Model’, Historical Methods 29 (3): 126-134.
S. Aumann, H.-H. Ebeling, H.-R. Fricke, et al. (1999). ‘From Digital Archive to Digital Edition’,
Historical Social Research 24 (1): 101-144.
R.H. Baayen, H. van Halteren and F.J. Tweedie (1996). ‘Outside the Cave of Shadows. Using Syn-
tactic Annotation to Enhance Authorship Attribution’, Literary and Linguistic Computing 11 (3):
121-131.
J. Bacher (1989). ‘Einführung in die Clusteranalyse mit SPSS-X für Historiker und Sozialwissens-
chaftler’, Historical Social Research 14 (2): 6-167.
J.T. Barry, S. Walby and B. Francis (1990). ‘Graphical Exploration of Work History Data’, Quad.
Statist. Mat. Appl. Sci. Econ Sociali. 12: 65-74.
T. Bengtsson (1999). ‘The Vulnerable child. Economic Insecurity and Child Mortality in Pre-in-
dustrial Sweden: a Case Study of Västanfors, 1757-1850’, European Journal of Population/Revue
Européenne de Démographie 15 (2): 117-151.
T. Bengtsson and G. Broström (1997). ‘Distinguishing Time Series Models by Impulse Response: a
Case Study of Mortality and Population Economy’, Historical Methods 30 (4): 165-171.
T. Bengtsson and M. Dribbe (2002). ‘Fertility Response to Economic Stress. Deliberate control or
Reduced Fecundability?’ Lund Papers in Economic History 78.
T. Bengtsson and M. Lindström (2003). ‘Airborne Infectious Diseases during Infancy and Mortality
in Later Life in Southern Sweden, 1766-1894’, International Journal of Epidemiology 32: 286-294.
G. Bloothooft (1994). ‘Corpus-based Name Standardization’, History and Computing 6 (3): 153-167.
G. Bloothooft (1995). ‘Multi-Source Family Reconstruction’, History and Computing 7 (2): 90-103.
G. Bloothooft (1998). ‘Assessment of Systems for Nominal Retrieval and Historical Record Lin-
kage’, Computers and the Humanities 31: 39-56.
H.-P. Blossfeld and G. Rohwer (2002). Techniques of Event History Modelling . New Approaches to
Causal Analysis. Mahwah, New Jersey: Erlbaum.

113 References
 O. Boonstra (1994a). ‘Automatisering en het Kadaster. Het Gebruik van de Computer bij Histo-
risch-kadastraal Onderzoek’, Cahier VGI 8: 114-123.
O. Boonstra and M. Panhuysen (1999). ‘From Source-oriented Databases to Event-history Data files:
a Twelve-step Action Plan for the Analysis of Individual and Household Histories’, History and
Computing 10 (1-3): 1-9.
O.W.A. Boonstra (1990). ‘Supply-side Historical Information Systems. The Use of Historical Data-
bases in a Public Record Office’, Historical Social Research 15: 66-71.
O.W.A. Boonstra (1994b). ‘Mapping the Netherlands, 1830-1994: The Use of nlkaart’, in: M. Goe-
rke, Coordinates for Historical Maps. St. Katharinen: Halbgraue Reihe, 156-161.
O.W.A. Boonstra (2001). ‘Breukvlakken in de Eenwording van Nederland’, in: J.G.S.J.v. Maarse-
veen and P.K. Doorn, Nederland een eeuw geleden geteld. Een terugblik op de samenleving rond 1900
Amsterdam: iisg, 277-298.
O.W.A. Boonstra, L. Breure and P.K. Doorn (1990). Historische Informatiekunde. Hilversum: Verlo-
ren.
O.W.A. Boonstra, P.K. Doorn and F.M.M. Hendrickx (1990). Voortgezette Statistiek voor Historici.
Muiderberg: Coutinho.
K. Börner and C. Chen (2002). ‘Visual Interfaces to Digital Libraries. Lecture Notes in Computer
Science.’ Heidelberg: Springer-Verlag.
L. Borodkin (1996). ‘History and Computing in the ussr/Russia: Retrospection, State of Art, Pers-
pectives’. Derived from the World Wide Web: ������������������������������������
http://www.ab.ru/~kleio/aik/aik.html
L. Borodkin and M. Svishchev (1992). ‘El Sector Privado de la Economia Sovietica en Los Anos
Veinte’, Revista de Historia Económica 10 (2): 241-262.
B. Bos and G. Welling (1995). ‘The Significance of User-Interfaces for Historical Software’, in: G.
Jaritz, I.H. Kropac, and P. Teibenbacher, The Art of communication. Proceedings of the Eight Inter-
national Conference of the Association for History and Computing, Graz, Austria, August 24-27, 1993
Graz: Akademische Druck- iund Verlaganstalt, 223-236.
J. Bradley (1994). ‘Relational Database Design and the Reconstruction of the British Medical Profes-
sion: Constraints and Strategies’, History and Computing 6 (2): 71-84.
L. Breure (1992). ‘Tools for the Tower of Babel: Some Reflections on Historical Software Enginee-
ring’, in: Eden or Babylon? On Future Software for Highly Structured Historical Sources. St. Kathari-
nen: Max-Planck-Insitut für Geschichte, Göttingen, 23-36.
L. Breure (1994a). ‘How To Live With XBase: the Socrates Approach’, in: F. Bocchi and P. Denley,
Storia & Multimedia. Proceedings of the Seventh International Congress Association for History &
Computing. Bologna: Grafis Edizioni, 477-484.
L. Breure (1994b). ‘SocrATES: Tools for Database Design and Management’, in: H.J. Marker and
K. Pagh, Yesterday. Proceedings from the 6th international conference Association of History and
Computing, Odense 1991 Odense: Odense University Press, 140-148.
L. Breure (1995a). ‘Altis. A Model-based Approach to Historical Data-entry’, Cahier VGI 9: 178-188.
L. Breure (1995b). ‘Interactive Data Entry: Problems, Models, Solutions’, History and Computing 7
(1): 30-49.
L. Breure (1999). ‘In Search of Mental Structures: A Methodological Evaluation of Computerized
Text Analysis of Late Medieval Religious Biographies’, History and Computing 11 (1-2): 61-78.
M. Brod (1998). ‘Computer Simulation of Marriage Seasonality’, History and Computing 10 (1-3):
10-16.
A.S. Bryk and S.W. Raudenbush (1992). Hierachical Linear Models. Applications and Data Analysis
Methods. Newbury Park: Sage.
P. Burke (2001). Eyewitnessing: The Uses of Images as Historical Evidence. Ithaca: Cornell University
Press.
L. Burnard (1987). ‘Primary to Secondary: Using the Computer as a Tool for Textual Analysis in His-
torical Research’, in: P. Denley and D. Hopkin, History and Computing. Manchester: Manchester
University Press, 228-233.

114 References
 L. Burnard (1989). ‘Relational Theory, sql and Historical Practice’, in: C. Harvey, History and Com-
puting II. Manchester, New York: Manchester University Press, 63-71.
L. Burnard (1990). ‘The Historian and the Database’, in: E. Mawdsley, N. Morgan, L. Richmond, and
R. Trainor, History and Computing III. Historians, Computers and Data. Applications in Research
and Teaching. Manchester, New York: Manchester University Press, 3-7.
J. Burrows (2003). ‘Questions of Authorship: Attribution and Beyond’, Computer and the Humanities
37: 5-32.
J.F. Burrows (1987). Computation into Criticism: A Study of Jane Austen and an Experiment in Method.
Oxford: Clarendon Press.
J. Burt and T.B. James (1996). ‘Source-Oriented Data Processing. The Triumph of the Micro over
the Macro?’ History and Computing 8 (3): 160-168.
J.W. Buzydlowski, H.D. White and X. Lin (2002). ‘Term Co-occurrence Analysis as an Interface for
Digital Libraries’, in: C.C. K. Börner, Visual Interfaces to Digital Libraries. Heidelberg: Springer
Verlag, 133-144.
C. Campbell and J. Lee (2001). ‘ Free and unfree Labor in Qing China: Emigration and Escpae
among the Bannermen of Northeast China, 1789-1909’, History of the Family 6 (4): 455-476.
C. Campbell and J.Z. Lee (1996). ‘A Death in the Family: Household Structure and Mortality in
Rural Liaoning: Life-event and Time-series Analysis’, History of the Family 1 (3): 297-328.
K.D. Cartwright (2000). ‘Shotgun Weddings and the Meaning of Marriage in Russia: an Event
History Analysis’, History of the Family 5 (1): 1-22.
C. Chen and K. Börner. (2002). ‘Top Ten Problems in Visual Interfaces to Digital Libraries’, in: K.
Börner and C. Chen, Visual Interfaces to Digital Libraries. Heidelberg: Springer Verlag, 226-231.
T. Coppock (1999). ‘Information Technology and Scholarship: Applications in the Humanities and
Social Sciences.’ Oxford: Oxford University Press for the British Academy.
L. Corti (1984a). ‘Automatic processing of art history data and documents’. Pisa: Scuola Normale
Superiore.
L. Corti (1984b). ‘Census: Computerization in the history of art.’ Los Angeles: The J. Paul Getty
Trust.
L. Corti and M. Schmitt (1984). ‘Automatic processing of art history data and documents. Second
International Conference - proceedings.’ Pisa: Scuola Normale Superiore.
P. Cowley and J. Garry (1998). ‘The British Conservative Party and Europe: the Choosing of John
Major’, British Journal of Political Science 28 (3): 473-499.
G. Crane (2000). ‘Designing Documents to Enhance the Performance of Digital Libraries. Time,
Space, People and a Digital Library on London’, D-Lib Magazine 6 (7/8).
G. Crane, D.A. Smith and C.E. Wulfman (2001). ‘Building a Hypertextual Digital Library in the
Humanities: a Case Study on London’, Proceedings of the first ACM/IEEE-CS joint conference on
Digital libraries: 426-434.
C. Daassi, M. Dumas, M.-C. Fauvet, et al. (2000). ‘Visual Exploration of Temporal Object Databa-
ses.’ Presented at bda 2000, Blois, France.
O. Darné and C. Diebolt (2000). ‘Explorations in Monetary Cliometrics. The Reichsbank: 1876-
1920’, Historical Social Research 25 (3-4): 23-35.
H.R. Davies (1992). ‘Automated Record Linkage of Census Enumerators’ Books an Registration
Data: Obstacles, Challenges and Solutions’, History and Computing 4 (1): 16-26.
M. Debuisson (2001). ‘The Decline of Infant Mortality in the Belgian Districts at the Turn of the
20th Century’, Belgisch Tijdschrift voor Nieuwste Geschiedenis 31 (3-4): 497-527.
H.E. Delger (2003). Nuptiality and Fertility: an Investigation into Local Variations in Demographic
Behaviour in Rural Netherlands about 1800 Hilversum: Verloren.
F. Deng (1997). ‘Information Gaps and Unintended Outcomes of Social Movements: The 1989
Chinese Student Movement’, American Journal of Sociology 102 (4): 1085-1112.
P. Denley (1994a). ‘Models, Sources and Users: Historical Database Design in the 1990s’, History
and Computing 6 (1): 33-43.

115 References
 P. Denley (1994b). ‘Source-Oriented Prosopography: Kleio and the Creation of a Data Bank of
Italian Renaissance University Teachers and Students’, in: F. Bocchi and P. Denley, Storia &
Multimedia. Proceedings of the Seventh International Congress Association for History & Computing.
Bologna: Grafis Edizioni, 150-160.
K. Depuydt and T. Dutilh-Ruitenberg (2002). ‘tei encoding for the Integrated Language Database
of 8th to 21st-Century Dutch’, in: C. Povlsen, Proceedings of the Tenth EURALEX International
Congress, EURALEX 2002 Copenhagen, Denmark, August 13-17, 2002 683-688.
R. Derosas (1999). ‘Residential Mobility in Venice, 1850-1869’, Annales de Démographie Historique
(1): 35-61.
C. Diebolt and V. Guiraud (2000). ‘Long Memory Time Series and Fractional Integration. A
Cliometric Contribution to Franch and German Economic and Social History’, Historical Social
Research 25 (3-4): 4-22.
C. Diebolt and J. Litago (1997). ‘Education and Economic Growth in Germany before the Second
World War: an Econometric Analysis of Dynamic Relations’, Historical Social Research 22 (2):
132-149.
A. Diekmann and H. Engelhardt (1999). ‘The Social Inheritance of Divorce: Effects of Parent’s
Family Type in Postwar Germany’, American Sociological Review 64 (6): 783-793.
J.d. Does and J. Voort van der Kleij (2002). ‘Tagging the Dutch parole Corpus’, in: M. Theune,
Computational Linguistics in the Netherlands 2001 Selected Papers from the Twelfth CLIN Meeting.
Amsterdam, New York: Rodopi, 62-76.
P. Doorn (2000). ‘The Old and the Beautiful. A Soap Opera about Misunderstanding between His-
torians and Models’, in: L. Borodkin and P. Doorn, Data Modelling Modelling History. Proceedings
of the XI International Conference of the Association for History and Computing, Moscow, August
1996 Moscow University Press, 2-29.
P.K. Doorn and J.T. Lindblad (1990). ‘Computertoepassingen in de Economische Geschiedenis, in
het bijzonder bij Tijdreeksanalyse’, Tijdschrift voor Geschiedenis 103 (2): 326-341.
S. Drobnic, H.-P. Blossfeld and G. Rohwer (1999). ‘Dynamics of Women’s Employment Patterns
over the Family Life Course: a Comparison of the United States and Germany’, Journal of Mar-
riage and the Family 61 (1): 133-146.
T. Dutilh and T. Kruyt (2002). ‘Implementation and Evaluation of parole pos in a National
Context’, in: C.P.S. Araujo, Proceedings of the third International Conference on Language Resources
and Evaluation, ELRA. Paris, 1615-1621.
M.J. Egger and J.D. Willigan (1984). ‘An Event-history Analysis of Demographic Change in Renais-
sance Florence’, American Statistical Association, 1984 proceedings of the Social Statistics Section:
615-620.
J. Everett (1995). ‘Kleio 5.1.1: A source-oriented data processing system for historical documents.
Technical review’, Computer and the Humanities 29: 307-316.
M. Feeney and S. Ross, (1993). Information technology in humanities scholarship: British achievements,
prospects and barriers. London: British Library and British Academy.
R.W. Fogel and S.L. Engerman (1974). Time on the Cross. Boston ; Toronto: Little, Brown and Com-
pany.
C. Folini (2000). ‘How to bring Barzabal Facin on the screen? A student in search of suitable data-
base architecture’, History and Computing 12 (2): 203-214.
I. Foster and C. Kesselman (1999). The Grid: Blueprint for a New Computing Infrastructure. Los Ange-
les: Morgan Kaufmann.
B. Francis and J. Pritchard (1998). ‘Visualisation of Historical Events using Lexis Pencils’. Derived
from the World Wide Web: h�����������������������������������������������������������������
ttp://www.agocg.ac.uk/reports/visual/casestud/francis/francis.pdf
F.S. Frey and J.M. Reilly (1999). Digital Imaging for Photographic Collections - Foundations for Techni-
cal Standards. Rochester: Image Permanence Institute, Rochester Institute of Technology.
J.B. Friedman (1992). ‘Cluster Analysis and the Manuscript Chronology of William du Stiphel, a
Fourteenth-Century Scribe at Durham’, History and Computing 4 (2): 75-97.
A.H. Galt (1986). ‘Social Class in a Mid-Eighteenth-Century Apulian Town: Indications from the
Castato Onciario’, Ethnohistory 33 (4): 419-447.

116 References
 N. Gershon and S.G. Eick (1997). ‘Guest Editors’ Introduction to Special Issue on Information
Visualization’, IEEE Computer Graphics and Applications 17 (4): 29-31.
T. Gevers and A.W.M. Smeulders (2004). ‘Content-based Image Retrieval: An Overview Survey
on content-based image retrieval’, in: G. Medioni and S.B. Kang, Emerging Topics in Computer
Vision. New York: Prentice Hall,
A. Gilmour-Bryson (1987). ‘Computers and Medieval Historical Texts’, in: P. Denley and D. Hopkin,
History and Computing. Manchester: Manchester University Press, 3-9.
M. Goerke (1994). Coordinates for Historical Maps. St. Katharinen: Halbgraue Reihe.
P. González (1998). Computerization of the Archivo General de Indias: Strategies and Results: Council
on Library and Information Resources [Also full-text availabe in html].
L. Goodman (1959). ‘Some Alternatives to Ecological Correlation’, American Journal of Sociology 64:
610-625.
R.C. Graul and W. Sadée. ‘Evolutionary Relationships Among G Protein-Coupled Receptors Using a
Clustered Database Approach’. Derived from the World Wide Web: ���������������������������
http://itsa.ucsf.edu/~gram/
home/gpcr
D. Greasley and L. Oxley (1998). ‘Comparing British and Americal Economic and Industrial Perfor-
mance 1860-1993: a Time Series Perspective’, Explorations in Economic History 35 (2): 171-195.
M. Greenhalgh (1987). ‘Databases for Art Historians: Problems and Possibilities’, in: P. Denley and
D. Hopkin, History and Computing. Manchester: Manchester University Press, 156-167.
D. Greenstein (1997). ‘Bringing Bacon Home: The Divergent Progress of Computer-Aided Histori-
cal Research in Europe and the United States’, Computers and the Humanities 30: 351-364.
D. Greenstein and L. Burnard (1995). ‘Speaking with One Voice: Encoding Standards and the
Prospect for an Integrated Approach to Computing in History’, Computers and the Humanities
29: 137-148.
D.I. Greenstein (1989). ‘A Source-Oriented Approach to History and Cpmputing: The Relational
Database’, Historical Social Research 14 (51): 9-16.
I.N. Gregory and H.R. Southall (2000). ‘Spatial Frameworks for Historical Censuses – the Great
Britain Historical gis’, in: P.K. Hall, R. McCaa, and G. Thorvaldsen, Handbook of Historical
Microdata for Population Research. Minneapolis: Minnesota Population Center, 319-333.
G.C. Grinstein and M.O. Ward (2002). ‘Introduction to Data Visualization’, in: U. Fayyad, G.C.
Grinstein, and A. Wierse, Information Visualization in Data Mining and Knowledge Discovery. San
Francisco: Morgan Kaufmann, 21-46.
E. Gröller (2001). ‘Insight into Data through Visualization’, in: P. Mutzel, M. Jünger, and S. Leipert,
GD 2001 Heidelberg: Springer-Verlag, 352-366.
F. Guérain-Pace and X. Lesage (2001). ‘Le Systeme Urbain Français. Les Mesures de l’ Inégalité de
Distributions de Type Partetien’, Histoire & Mesure 16 (102): 157-183.
M.P. Gutmann and G. Alter (1993). ‘Family Reconstitution as Event-History Analysis’, in: D. Reher
and R. Schofield, Old and New Methods in Historical Demography. Oxford: Clarendon, 159-177.
M.P. Guttman (2002). ‘Preface’, in: A.K. Knowles, Past Times, Past Place. GIS for history. Redlands,
CA. 2002: esri Press,
D. Haks (1999). ‘Two Examples of the Impact of Computer Technology om Historical Editing: The
Correspondence of William of Orange 1533-1584 and the Resolutions of the States general 1626-
1651’, Journal of the Association for History and Computing 2 (3): pages n.a.
P. Hartland and C. Harvey (1989). ‘Information Engineering and Historical Databases’, in: P. Den-
ley, S. Fogelvik, and C. Harvey, History and Computing II. Manchester, New York: Manchester
University Press, 44-62.
R. Hartmann. ‘Prometheus. Das Verteilte Digitale Bildarchiv für Forschung und Lehre’. Derived
from the World Wide Web: �����������������������������������
http://www.prometheus-bildarchiv.de
C. Harvey (1990). ‘The Nature and Future of Historical Computing’, in: E. Mawdsley, N. Morgan, L.
Richmond, and R. Trainor, History and Computing III. Historians, Computers and Data. Applica-
tions in Research and Teaching. Manchester, New York: Manchester University Press, 205-211.
C. Harvey and E. Green (1994). ‘Record Linkage Algorithms: Efficiency, Selection and Relative
Confidence’, History and Computing 6 (3): 143-152.

117 References
 C. Harvey, E.M. Green and P.J. Corfield (1996). ‘Record Linkage Theory and Practice: an Experi-
ment in the Application of Multiple Pass Linkage Algorithms’, History and Computing 8 (2):
78-89.
C. Harvey and J. Press (1992). ‘Relational Data Analysis: Value, Concepts and Methods’, History and
Computing 4 (2): 98-109.
C. Harvey and J. Press (1993). ‘Structured Query Language and Historical Computing’, History and
Computing 5 (3): 154-168.
C. Harvey and J. Press (1996). Databases in Historical Research. Wiltshire: Antony Rowe.
A. Hayami and S. Kurosu, (2001). ‘Regional Diversity in Demographic and Family Patterns in Prein-
dustrial Japan’, Journal of Japanese Studies 27 (2): 295-321.
E.A. Henderson (2000). ‘When States Implode: the Correlates of Africa’s Civil Wars, 1950-92’,
Studies in Comparative International Development 35 (2): 28-47.
C.F. Hermann and M.G. Hermann (1967). ‘An Attempt to Simulate the Outbreak of World War I’,
American Political Science Review 61 (2): 400-416.
T. Hershberg (1981). ‘The Philadelphia History Project.’ Philadelphia, ny.
E. Higgs (1998). History and Electronic Artefacts. Oxford.
S. Hockey (1999). ‘Is There a Computer in this Class?’ Derived from the World Wide Web: http://
�������
www.iath.virginia.edu/hcs/hockey.html
A. Hodgkin (1987). ‘History and Computing: Implications for Publishing’, in: P. Denley and D.
Hopkin, History and Computing. Manchester: Manchester University Press, 256-261.
D.I. Holmes and R.S. Forsyth (1995). ‘The Federalist Revisited: New Directions in Authorship Attri-
bution’, Literary and Linguistic Computing 10 (2): 111-127.
D.I. Holmes, L.J. Gordon and C. Wilson (1999). ‘A Widow and her Soldier: A Stylometric Analysis
of the ‘Picket Letters’ ’, History and Computing 11 (3): 159-179.
J.J. Hox (2002). Multilevel Analysis: Techniques and Applications. Mahwah, NJ [etc.]: Lawrence Erl-
baum Associates.
E. Hyvönen, S. Saarela and K. Viljanen (2003). ‘Intelligent Image Retrieval and Browsing Using
Semantic Web Techniques - a Case Study.’ Presented at sepia Conference 2003, Helsinki.
N.M. Ide (1995). ‘The tei: History, Goals, and Future’, Computers and the Humanities 29: 5-15.
L. Isaac, L. Christiansen, J. Miller, et al. (1998). ‘Temporally Recursive Regression and Social Histo-
rical Inquiry: an Example of Cross-movement Militancy Spillover’, International Review of Social
History 43 (Supplement 6): 9-32.
A. Janssens (1989). ‘Een ‘Direct-entry Methodology’ voor Negentiende Eeuwse Bevolkingsregisters’,
Cahiers voor Geschiedenis en Informatica 3: 19-41.
M. Jensen (2003). ‘Visualizing Complex Semantic Timelines’. Derived from the World Wide Web:
http://newsblip.com/tr/
M. Katzen (1990). ‘Scholarship and Technology in the Humanities.’ London: Bowker Saur.
K.S.B. Keats-Rohan (1999). ‘Historical Text Archives and Prosopography: the coel Database sys-
tem’, History and Computing 10 (1-3): 57-72.
M. Keiding (1990). ‘Statistical Inference in the Lexis Diagram’, Philosophical Transactions of the Royal
Society of London, Series A (332).
S. Kenna and S. Ross (1995). ‘Networking in the Humanities.’ London, etc: Bowker Saur.
E. Keogh, S. Lonardi and B.Y.-c. Chiu (2002). ‘Finding Surprising Patterns in a Time Series Data-
base in Linear Time and Space.’ Presented at Eighth acm sigkdd International Conference on
Knowledge Discovery and Data Mining.
D.V. Khmelev (2000). ‘Disputed Authorship Resolution through Using Relative Entropy for Markov
Chains of Letters in Human Language Texts’, Journal of Quantitative Linguistics 7 (3): 201-207.
D.V. Khmelev and F.J. Tweedie (2001). ‘Using Markov Chains for Identification of Writers’, Literary
and Linguistic Computing 16 (4): 299-307.
G. King (1997). A Solution to The Ecological Inference Problem: Reconstructing Individual Behavior
From Aggregate Data. Princeton: Princeton University Press.
G. King and L. Zeng (2001). ‘Explaining Rare Events in International Relations’, International Orga-
nization 55 (3): 693-715.

118 References
 S. King (1992). ‘Record Linkage in a Protoindustrial Community’, History and Computing 4 (1): 27-
33.
S. King (1994). ‘Multiple-source Record Linkage in a Rural Industrial Community, 1680-1820’,
History and Computing 6 (3): 133-142.
W. Kintsch (2003). ‘On the notion of theme and topic in psychological peocess models of text
comprehension’, in: W.v. Peer, Parsing for the theme. A computer based approach. Amsterdam,
Philadelphia: John Benjamins Publishing, 158-170.
M.G. Kirschenbaum (2002). ‘Editor’s Introduction: Image-based Humanities Computing’, Compu-
ter and the Humanities 36: 3-6.
E. Klijn and Y.D. Lusenet (2002). In the Picture. Preservation and Digitisation of European Photogra-
phic Collections. Amsterdam: European Commission on Preservation and Access.
E. Klijn and L. Sesink (2003). ‘sepia Working Group on Descriptive Models and Tools.’ Presented at
sepia Conference 2003, Helsinki.
A.K. Knowles (2002). ‘Introducing Historical gis’, in: A.K. Knowles, Past Times, Past Place. GIS for
history. Redlands, ca.: esri Press.
M. Kobialka (2002). ‘Can there be such a thing as a postmodern archive?’, in: J. Frow, The New
Information Order and the Future of the Archive. Institute for Advanced Studies in the Humanities
- The University of Edinburgh.
J. Kok (1997). ‘Youth Labour Migration and its Family Setting, the Netherlands 1850-1940’, History
of the Family. An International Quarterly 2: 507-526.
J. Komlos and M. Artzrouni (1994). ‘Ein Simulationsmodell der Industriellen Revolution’, Viertelja-
hrschrift für Sozial- und Wirtschaftsgeschichte 81 (3): 324-338.
I. Koprinska and S. Carrato (2001). ‘Temporal Video Segmentation: a Survey’, Signal Processing:
Image Communication 16 (477-500).
J.M. Kousser (2001). ‘Ecological Inference from Goodman to King’, Historical Methods 34 (3): 101-
126.
I. Kreft and J.d. Leeuw (1998). Introducing Multilevel Modelling. London: Sage.
G. Kristiansson (2000). ‘Building a National Topographic Database’. Derived from the World Wide
Web: �������������������������������������������������������������������
http://www.geog.port.ac.uk/hist-bound/project_rep/NAD_more_info.htm
H. Kropac (1997). ‘Electronical Documentation vs. Scholarly Editing?’ Derived from the World Wide
Web: ������������������������������������������������
http://www.hist.uib.no/acohist/kropac/kropac.htm
G.P. Landow (1996?). ‘Hypertext, Scholarly Annotation, and the Electronic Edition’.
E. Lecolinet, L. Robert and F. Role (2002). ‘Text-image Coupling for Editing Literary Sources’, Com-
puters and the Humanities 36: 49-73.
Q. Li and R. Reuveny (2003). ‘Economic Globalization and Democracy: an Empirical Analysis’,
British Journal of Political Science 33 (1): 29-54.
G. Lind (1994). ‘Data Structures for Computer Prosopography’, in: H.J. Marker and K. Pagh, Yester-
day. Proceedings from the 6th international conference Association of History and Computing, Odense
1991 Odense: Odense University Press, 77-82.
C. Litzenberger (1995). ‘Computer-based Analysis of Early-modern English Wills’, History and
Computing 7 (3): 143-151.
J. Ljungberg (2002). ‘About the Role of Education in Swedish Economic Growth, 1867-1995’, Histo-
rical Social Research 27 (4): 125-139.
M. Louwerse (2003). ‘Computational retrieval of themes’, in: W.v. Peer, Parsing for the theme. A com-
puter based approach. Amsterdam, Philadelphia: John Benjamins Publishing, 189-212.
M. Louwerse and W.v. Peer (2002). ‘Thematics. Interdisciplinary Studies.’ in Converging Evidence in
Language and Communication Research. Amsterdam, Philadelphia: John Benjamins.
K.A. Lynch and J.B. Greenhouse (1994). ‘Risk Factors for Infant Mortality in Nineteenth-century
Sweden’, Population Studies 48 (1): 117-135.
C. Martindale and D. McKenzie (1995). ‘On the Utility of Content Analysis in Author Attribution:
The Federalist’, Computers and the Humanities 29 (4): 259-270.
E. Mawdsley, N. Morgan, L. Richmond, et al. (1990). ‘History and Computing III. Historians, Com-
puters and Data. Applications in Research and Teaching.’ Manchester, New York: Manchester
University Press.

119 References
 H.J. McCammon (1999). ‘Using Event History Analysis in Historical Research: with Illustrations
from a Study of the Passage of Women’s Protective Legislation’, in: L.L. Griffin and M. van der
Linden, New Methods for Social History (International Review of Social History, Supplement 6).
Cambridge: cup, 33-56.
W. McCarty (1999). ‘Humanities Computing as Interdiscipline’. Derived from the World Wide
Web: �����������������������������������������������������
http://www.kcl.ac.uk/humanities/cch/wlm/essays/inter/.
W. McCarty (2001). ‘Looking Through an Unknown, Remembered Gate: Millennial Speculations on
Humanities Computing’. Derived from the World Wide Web: http://www.cch.kcl.ac.uk/legacy/
staff/wlm/essays/victoria/
B.H. McCormick, T.A. Defanti and M.D. Brown (1987). ‘Visualization in Scientific Computing - A
Synopsis’, Computer Graphics & Application 7: 61-70.
L.J. McCrank (2002). Historical Information Science. An Emerging Discipline. Medford, New Jersey:
Information Today.
J. McGann (1991). ‘What is Critical Editing?’ TEXT: Transactions of the Society for Textual Scholarship
5: 15-29.
J. McGann (1992). A Critique of Modern Textual Criticism. Charlottesville: up of Virginia.
J. McGann (1995). ‘The Rationale of HyperText’. Derived from the World Wide Web: http://www.
iath.virginia.edu/public/jjm2f/rationale.html
J. McGann (2002). ‘Dialogue and Interpretation at the Interface of Man and Machine. Reflections
on Textuality and a Proposal for an Experiment in Machine Reading’, Computer and the Humani-
ties 36: 95-107.
T. McIntosh (2001). ‘Urban Demographic Stagnation in early Modern Germany: a Simulation’,
Journal of Interdisciplinary History 31 (4): 581-612.
J.C. Meister (2003). ‘Parsing for the theme. A computer based approach’, in: W.v. Peer, Thematics.
Interdiciplinary Studies. Amsterdam, Philadelphia: John Benjamins Publishing, 407-431.
T. Merriam (2002). ‘Linguistic Computing in the Shadow of Postmodernism’, Literary and Linguistic
Computing 17 (2): 181-192.
R. Metz (1988a). ‘Ansätze, Begriffe und Verfahren der Analyse Ökonomischer Zeitreihen’, Histori-
cal Social Research 13 (3): 23-103.
R. Metz (1988b). ‘Erkenntnisziele Zeitreihenanalytischer Forschung’, Historical Social Research 13
(3): 6-22.
R. Metz (1993). ‘Probleme der Statistischen Analyse langer historischer Zeitreihen’, Vierteljahrschrift
für Sozial- und Wirtschaftsgeschichte 80 (4): 457-486.
H. Mielants and E. Mielants (1997). ‘The Importance of Simulation as a Mode of Analysis: Theoreti-
cal and Practical Implicaitons and Considerations’, Belgisch Tijdschrift voor Nieuwste Geschiedenis
27 (3-4): 293-322.
C. Monroy, R. Kochumman, R. Furuta, et al. (2002a). ‘ Interactive Timeline Viewer (ItLv): A Tool
to Visualize Variants Among Documents’, in: K. Börner and C. Chen, Visual Interfaces to Digital
Libraries. Lecture Notes in Computer Science. Heidelberg: Springer-Verlag, 39-49.
C. Monroy, R. Kochumman, R. Furuta, et al. (2002b). ‘Visualization of Variants in Textual Colla-
tions to Analyze the Evolution of Literary Works in the Cervantes Project.’ ECDL 2002: 638-653.
R.J. Morris (1995). ‘Death, Property and the Computer - Strategies for the Analysis of English Wills
in the First Half of the Nineteenth Century’, in: P. Teibenbacher, The Art of communication.
Proceedings of the Eight International Conference of the Association for History and Computing, Graz,
Austria, August 24-27, 1993 Graz: Akademische Druck- iund Verlaganstalt, 164-178.
F. Mosteller and D.L. Wallace (1964). Applied Bayesian and Classical Inference: The Case of the Federa-
list Papers. Reading: Addison-Wesley.
V. Mueller-Benedict (2000). ‘Confirming Long Waves in Time Series of German Student Popula-
tions 1830-1990 Using Filter Techniques and Spectral Analysis’, Historical Social Research 25
(3-4): 36-56.
C. Mullings (1996). ‘New Technologies for the Humanities.’ London: Bowker Saur.
J.W. Nibbering and J. DeGraaff (1998). ‘Simulating the Past: Reconstructing Historical Land Use
and Modeling Hydrological Trends in a Watershed Area in Java’, Environment and History 4 (3):
251-278.

120 References
 H. Obinger and U. Wagschal (2001). ‘Families of Nations and Public Policy’, West European Politics
24 (1).
B.S. Okun (1995). ‘Distinguishing Stopping Behavior from Spacing Behavior with Indirect
Methods’, Historical Methods 28 (2): 85-96.
J. Oldervoll (1992). ‘Wincens, a Census System for the Nineties?’ in: Eden or Babylon? On Future
Software for Highly Structured Historical Sources. St. Katharinen: Max-Planck-Insitut für Ges-
chichte, Göttingen, 37-52.
J. Oldervoll (1994). ‘Why don’ t We All use dBase?’ in: H.J. Marker and K. Pagh, Yesterday. Pro-
ceedings from the 6th international conference Association of History and Computing, Odense 1991
Odense: Odense University Press, 135-139.
B. Opheim (2000). ‘Political Networks and Factions: Online Prosopography of Medieval Scandina-
vian Sagas’, History and Computing 12 (1): 43-57.
S.E. Ostrow (1998). ‘Digitizing Historical Pictorial Collections for the Internet’. Derived from the
World Wide Web: http://www.clir.org/PUBS/reports/ostrow/pub71.html
T. Ott and F. Swiaczny (2001). Time-integrative Geographic Information Systems: Management and
Analysis of Spatio-temporal Data. Heidelberg: Springer Verlag.
W. Ott (2002). ‘Textual Criticism / Scholarly Editing’. Derived from the World Wide Web: http://
www.allc.org/reports/map/textual.html
M. Overton (1995). ‘A Computer Management System fro Probate Inventories’, History and Compu-
ting 7 (3): 135-142.
A.C. Pacek and B. Radcliff (2003). ‘Voter Participation and Party-Group Fortunes in European
Parliament Elections, 1979-1999: a Cross-national Analysis’, Political Research Quarterly 56 (1):
91-95.
M.E. Palmquist, K.M. Carley and T.A. Dale (1997). ‘Applications of Computer-Aided Text Analysis:
Analyzing Literary and Nonliterary Texts’, in: C.W. Roberts, Text Analysis for the Social Sciences.
New Jersey: Erlbaum, 171-189.
D. Parker (2001). ‘The World of Dante: a Hypermedia Archive for the Study of the Inferno’, Literary
and Linguistic Computing 16 (3): 287-297.
C. Plaisant, B. Milash, A. Rose, et al. (1996). ‘LifeLines: Visualizing Personal Histories.’ Presented
at chi’ 96, Vancouver, bc.
A. Prescott (1997). ‘The Electronic Beowulf and Digital Restoration’, Literary and Linguistic Compu-
ting 12: 185-195.
C.T.D. Price, F.G. O’ Brien, B.P. Shelton, et al. (1999). ‘Effects of Salicylate and Related Compounds
on Fusidic Acid mics in Staphylococcus Aureus’, Journal of Antimicrobial Chemotherapy 44: 57-
64.
S.A. Raaijmakers (1999). ‘Woordsoorttoekenning met Markov-modellen’, in: Jaarboek van de
Stichting Instituut voor Nederlandse Lexicologie, overzicht van het jaar 1998 82-90.
L.E. Raffalovich (1999). ‘Growth and Distribution: Evidence from a Variable-parameter Cross-natio-
nal Time-series Analysis’, Social Forces 78 (2): 415-432.
L.E. Raffalovich and D. Knoke (1983). ‘Quantitative Methods for the Analysis of Historical Change’,
Historical Methods 16 (4): 149-154.
D. Reher and R. Schofield (1993). ‘Old and New Methods in Historical Demography.’ Oxford: Cla-
rendon.
A. Reid (2001). ‘Neonatal Mortality and Stillbirths in early Twentieth Century Derbyshire, England’,
Population Studies 55 (3): 213-232.
K.F.J. Reinders (2001). Feature-based Visualization of Time-dependent Data. Delft: Diss. tu Delft.
A. Renear, E. Mylonas and D. Durand (1993). ‘Refining our Notion of What Text Really Is: The
Problem of Overlapping Hierarchies’. Derived from the World Wide Web: http://www.stg.brown.
edu/resources/stg/monographs/ohco.html
A. Ritschl (1998). ‘Reparation Transfers, The Borchardt Hypothesis and the Great Depression in
Germany, 1929-32: a Guided Tour for Hard-headed Keynesians’, European Review of Economic
History 2 (1): 49-72.
P.M.W. Robinson, W. Gabler and H. Walter (2000). ‘Making Texts for the Next Century.’

121 References
 G. Rockwell (1999). ‘Is Humanities Computing an Academic Discipline?’ Derived from the World
Wide Web: http://www.iath.virginia.edu/hcs/rockwell.html.
J. Rudman, D.I. Holmes, F.J. Tweedie, et al. (1997). ‘The State of Authorship Attribution Studies: (1)
The History and the Scope; (2) The Problems – Towards Credibility and Validity.’ Derived from
the World Wide Web: http://www.cs.queensu.ca/achallc97/papers/s004.html
���������������������������������������������������
R. Ruusalepp (2000). ‘Multiple-Source Nominal Record Linkage: An Interactive Approach with
Kleiw’, in: P. Doorn, Data Modelling Modelling History. Proceedings of the XI International
Conference of the Association for History and Computing, Moscow, August 1996 Moscow: Moscow
University Press, 320-332.
J.A. Rydberg-Cox, R.F. Chavez, D.A. Smith, et al. (2002). ‘Knowledge Management in the Perseus
Digital Library’, Ariadne 25.
J. Schellekens (1995). ‘Illegitimate Fertility Decline in England, 1851-1911’, Journal of Family History
20 (4): 365-377.
C. Schonhardt-Bailey (1998). ‘Parties and Interests in the ‘Marriage of Iron and Rye‘’, British Journal
of Political Science 28 (2): 291-332.
R. Schor (1996). Histoire de l’ immigration en France de la fin du XIXe siècle à nos jours. Paris: Armand
Colin.
A.T. Schreiber, B. Dubbeldam, J. Wielemaker, et al. (2001). ‘Ontology-based Photo Annotation’,
IEEE Intelligent Systems 16 (66-74).
S. Schreibman (2002). ‘Computer-mediated Texts and Textuality: Theory and Practice’, Computers
and the Humanities 36 (3): 283-293.
A. Schuurman and G. Pastoor (1995). ‘From Probate Inventories to a Data Set for the History of the
Consumer Society’, History and Computing 7 (3): 126-134.
S. Scott, S.R. Duncan and C.J. Duncan (1998). ‘The Origins, Interactions and Causes of the Cycles
in Grain Prices in England, 1450-1812’, Agricultural History Review 46 (1): 1-14.
D.A. Smith (2002). ‘Detecting Events with Date and Place Information in Unstructured Text’. Deri-
ved from the World Wide Web: www.perseus.tufts.edu/Articles/datestat.pdf
�������������������������������������������
D.A. Smith, J. Rydberg-Cox and G.R. Crane (2000). ‘The Perseus Project: a Digital Library for the
Humanities’, Literary and Linguistic Computing 15 (1): 15-25.
B.K. Song (2002). ‘Parish typology and the Operation of the Poor Laws in early Nineteenth-Century
Oxfordshire’, Agricultural History Review 50 (2): 203-224.
S.J. South (1999). ‘Historical Changes and Life Course Variation in the Determinants of Premarital
Childbearing’, Journal of Marriage and the Family 61 (3): 752-763.
W.A. Speck (1994). ‘History and Computing: Some Reflections on the Past Decade’, History and
Computing 6 (1): 28-32.
R. Spree (1997). ‘Klassen- und Schichtbildung im Medium der Privaten Konsums: Vom Späten
Kaiserreich in die Weimarer Republik’, Historical Social Research 22 (2): 29-80.
D.J. Staley (1998). ‘Designing and Displaying Historical Information in the Electronic Age’, Journal
of the Association for History and Computing 1 (1).
D.J. Staley (2003). Computers, Visualization, and History: How New Technology Will Transform Our
Understanding of the Past. Armonk, N.Y.: M.E. Sharpe.
L.L. Stewart (2003). ‘Charles Brockden Brown: Quantitative Analysis and Literary Interpretation’,
Literary and Linguistic Computing 18 (2): 129-138.
W. Stier (1989). ‘Basic Concepts and new Methods of Time Series Analysis in Historical Social
Research’, Historical Social Research 14 (1): 3-24.
M. Thaller (1980). ‘Automation on Parnassus. clio - A databank oriented system for historians’,
Historical Social Research 15: 40-65.
M. Thaller (1987). ‘Methods and Techniques of Historical Computation’, in: P. Denley and D. Hop-
kin, History and Computing. Manchester: Manchester University Press, 147-156.
M. Thaller (1989). ‘The Need of a Theory of Historical Computing’, in: P. Denley, S. Fogelvik, and C.
Harvey, History and Computing II. Manchester, New York: Manchester University Press, 2-11.
M. Thaller (1993a). ‘Historical Information Science: Is There such a Thing? New Comments on an
old Idea’, in: T. Orlandi, Seminario Discipline Umanistiche e Informatica. Il Problema dell’ Integra-
zione. Roma.

122 References
 M. Thaller (1993b). ‘What is ‘Source Oriented Data Processing?’ ; What is a ‘Historical Information
Science?’ ’, in: L.I. Borodkin and W. Levermann, Istoriia i comp’ iuter. Novye informationnye tekh-
nologii v istotricheskikh issledovanii akh i obrazovanii. St. Katharinen, 5-18.
M. Thaller (1996). ‘Digital Manuscripts: Editions v. Archives’. Derived from the World Wide Web:
http://gandalf.aksis.uib.no/allc-ach96/Panels/Thaller/thaller1.html
S. Thernstrom (1973). The other Bostonians : Poverty and Progress in the American Metropolis, 1880-
1970 Cambridge, Mass.: Harvard University Press.
P. Tilley and C. French (1997). ‘Record Linkage of Nineteenth-century Census Returns. Automatic
or Computer Aided?’ History and Computing 9 (1-3): 122-133.
E. Tufte (1983). The Visual Display of Quantitative Information. Cheshire: Graphics Press.
F.J. Tweedie, S. Singh and D.I. Holmes (1996). ‘Neural Network Applications in Stylometry: The
Federalist Papers’, Computers and the Humanities 30: 1-10.
E. Urbina, R.K. Furuta, A. Goenka, et al. (2002). ‘Critical Editing in the Digital Age: Information
and Humanities Research’, in: J. Frow, The New Information Order and the Future of the Archive.
Institute for Advanced Studies in the Humanities - The University of Edinburgh,
E. Vanhoutte (1999). ‘Where is the editor? Resistance in the creation of an electronic critical edi-
tion’, Human IT. Tidskrift för studier av IT ur ett humanvetenskapligt perspektiv 1.
S.v.d. Velden and P.K. Doorn (2001). ‘The Striking Netherlands: Time Series Analysis and Models
of socio-economic Development and Labour Disputes, 1850-1995’, Historical Social Research 26:
222-243.
J.E. Vetter, J.R. Gonzalez and M.P. Gutmann (1992). ‘Computer-Assisted Record Linkage Using a
Relational Database System’, History and Computing 4 (1): 34-51.
J. Viscomi (2002). ‘Digital Facsimiles: Reading the William Blake Archive’, Computers and the Hu-
manities 36: 27-48.
C.C. Webb and V.W. Hemingway (1995). ‘Improving Access: A Proposal to Create a Database for
Probate Records at Borthwick Institute’, History and Computing 7 (3): 152-155.
G. Welling (1993). ‘A Strategy for Intelligent Input Programs for Structured Data’, History and
Computing 5 (1): 35-41.
G. Welling (1998). The Prize of Neutrality. Trade relations between Amsterdam and North America 1771-
1817 A study in computational history. Hilversum: Verloren.
G.M. Welling (1992). ‘Intelligent Large-scale Historical Direct-data-entry Programming’, in: J.
Smets, Histoire et Informatique. Actes du Congrès. Ve Congrès ‘History & Computing‘ 4-7 Septembre
1990 à Montpellier. Montpellier, 563-571.
J.J.v. Wijk and E.v. Selow (1999). ‘Cluster and Calendar-based Visualization of Time Series Data.’
Presented at Proceedings 1999 ieee Symposium on Information Visualization (InfoVis’ 99),
October 25-26, 1999.
I. Winchester (1970). ‘The Linkage of Historical Records by Man and Computer’, Journal of Interdis-
ciplinary History 1: 107-124.
R.L. Woods (1987). ‘Skills for Historians: Getting Something Done with a Computer’, in: P. Denley
and D. Hopkin, History and Computing. Manchester: Manchester University Press, 205-210.
M. Woollard (1999). ‘Introduction: What is History and Computing? An Introduction to a Problem’,
History and Computing 11 (1-2): 1-8.
E.A. Wrigley (1973). Identifying People in the Past. London.
K. Yamaguchi (1991). Event History Analysis. Beverly Hills / New York: Sage.
D. Zeldenrust (2003). ‘Picture the Past, the Use of Documentary Photographic Images in Historical
Research.’ Presented at sepia Conference 2003, Helsinki.
Z. Zhao (1996). ‘The Demographic Transition in Victorian England and Changes in English
Kinship Networks’, Continuity and Change 11 (2): 243-272.

123 References
124 References
 Acknowledgements

This study has been made possible thanks to the generous co-operation of the Insti-
tute of Information and Computing Sciences (Instituut voor Informatica en Infor-
matiekunde) of the University of Utrecht, and the Faculteit Letteren of the Radboud
University Nijmegen, who allowed Leen Breure and Onno Boonstra to work for the
niwi project on past, present and future of historical information science.
Parts of this study have been discussed at various occasions with a wide range of
specialists in the field. We would like to thank all those that were so kind to provide
us with new information or with critical remarks during these sessions. In particular,
we would like to express our gratitude to:
− the participants of the workshop ‘Past, present & future of historical information
science’, 9 August 2003, International ahc Conference, Tromsø, Norway;
− the participants of the workshop ‘Historical information science’, 24 October 2003,
Amsterdam;
− the participants at the vgi symposium ‘Past, present & future of historical informa-
tion science’, 17 November 2003, Amsterdam;
− dr. Peter Boot, Constantijn Huygens Instituut, Den Haag;
− prof. dr. Hans Bennis and drs. Edwin Brinkhuis, Meertens Instuut, Amsterdam;
− dr. Donald Haks and dr. Rik Hoekstra, Instituut voor Nederlandse Geschiedenis,
Den Haag;
− dr. Henk Wals, International Institute for Social History, Amsterdam;
− dr. Hans Voorbij and prof. dr. Mark Overmars, Instituut Informatica en Informa-
tiekunde, Universiteit Utrecht;
− dr. Karina van Dalen, Afdeling Neerlandistiek, Nederlands Instituur voor Weten-
schappelijke Informatiediensten, Amsterdam;
− the National Archives, Den Haag;
− dr. Truus Kruyt, Instituut voor Nederlandse Lexicologie, Den Haag;
− dr. Martin Bossenbroek, drs. Marco de Niet, Mr. Margariet Moelands, Koninklijke
Bibliotheek, Den Haag;
− prof. dr. Eep Talstra, Faculteit Theologie, Vrije Universiteit Amsterdam;
− dr. George Welling, Vakgroep Alfa-informatica, Rijksuniversiteit Groningen;
− dr. Joost Kircz, kra-Publishing Research, Amsterdam
− dr. Seamus Ross, Humanities Computing and Information Management / hatii
(Humanities Advanced Technology and Information Institute), University of
Glasgow, uk

The first draft of this report was discussed on an international meeting, held in ‘De
Sparrenhorst’, Nunspeet, 13-15 February 2004, by the following experts:
− prof. dr. Manfred Thaller, Historisch-Kulturwissenschaftliche Informationsverar-
beitung, Universität zu Köln, Germany;
− prof. Gunnar Thorvaldsen, Norwegian Historical Data Centre, Faculty of Social
Sciences, University of Tromsø, Norway;

125 Acknowledgements
 − prof. Leonid Borodkin, Historical Informatics Lab, Moscow State University, Rus-
sia;
− dr. Matthew Woollard, Arts and Humanities Data Service, History, University of
Essex, uk;
− dr. Jan Oldervoll, Historisk Institutt, University of Bergen, Norway;
− prof. dr. Henk Koppelaar, Department of Information Technology and Systems,
Technical University Delft;
− prof. dr. Martin Kersten, Instituut voor Wiskunde en Informatica, Amsterdam
University;
− prof. dr. Jaap van den Herik, Department of Computer Science, Universiteit Maas-
tricht;
− prof. Bob Morris, Economic and Social History, University of Edinburgh, uk;
− prof. Dr. Ingo H. Kropač, Institut für Informationsverarbeitung in den Geisteswis-
senschaften, Karl-Franzens-Universität Graz, Austria;
− Alan Morrison, Oxford Text Archive / ahds literature, language and linguistics,
Oxford University, uk;

We are especially grateful to Matthew Woollard, who not only made many wise and
useful comments on the first draft of this text, but who also checked and corrected
our broken English.

126 Acknowledgements
127 Acknowledgements
128 Acknowledgements