INFORMATION SCIENCE AND KNOWLEDGE MANAGEMENT

Classifying Science
Phenomena, Data, Theory, Method, Practice

By
Rick Szostak
CLASSIFYING SCIENCE
 Information Science and Knowledge Management
Volume 7

Editor-in-Chief:
J. Mackenzie Owen, University of Amsterdam, Amsterdam

Editorial Board:
M. Bates, University of California, Los Angeles
P. Bruza, The University of Queensland, Brisbane
R. Capurro, Hochschule der Medien, Stuttgart University of Applied Sciences, Stuttgart
E. Davenport, Napier University, Edinburgh
R. Day, Wayne State University, Detroit
M. Hedstrom, University of Michigan, Ann Arbor
A.M. Paci, Istituto di Studi Sulla Ricerca e Documentazione Scientifica, Roma
C. Tenopir, University of Tennessee, Knoxville
M. Thelwall, University of Wolverhampton, Wolverhampton

The titles published in this series are listed at the end of this volume.
CLASSIFYING SCIENCE
Phenomena, Data, Theory, Method, Practice

by

RICK SZOSTAK
University of Alberta,
Edmonton, Alberta, Canada
A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 1-4020-3094-0 (HB)

ISBN 1-4020-3095-9 (e-book)

Published by Springer,
P.O. Box 17, 3300 AA Dordrecht, The Netherlands.

Sold and distributed in North, Central and South America

by Springer,
101 Philip Drive, Norwell, MA 02061, U.S.A.

In all other countries, sold and distributed

by Springer,
P.O. Box 322, 3300 AH Dordrecht, The Netherlands.

Printed on acid-free paper

All Rights Reserved

© 2004 Springer
No part of this work may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, microfilming, recording
or otherwise, without written permission from the Publisher, with the exception
of any material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work.

Printed in the Netherlands.

CONTENTS

List of Tables vii

Preface ix

Chapter 1: Classifying Science 1

1.1. A Simple Classificatory Guideline 3
1.2. The First “Cut” (and Plan of Work) 5
1.3. Some Preliminaries 9

Chapter 2: Classifying Phenomena and Data 23

2.1. Classifying Phenomena 23
2.2. Classifying Data 45

Chapter 3: Classifying Theory 51

3.1. Typology of Theory 55
3.2. What Is a Theory? 74
3.3. Evaluating Theories 78
3.4. Types of Theory and the Five Types
of Causation 80
3.5. Classifying Individual Theories 82
3.6. Advantages of a Typology of Theory 95

Chapter 4: Classifying Method 99

4.1. Classifying Methods 101
4.2. Typology of Strengths and Weaknesses
of Methods 103
4.3. Qualitative Versus Quantitative Analysis
Revisited 109
4.4. Evaluating Methods 113
4.5. Classifying Particular Methods Within
The Typology 116
4.6. Advantages of a Typology of Methods 144
vi Contents

Chapter 5: Classifying Practice 155

5.1. Errors and Biases in Science 158
5.2. Typology of (Critiques of) Scientific
Practice 161
5.3. Utilizing This Classification 192
5.4. The Five Types of Ethical Analysis 194

Chapter 6: Drawing Connections Across

These Classifications 199
6.1. Theory and Method 199
6.2. Theory (Method) and Phenomena (Data) 203
6.3. Better Paradigms 208
6.4. Critiques of Scientific Practice: Are They
Correlated with Other Classifications? 213

Chapter 7: Classifying Scientific Documents 217

7.1. Faceted or Enumerative? 219
7.2. Classifying By Phenomena Studied 221
7.3. Classifying By Theory Used 225
7.4. Classifying By Method Used 227
7.5 Links Among Subjects 228
7.6. Type of Work, Language, and More 229
7.7. Critiques of Scientific Practice 230
7.8. Classifying Philosophy 231
7.9. Evaluating the System 232

Chapter 8: Concluding Remarks 239

8.1. The Classifications 239
8.2. Advantages of These Various Classifications 241
8.3. Drawing Connections Across Classifications 245
8.4. Golden Mean Arguments 247
8.5. Why Should Science Be Believed? 249
8.6. How Can Science Be Improved? 250
8.7. How Should Science Be Taught? 259

References 269
Index 279
List of Tables vii

LIST OF TABLES

Table 2.1: The Phenomena of Interest to

Human Scientists 27

Table 3.1: Typology of Selected Theories 94

Table 4.1: Typology of Strengths and Limitations

Of Methods 138
PREFACE

I have always been interdisciplinary in orientation, though I only became

self-consciously interdisciplinary, and thus concerned with the theory and
practice of interdisciplinarity, after a decade of scholarship as an economic
historian. In my research in economic history, I was always looking outward
for the role that culture or politics or natural science understanding might
have played in the economic transformations that I studied. I was also very
suspicious of the way that one method and one theory dominated the
Economics profession; my own research indicated that the discipline’s
collective understanding of economic development would benefit from an
openness to multiple theories and methods.
I began to wonder if there was not some way to map the phenomena
studied by scientists, so that both scholars and students could readily place
any subject they studied in context, and see what other phenomena it might
be related to. My initial motivation was that, without such a map, I often
found as an economic historian that I was unsure of what phenomena to look
for, much less where in the scholarly literature to find analysis of these. I
soon became aware that such a map would also help students in the difficult
task of drawing connections across different courses. I have, over the last
several years, written a book and several articles along these lines; this
material is summarized and built upon in the second chapter of this book.
As I developed my classifications of the phenomena studied by scientists
– and note that the word ‘science’ is used in this book so that it encompasses
the natural and social sciences, plus the humanities; the logic of such a
definition is explored in chapter 5 – I wondered if there might be some way
to classify scientific methods and theories as well. The same goals would be
served by such classifications: to alert scholars and students to the range of
possibilities, to allow them to place their present studies within a larger
x Preface

vision of the scientific enterprise, and to encourage interdisciplinary

exploration. The ‘eureka’ moment came with the discovery that asking the
5W questions – who, what, where, when, and why – allowed a simple yet
powerful typology of types of scientific theory (in chapter 3). These five
questions served moreover to highlight the major strengths and weaknesses
of not only different theory types but also (in chapter 4) the dozen scientific
methods. These classifications in turn allow a concise and powerful
argument to be made that scientists and scientific communities should utilize
multiple theories and methods.
As Allyn Young once said, to a boy with a new hammer the whole world
looks like a nail. Given these successes, it was only natural to turn the 5W
questions upon other aspects of science. Indeed, asked of science itself the
5W questions suggest that the scientific enterprise is best understood in
terms of phenomena, data, theory, method, and practice (in chapter 1). The
5W questions are used in chapter 2 to provide guidance in scientific
treatment of phenomena and data. Then in chapter 5 the 5W questions are
employed once more to provide a classification of critiques of scientific
practice or, in other words, the possible errors and biases that can afflict the
scientific project. With these various classifications in place, the obvious
next step was to discuss what connections might be drawn across different
classifications: what is and should be the connection between particular
types of phenomena, data, theory, method, and practice? This task is taken
up in chapter 6.
The classifications were developed for their own inherent utility. Their
advantages individually and collectively are summarized in chapter 8. Yet
taken together the classifications underpin a vision of a superior scientific
enterprise, where scientific communities are encouraged to draw on a wider
range of phenomena, data, theory, and method; where self-conscious
awareness of biases and errors allows these to be substantially reduced in
effect; and where interdisciplinary scholarship integrates the varied
researches of specialized scholars into a coherent whole. The unity of
science has often been falsely pursued in terms of one grand theory or
method; this book suggests that science is instead unified around a finite and
inter-related set of phenomena that can be treated with a finite (though large)
and complementary set of data, theory types, and method. This vision of
how science should proceed is developed throughout the book and
summarized in chapter 8.
I have long been an advocate of clarity in scientific writing. Rhetorical
flourish has its place, but should not be pursued at the expense of
understanding. (This point will be returned to in chapter 3.) Scientific
advance depends on individual scientists being clear about the method(s)
they use, the phenomena they study, the data they use to proxy for these
Preface xi

phenomena, the theory(ies) they advocate and/or test, and the steps they have
taken to reduce possibilities of error. Clarity is more carefully pursued in
some scientific fields than others. The classifications developed in this book
both encourage and facilitate the careful specification of these key elements
of a research project. Moreover they allow this to be done in the context of a
common vocabulary. At the moment, scholarly communication, even within
disciplines but especially beyond, is rendered problematic by the fact that
different scholars attach quite different meanings to the same word. A
common vocabulary should encourage interdisciplinary communication, and
also communication to the wider public.
At present, and even moreso with the broader vision of science advocated
in this book, scientists face real difficulties in identifying relevant literature
from outside their own community of scholars. Library catalogues are
organized around disciplines rather than phenomena. Moreover while general
works on particular theories or methods can be found (though even here
there are problems in that these may be known by different names in
different fields), it is virtually impossible to search for works that apply a
particular theory or method. Chapter 7 shows how the classifications
developed in this book could form the basis of a superior system of scientific
document classification. While such a project should not be entertained
lightly, the present efforts to develop classifications more suited to the
electronic age provides an opportunity for re-engineering systems of
classification.
What is interdisciplinarity? I have defined this elsewhere as an openness
to the phenomena, theories, and/or methods of multiple disciplines. In this
book interdisciplinary scholarship is contrasted with specialized scholarship,
where particular communities of scholars employ a subset of scientific
theories and methods to analyze a particular subset of phenomena. It is
argued that both types of research are essential to the scientific enterprise.
Specialized communities of scholars can accomplish much because of their
shared expertise, vocabulary, and sense of purpose. On their own, though,
they will produce only a congeries of insights. Even if specialized
researchers become more theoretically and methodologically open, as they
should, interdisciplinary scholars are necessary to tie these pieces into a
coherent whole. Moreover, major advances in scientific understanding
generally come from the combination of insights from different fields.
Interdisciplinary scholars also have an important role to play in pointing out
when specialized scholars would benefit from a wider theoretical or
methodological compass. While much of this book will be couched in
general terms applicable to all scientists, there will be numerous points at
which specific advice is given to specialized or interdisciplinary researchers;
this advice is summarized in chapter 8.
xii Preface

It is no accident that this book advocates theoretical and methodological

flexibility, and also sees a key role for interdisciplinary research. The
different theories and methods that dominate different disciplines serve as
perhaps the major barrier to cross-disciplinary conversation. They also
provide a perverse rationale for disciplinarians to disparage the theoretical or
methodological impurity of interdisciplinary research. I suspect that the
battle for an interdisciplinary perspective and the battle for theoretical and
methodological flexibility are closely related; indeed the latter can be seen as
a component of the former. Once specialized communities of scholars come
to recognize the value of looking at a wider range of theory and method, they
must also see the value of interdisciplinary scholarship.
This book, then, is the work of a scholar committed to both
interdisciplinary scholarship and to theoretical and methodological flexibility
(though not without limit; one major purpose of the book is to detail the
strengths and weaknesses of different theory types and methods in different
applications). It shows how this vision of a scholarly enterprise can feasibly
be pursued, and the advantages of doing so. The book also relies upon and
expresses another authorial commitment: to openmindedness itself. This is a
subject I have addressed in previous work, where I have urged greater
openmindedness upon scholars and citizens alike. In this book,
openmindedness is urged with respect to a host of extreme philosophical
arguments – such as that science is impossible, or that it is perfect – that
would bring this work into question. This is done in the form of a Golden
Mean argument, developed in chapter 1. Time after time, it is shown that
only a slight backing away from an extreme perspective is required to
support the present project. These specific Golden Mean arguments are
summarized in chapter 8. It is hoped that this book serves to buttress a plea
for the pursuit of the Golden Mean with respect to these and other issues.
The majority of scientists just ‘do’ science, and thus spend little time
worrying about methodological issues. As long as the scientific community
of which they are part is pursuing a progressive research program, these
scientists can be productive. This book will suggest that a self-awareness of
the biases and errors that can afflict scholarship, and a sense of the broad
outlines of the scientific enterprise, will make such scientists even more
productive. All scientists would benefit from a familiarity with the strengths
and weaknesses of different theory types and methods. Certain types of
scientist may especially appreciate the advice contained herein: those in
graduate school or early in their careers who are still finding their way in the
scientific enterprise; those who have come to appreciate that their research
could benefit from a broader perspective and wonder how to proceed; those
in fields such as sociology which are by their nature open to a wide range of
theory and method; those who have come to wonder if the scientific
Preface xiii

community of which they are a part would be more productive if guided by

somewhat different principles; and, of course, those with a particular interest
in methodological or theoretical flexibility, or in classification itself.
This book should be of special interest to those in Philosophy of Science,
History of Science, and Social Studies of Science. Classification, it is often
noted, is the first step in science. Yet scholars of science lack a good
classification of “types of science”. This book provides such a classification,
and explores the strengths and weaknesses of different types of science. It
provides an organizing device, and a comprehensive set of questions, for the
study of science.
Since the book points the way toward a superior system for classifying
scientific documents, it should be of interest to those in information science.
Information scientists have long noted that researchers make relatively little
use of subject catalogues, though such use has increased somewhat with the
advent of electronic catalogues. Scientists often do not find what they need
to know in such catalogues: what research has been done that is similar to
that which they wish to undertake. Scientists need to be able to search by
(combinations of) phenomena, theory, method, and critique. The
classifications developed in this book, in conjunction with digital technology,
make this quite possible.
The book is useful for teaching as well as research. Students learn best
when they have some means of organizing their understanding into a
coherent whole. The book’s use as a text extends far beyond programs in the
study of science or information science. Many universities and colleges
struggle at present to provide an overview of (at least) the theories and
methods used by scholars. This book provides an easy structure for such a
course(s). Chapter 8 discusses how the various classifications could aid
student learning.
Interdisciplinarians – scholars, that is, who are self-consciously
interdisciplinary – should value this work for research as well as teaching.
In doing interdisciplinary research, one must identify the theories and
methods relevant to a particular question. This book can aid this process
immensely, by not only identifying the full range of possibilities, but
describing what uses particular theories and methods best serve. Chapter 8
will summarize a 12-step process for interdisciplinary research, and discuss
how each of the major classifications developed in this book would facilitate
key steps in interdisciplinary analysis.
This book draws upon a great deal of my previous research. The
discussion of phenomena in chapter 2 is grounded in several publications,
particularly “Toward a Unified Human Science”, Issues in Integrative
Studies, 2000, “Putting Social Structure in its Place, Schematically” Issues in
Integrative Studies, 2001, and A Schema For Unifying Human Science:
xiv Preface

Interdisciplinary Perspectives on Culture, Selinsgrove PA: Susquehanna

University Press, 2003. The classification of theories in chapter 3 draws upon
“Classifying Natural and Social Scientific Theories” Current Sociology 51:1,
Jan, 2003. Some of this material, plus key segments of chapters 4 and 7, was
contained in “Classifying Scientific Theories and Methods,” Knowledge
Organization 30:1, 2003. A summary of the classifications in chapters 2, 3,
and 4, as well as the classification of ethical approaches in chapter 5, and a
discussion of how these could inform curricular reform initiatives, can be
found in “’Comprehensive’ Curricular Reform: Providing Students With an
Overview of the Scholarly Enterprise,” Journal of General Education 52:1,
2003. The classification of ethical approaches in chapter 5 is developed in
more detail, and applied, in “Politics and the Five Types of Ethical Theory,”
International Journal of Politics and Ethics 2:4, 2002, and Unifying Human
Ethics, Ms. under review. My “How to Do Interdisciplinarity: Integrating
the Debate,” and “Intuition and Interdisciplinarity: A Reply to Mackey,”
both in Issues in Integrative Studies, 2002, discuss how the various
classifications developed in this book could inform the process of
interdisciplinary analysis; this material is treated in chapter 8. I thank each
of these outlets for allowing me to reproduce ideas from these previous
works in this book. While this book draws heavily on these previous
publications, it is much more than the stitching together of disparate works.
The book has a logic of its own, and many of the papers mentioned above
were written after the general project of this book was clearly in view.
I have benefited greatly from anonymous referees of all of these previous
publications. I have also consistently received great advice from editors, and
would like to thank Stan Bailis, Jay Wentworth, Hans Feldmann, Susan
McDaniel, Hope Olson, Marilyn Amey, and Patrick Hayden. Among those I
have acknowledged along the way, and would like to thank here again for
their advice and encouragement are Ray Miller, Fran Navakas, Michael
Bowler, Kathleen Maloney, Ross Emmett, Bruce Janz, Ronald King,
Elizabeth Jones, Wes Cooper, and Bill Newell.
An earlier version of the present manuscript was read in its entirety by
Adolf Buse, Reuben Kaufman, Julie Thompson Klein, and Bill Newell.
Their copious commentary has aided me immensely. This book is much
better as a result of their kind and sage counsel. I am immensely grateful. I
would also thank Steve Fuller and David Mitch for helpful advice on parts of
this manuscript.
I have presented some or all of this material at conferences of the
Association of American Colleges and Universities, Association for
Integrative Studies, Association for General and Liberal Studies, Canadian
Economics Association, International Economic History Association, Social
Preface xv

Science History Association, Society for Values in Higher Education,

Society for Social Studies of Science, Association for Graduate Liberal
Studies Programs, and Society for the Advancement of Behavioral
Economics. I thank all of the conference organizers for these opportunities
to present this material; I received useful input on every occasion. I would
also like to thank Linda Hanson for inviting me to discuss this material at
Ball State University; I greatly appreciated the opportunity to discuss this
endeavor with a wide range of scholars. I also made a presentation to the
Science and Humanities Circle at the University of Alberta; I would thank
Pamela Asquith and Andrew Ede for guiding me to very useful literature on
that occasion. Alex Rueger provided me with very useful advice on that
occasion and thereafter regarding the classification of natural science theory.
The first draft of this book, and many of the constituent articles noted
above, were produced during the academic year 2001-2, while I held a
McCalla Professorship at the University of Alberta. I thank the University of
Alberta, and the McCalla Award Committee, for this wonderful opportunity
to be temporarily released from teaching responsibilities so that I could focus
on my research.
I received valuable advice and encouragement from John Mackenzie
Owen, the editor-in-chief of the book series on Information Science and
Knowledge Management at Kluwer. Robbert van Berckelaer, and the staff at
Kluwer more generally, have been a pleasure to deal with.
Last, but far from least, I thank my wife Anne-Marie, and our children
Mireille, Julien, and Theodore. I thank them for urging me away to play, and
for understanding that I could not always oblige. I thank them for letting me
‘play’ on our computer so much. I thank them, most of all, for never letting
me forget what is important in life.
Chapter 1
CLASSIFYING SCIENCE

“Knowledge is classification.” John Dewey

Classification is the essential first step in science. Chemical reactions

cannot be analysed unless one chemical element is first distinguished from
another, atomic reactions cannot be comprehended without some sense of
subatomic particles, nor can the characteristics associated with creativity be
identified without some sense of personality dimensions. Classification of
science is likewise essential to its study.1
The burgeoning philosophy of science literature has identified some
broad distinctions such as “nomothetic” versus “idiographic” science,
“quantitative” as opposed to “qualitative” science, or “natural” versus
“social” science.2 Such distinctions will be addressed in later chapters. For
1
Mukherjee (1983) notes that scientists classify that which is perceived but whose
constitution and causes are not fully comprehended. “Types of science” thus seems a
worthy subject for classification. Mukherjee emphasizes that classification helps scientists
to describe, understand, and diagnose. Charles Sanders Peirce appreciated that,
“Classification [of science] is itself the business of a science, but which science?” (Kent,
1987, 18). Peirce was scandalized that “there was no agreement as to what constituted a
science for classification. Nor was there any uniformity as to where a given science ended
and another began ...” (1987, 85). Galison (1996, 13-14), citing many other scholars of
science, speaks of “the extraordinary variety of scientific languages, practices, purposes,
and forms of argumentation” (See also Fuller, 1997, 13-14). The diversity of science has
encouraged philosophers of science to move away from grand theorizing to develop
philosophies of particular disciplines or fields. This book celebrates both specialization
and integration (see below); its classifications allow specialized research to be tied into a
broader whole: this is as true for the study of science as for science itself.
2
Whitley (2000, xi) speaks of further simple dichotomies such as hard versus soft science:
“Such broad contrasts, which still seem popular, are clearly inadequate to deal with the
2 Chapter 1

now, it is sufficient to note that all of these distinctions are problematic: the
boundaries between classes are disputed, cases which seem to combine
elements of each class are legion, and some scholars are not as clear as they
should be about what they are attempting to classify, and along what
dimension(s). This book will develop classifications that are at once more
detailed and less problematic than these distinctions.
Questions of whether some types of science are better than others, or
whether different types of science are best suited to different purposes, can
hardly be asked intelligibly without first classifying types of science. Less
obviously, enquiries about the very possibility and limitations of scientific
enquiry would benefit from such a classification, just as the periodic table
aids those who wonder what chemicals can do. Moreover, without some
classification of science, scholars of science will communicate poorly.
“Communication between researchers is impossible unless they can identify
to one another the entities that they are talking about” (Ziman, 2000, 119).
Finally, we shall see that these classifications aid the performance of science
in a variety of ways.
A classification of science should ideally possess two qualities. First, it
should identify an exhaustive set of types of science, as the periodic table
identifies an exhaustive set of possible chemical elements. This would
provide insight into the possibilities and limits of the scientific enterprise.3
Such a classification might, in particular, encourage practicing scientists to
be open to types of science different from those they pursue. Second, the
classification should reflect some theoretical order. This would aid efforts to
identify the strengths and weaknesses of different types of science, and also
the potential for integrating across different types of science – as the periodic
table guides chemists as to the properties of particular elements, both alone
and in combination. Note that a theoretical grounding has the additional
important advantage of allowing the classification of new types of science (or
elements) even before these have been found or created. Both goals could
potentially be achieved either with a unidimensional classification or with a
multidimensional typology; both will be utilized in what follows.
While some elements had been identified before the development of the
periodical table, gaps in the table were then filled in by the search for or
creation of missing elements. This process betrays a mixture of induction
(various particulars are identified, and organized into a larger structure) and

major differences between, say, physics and chemistry in the twentieth century, let alone
the biological sciences.”
3
“Every program of social inquiry, every theoretical vision, I suggest, implies some model
of closure, some sense of how individual bits of insight fit, or fail to fit, into some larger
project of enlightenment” (Rule, 1997, 26). Scientific research programs should instead
proceed from an understanding of the full range of possibilities.
1. Classifying Science 3

deduction (a general approach is hypothesized, and particulars derived or

predicted from this). Both deduction and induction will likewise be utilised
in what follows: existing literature in the history, philosophy, social study,
and practice of science is sifted in order to identify different types of science,
while guides to classification are developed that help structure and even
identify types. The roles of induction and deduction more generally in
science are discussed in chapter 4.
Aristotle identified three types of science: the theoretical, practical, and
productive. These divisions embrace the modern natural and social sciences
as well as the humanities (McKeon, 1994, 13-4). This book follows Aristotle
in this broad usage of the word “science”. While such a usage is unusual
these days – though Whitley (2000) also treats science as broadly equivalent
to scholarship4 – the logic of the classifications to be developed in later
chapters calls for a broad perspective. Individual theories and methods are
often used across these scholarly domains. Breadth of definition allows
comparison of how research proceeds in these domains. It would be
undesirable to arbitrarily constrain the scope of analysis at the outset. For
scholars who study science itself, the broad approach has the additional
advantage of classifying not only that which they might wish to study, but
also the theories and methods they might wish to apply in their studies.

1. A SIMPLE CLASSIFICATORY GUIDELINE

The key deductive guideline in this book is grounded in one of the

simplest organizing principles possible: the 5W questions. That is, at each
stage of the investigation the five questions – Who, What, Where, When, and
Why – are asked. One key goal of a classificatory system is that it helps
users to recall diverse classes and where they fit. Since all are familiar (or
can quickly become so) with the five Ws, the 5W guideline is ideally suited
to the recall function. This book will show that this incredibly simple device
is also well suited to guiding scholars to ask and answer interesting questions

4
Whitley argues that the natural sciences can only be understood in
comparison with the social sciences and humanities, Dupré (2001, 4)
notes that (natural) science cannot answer some of humanity’s most
profound questions. He advocates “an approach that draws both on the
empirical knowledge derivable from the (various) sciences, and on the
wisdom and insight into human nature that can be derived from more
humanistic studies.” Rather than speaking of the “limits of science”
(Dupré, 2001, 17), this book includes humanistic study within science.
4 Chapter 1

about science. Given the tendency of both practicing journalists and scientists
to be guided by the five Ws, this result is perhaps not surprising.
Scientific research reports are generally more logical in format than the
research enterprise that they summarize. This project was not begun with the
five Ws in mind. This project was inspired by a sense of the desirability of
classifying science; a feasible means of doing so followed. Familiarity with
the use of the five Ws in the research of Marion Brady on interdisciplinary
teaching encouraged the investigation of the possibility of classifying science
in this way, after a fair bit of inductive contemplation had already been
performed.5 Once the approach seemed useful in a couple of contexts, its
wider applicability was naturally explored.
In addition to the 5W questions, scientists and journalists often ask
‘How?’ Indeed, this is a particularly important scientific question.
Philosophers, though, often distinguish scientific approaches as to whether
they are focussed on ‘Why?’ Or ‘How?’ This suggests that ‘How?’ is
actually a combination of (some of) the other four Ws. It has been noted that
when interviewers ask people a ‘How?’ question, they are generally
rewarded with a chronology of When and Where a process occurred
(Goldenberg, 1992, 118). ‘How?’ will thus be treated in what follows as
primarily a substitute for When/Where questions. When answers to ‘When?’
and ‘Where?’ are not obvious, asking ‘How?’ will prove a useful device.
‘How?’ can also serve as a check on the analysis: do the answers to that
question find a place in the 5W analysis?
It might be argued that if the ‘How?’ question were formulated as “How
does science proceed?” rather than “How do scientists do science?,” a
stronger distinction might be drawn: the 5W questions would interrogate the
inputs into science, and ‘How?’ would inquire about outputs.6 This book,
though, will link inputs and outputs throughout. It will argue that science
should proceed through the careful specification of phenomena, data, theory,
and method. It is hoped that classifications of these elements of science will
greatly facilitate communication across scientific communities, and thus the
cumulative development of scientific insight.

5
Brady notes that the 5W questions are interrelated. He proposes a superior educational
system whereby students are guided to explore the 5W questions as applied to pairs of
phenomena from a lengthy list. Some relationships will prove obvious, while others are
subtle or silly. Students are thus forced to think in a complex manner, appreciate the
importance of exploring causal links, and realize that all phenomena have diverse
influences. Crucially, they gain both an organizing system on which to hang diverse bits
of understanding, and an appreciation of those links and questions they have not explored.
See his website, http://digital.net/~mbrady/
6
I thank Bill Newell for drawing this possibility to my attention.
1. Classifying Science 5

The use of ‘How?’ as a test of whether the 5W questions have ignored

important aspects of science will help to establish that the 5W approach is
exhaustive, at least when classifying science. Of course, in establishing
exhaustiveness in this way, the “problem of induction” identified by David
Hume in the eighteenth century must be confronted: the fact that a million
white swans have been observed does not prove that black swans do not
exist. The fact that the 5Ws plus ‘how?’ are widely perceived in both
journalism (see Hough, 1984) and science to provide an exhaustive set of
questions provides a deductive rationale for this conclusion. Recall that
exhaustiveness is a highly desirable quality in any classification scheme. The
periodic table gives chemists an idea of what is possible; a less complete
table would guide chemists away from particular possibilities. Humanity
would have to rely upon less wonderful ways of inflating balloons if it did
not know about helium.

2. THE FIRST “CUT” (AND PLAN OF WORK)

The analysis has proceeded so far by reference to the rather uncongenial

formulation of “types of science”. Since science is a complex process, it is
hardly surprising that little analysis has been carried out at this level (at least
explicitly; as noted above some scholars are not as clear as they should be
about what they are attempting to classify, and along what dimension(s)). By
asking the 5W questions with respect to scientific investigation in general,
the classificatory enterprise can be broken into a (exhaustive) handful of key
components. These in turn can be defined in more congenial terms: “types of
theory”, for example, is much more suggestive of the goals of this book than
is “types of science”. This first “cut” then – or exercise in disaggregation to
use the jargon of the economist, or exercise in unpacking to use the jargon of
realist philosophy – serves to focus attention on a handful of key facets of
science. It can reasonably be hoped that these are the only facets of the
scientific process. Classification with respect to each of these will thus mean
collectively categorizing science itself.
Of any scientific investigation, it can be asked, “What is the scientist(s)
investigating?” Every investigation looks at something, be it atoms or
chemicals or personality characteristics or textual material or art (or often the
relationships among more than one “thing”). These things can be called
“phenomena”. Ideally, an exhaustive list of the phenomena of interest to
scientists (which should include all observable phenomena of interest to non-
scientists) would be provided, and placed within some orderly structure (see
Szostak 2000, 2003). Chapter 2 will reprise earlier research on the
6 Chapter 1

classification of human science phenomena,7 discuss how this can be

extended to natural science, and discuss how asking the 5W questions adds
insight to this organizing structure. Chapter 2 will also discuss to what
extent the data investigated by scientists represent the relevant phenomena,
and how data can be classified.
“Why is the scientist investigating that?” Ideally (baser motives will be
discussed below) the answer would be that they hope to enhance their
understanding of some aspect of the phenomena being studied. Note that
“understanding” is defined very broadly here, such that it would embrace the
efforts of those who instead emphasize “interpretation” (see chapter 3).8
Scientists organize their understanding in terms of theories. Indeed, the vast
majority of philosophers of science view “theory” as roughly the equivalent
of “scientific explanation” (Boylan and O’Gorman, 1995).9 A particular
investigation may be designed to derive, test, and/or apply theories, and these
theories may involve the definition of particular phenomena or (more
commonly) explication of links among these. This classificatory enterprise
should strive, then, to identify the various types of theory that might be
employed in scientific investigation, and then discuss their strengths and
weaknesses. A 5W typology will be employed to this end in chapter 3.
As noted above, a ‘How?’ question can often substitute for ‘Where?’ and
‘When?’ questions. “How is the scientist conducting the investigation?” All
scientists employ one or more of a small set of methods. Since there are only
a dozen of these, an exhaustive and manageable list of scientific methods can
be presented in chapter 4. At that time, the 5W questions will prove
invaluable guides to the identification of the relative strengths and
weaknesses of different methods. Note that questions of ‘where’ and ‘when’
a scientific investigation is undertaken can and will be subsumed within the
discussion of methods.
That leaves the ‘Who?’ question. Care must be taken here. To ask “Who
is the scientist investigating?” would, in this context, be a subset of the

7
Common practice (in North America) is followed here in defining “human science” as the
social and behavioral sciences, plus the humanities. Caws (1993) defines human science
as the study of processes involving human agency. The list of human science phenomena
in Table 2.1 provides a more precise definition of what human science encompasses.
8
Hempel and Oppenheim argued that the purpose of science was explanation, because
scientists were concerned with answering ‘Why?’ questions. Scriven (1988) notes, though,
that scientists engage each of the 5W questions with regard to the phenomena they study,
and answers to each of these questions can be termed explanation.
9
Boylan and O’Gorman render this judgment despite disagreeing with this view. They note
that Quine equated “theory” with “description”, although he said little about “explanation”.
Van Frassen emphasized that “explanation” requires a narrative approach; he, and Boylan
and O’Gorman in turn, nevertheless appreciate that such a narrative will be grounded in
theory.
1. Classifying Science 7

“What is the scientist investigating?” question dealt with above, for

phenomena were defined to include individual-level phenomena (see chapter
2). The question to be asked is thus “Who is doing the investigation?” So
far, the 5W questions have guided the classification of familiar elements of
science: phenomena, data, theories, and methods.10 The ‘Who?’ question
turns the focus on scientists themselves. Some may still cling to the ideal
that science is an entirely objective pursuit, and thus not dependant at all
upon who is doing the research. However, the past few decades have
witnessed several attacks on this and related scientific ideals. It is now
widely appreciated that scientists deviate to varying degrees from a selfless
and objective pursuit of understanding. Chapter 5 will discuss how to
categorize these concerns in terms of the 5W questions. It is chapter 5, then,
that investigates the variety of possible motives beyond that of enhancing our
collective understanding that might characterize a particular researcher. Of
particular note, in answering the ‘Why?’ question five broad types of ethical
analysis will be identified, and their implications for how the goals and
standards of scientists might be evaluated are discussed.
What is science? Philosophers are far from consensus on this seemingly
simple question (with the inevitable result that some scholars doubt that
science exists). Yet many philosophers and most scientists are confident that
they “know it when they see it” (Ziman, 2000). Philosophers of language
appreciate that it is generally impossible to provide precise definitions of any
term in a sentence or two. Wittgenstein thus suggested that the best
definition of a term would involve providing examples of its use. Thus the
word “game” could be defined by listing baseball, golf, chess, and other
examples. Note that there may be few or no commonalities among pairs of
examples, as with baseball and chess. While Wittgenstein did not mention
the possibility, it must seem that a precise definition of a word could be
obtained by providing an exhaustive list of examples (Szostak, 2003). In this
book, potentially exhaustive lists of phenomena, types of theory, and
methods will be provided. In chapter 5, science will be defined in terms of
these. Chapter 5 will then investigate how well this definition can
distinguish science from non-science. Chapters 3 and 4 will develop
definitions of scientific theory and method in a similar manner, and engage
with arguments that these terms cannot be defined. On a related matter, it
will be argued in chapter 2 that there is little or no place in the scientific
enterprise for “concepts” that cannot be defined precisely in terms of
phenomena, theory, or method.

10
Hall and Hall (1996), following Gilbert, argue that there are three major ingredients in
social research: theory, method, and data. The analysis so far has only added reference to
the phenomena that data are supposed to represent.
8 Chapter 1

One huge advantage inherent in these varied efforts at classification is

that they set the stage for an attempt to draw connections across particular
classifications. Chapter 6 discusses the relationship between theory and
method: to what extent do particular types of theory encourage the use of
particular methods, and vice versa? The relationship between method and
phenomena is then investigated. Scientific disciplines can generally be
defined in terms of both phenomena of interest and favored methods. This
chapter asks whether such strong links between phenomena and method can
be justified. The advantages and disadvantages of scientific “paradigms” are
also explored, and a superior scientific attitude toward these is outlined.
Finally, chapter 6 discusses the degree to which the various potential biases
or errors in scientific analysis identified in chapter 5 vary in importance by
phenomena, data, theory, or method. Research in the study of science
generally explores how some subset of these biases or errors has affected a
particular research endeavor. Yet the study of science rarely draws explicit
comparisons across different types of science (Whitley, 2000, 32). These
classifications allow analyses of science to be organized into a more coherent
whole, and will facilitate explicitly comparative work. The simple act of
identifying which types of science are most affected by particular biases or
errors can point the way to improvements in scientific practice.
Chapter 6 (and much of chapter 5) thus serve to justify the efforts at
classification in the preceding chapters. Chapters 2, 3, and 4 also close with
some discussion of the advantages of the particular classifications developed
therein. After all, systems of classification must be judged in terms of their
utility. Chapter 7 provides yet further justification, in discussing how these
classifications could form the basis of a vastly superior system for classifying
scientific documents. It is virtually impossible to search libraries or web
collections by theory, method, or type of bias or error, and quite difficult in
terms of phenomena or causal links. Scientists are thus severely limited in
their ability to identify related research, especially in fields other than their
own. Scientific productivity could be greatly enhanced if such searches were
possible. Researchers would be much better able to survey relevant
literature, and the scientific community would be much less prone to
“reinventing the wheel” by replicating forgotten research.
Chapter 8 summarizes the classifications, advantages of these,
connections drawn across classifications, and Golden Mean arguments (see
below) of preceding chapters. It then briefly extends these to engage three
key questions. First, should science be believed, and if so why? It will be
argued that science, while imperfect, can and does enhance human
understanding; while the errors and biases in science should be ameliorated
they do not overwhelm the scientific enterprise. Second, how might science
be improved? The classifications developed in this book can enhance the
1. Classifying Science 9

productivity of the scientific enterprise in a variety of ways. Most

importantly they encourage and allow all scientists, regardless of field, to
carefully specify the method, theory, and phenomena used in their study, and
in a manner that should be readily comprehensible to other scientists and the
educated public. Particular advice will be given to both specialized and
interdisciplinary researchers. Third, how should science be taught? The
classifications developed in this book provide an organizing structure for
students; this is especially valuable in integrating material from different
disciplines or fields.

3. SOME PRELIMINARIES

There are a handful of preliminaries that should be dealt with before

proceeding further. First, the scientific literature on classification should be
discussed, in order to establish the characteristics to be both sought and
avoided in a system of classification. Second, it is useful to discuss the
limitations of the few previous efforts to classify science. Third, and less
obviously, the Aristotleian Golden Mean will be discussed. This simple idea
will provide a guide through much philosophical dispute in what follows.
Fourth, it is useful to discuss interdisciplinarity, both to foreshadow the
implications of this book for interdisciplinary research and teaching, and to
indicate the sort of interdisciplinary vision that motivates this work.

3.1 Theories of Classification

Bailey (1994) can profitably be followed here. He notes that the main
goal of a classification system is to maximize both cross-group heterogeneity
and within-group homogeneity, but that in general there is a tradeoff between
these criteria. If, for example, that which we wish to classify is distributed
evenly along some dimension, dividing it into small groups will ensure
substantial homogeneity within groups, but also that neighbouring groups are
quite similar as well. The periodic table is thus a misleading guideline, for
many of the things one would wish to categorize are not so discretely
differentiable. While heterogeneity across and homogeneity within types of
science are goals to be pursued in this work, and generally with great
success, it is noteworthy that most categorization schemes have difficulties in
ascertaining where group boundaries are best situated and where particular
cases should be classified.
Bailey argues that the basic secret to categorization is the “ability to
ascertain the key or fundamental characteristics on which the classification is
to be based,” which in turn can only come from a mix of prior knowledge
10 Chapter 1

and theoretical guidance (1994,2). This advice accords well with the premise
above that both induction and deduction should be used to derive
classifications of science.
Bailey defines a typology as a classification that is multidimensional and
conceptual. That is, it is organized in terms of at least two theoretical
criteria. (While some scholars use taxonomy as a synonym for typology,
Bailey reserves taxonomy for multidimensional classifications based on
empirical criteria).11 Bailey notes that if there are many dimensions, it may
be advisable to group nearbly cells of a multidimensional table in order to
obtain a manageable number of classes (a process which is necessarily a bit
arbitrary,12 but often aided by the fact that various cells have no observed
cases). One advantage of the 5W approach to typology is that it generates a
manageable number of dimensions, and thus eliminates the need for
artificially grouping disparate cells into classes. Bailey warns scholars not to
sacrifice dimensions that identify important differences merely to achieve
manageability; the 5W approach will provide insight as well as
manageability. Nevertheless, it is important to show in what follows that 5W
typologies subsume diverse distinctions noted by previous scholars.
There are two broad types of classification. The first, “classification”
proper, involves a unidimensional rendering. For example, a unidimensional
list of scientific methods will be provided in chapter 4. The second type of
classification, typology, involves a multidimensional rendering. The 5W
questions will be used throughout this book to develop such typologies. Both
unidimensional and multidimensional classifications have advantages, and
thus the possibilities for both types of classification will be explored in each
of the next four chapters. Note that both types of classification can
potentially achieve the dual goals of identifying an exhaustive set of classes
and identifying some rationale for the existence of these classes.
Bailey (1994, 12) discusses several advantages of a typology. Scholars
can see where a particular class falls along each dimension. Scholars can
then see which classes are most similar (very important in modern
medical analysis) and most different. Scholars can also ensure that their

11
Bailey (1994, 32-4) is naturally enthusiastic about a third possibility of operational
taxonomy, in which conceptual typologies are found to have empirical validity. This
encourages a mixture of qualitative and quantitative approaches. Ragin (2000, 76-7)
criticizes the common tendency of sociologists to use exclusively the quantitative
technique of cluster analysis to identify clusters which share many but not all
characteristics; he worries that this detracts attention from the importance which
differences along any one dimension may have.
12
It may be explicitly arbitrary, as when index numbers are used, or pragmatic, as when the
goals of a particular research project determine how cells are clustered. Less arbitrarily,
cells may be eliminated if judged theoretically or practically impossible.
1. Classifying Science 11

classification is comprehensive, and thus provide an inventory so that they

“know at all times what types [classes] are available for analysis.” If it turns
out that many cells are empty, insight is provided into how the various
dimensions are related. With respect to science, scholars can thus aspire to
answer a host of questions which are rarely even asked, such as: “Which
methods are most similar?”, “What (types of) theory could possibly be
applied to a particular issue?”, “How can the types of theory best be
organized?”, “What methods are most complementary?”, and “What can be
learned about theory, method, and phenomena from observed correlations
among answers to the 5W questions?”
There are, though, several disadvantages inherent in any form of
classification (Bailey, 1994, 15-16). First, a classification scheme may be
merely descriptive. Many scholars nevertheless see this as a useful end in
itself, while others see descriptive classification as but a stepping stone to
analysis.13 It was suggested above that describing the various types of
science is a useful goal in its own right, but that the classifications developed
here will also be used to analyze the strengths and weaknesses of different
types of science. This potential drawback can thus be ignored in this work.
A more serious danger is “reification”: the creation of artificial entities
that then come to be viewed as real.14 This danger will be minimized here in
many ways. When classifying, scholarly consensus will be utilized wherever
possible, and competing arguments will be discussed where there is no
consensus. Moreover, the creation of artificial classes should be limited by
the utilization of the simplest of typological devices, the 5W questions. The
5W typologies facilitate both exhaustiveness and the identification of empty
cells: this should protect against the twin dangers of ignoring classes that
exist and imagining classes that do not. (It must, of course, be asked whether
empty cells reflect impossibility or merely a failure in scientific practice.) By
analyzing the similarities and differences across classes, the artificial
grouping of disparate cases or separation of similar cases can be avoided.
Typologies tend to be retained only if they provide insight into important
questions; the fact that these typologies will provide insights into scientific
practice provides further evidence that they reflect reality.

13
“Categories and their boundaries are centrally important in science, and sciences are
especially good at documenting and publicly arguing about the boundaries of categories”
(Bowker and Star, 1999, 296). This book will hopefully stimulate scientific attention to
the task of classifying science itself.
14
Some would minimize the dangers of reification. Newton, it is noted, started theorizing
from concepts of effort and movement, before developing a theory in terms of force, mass,
and acceleration. Since classification is an essential component of research, even a
misguided system may be better than none in guiding inquiry toward better systems.
12 Chapter 1

A distinction is often drawn between “natural” and “artificial” (non-

natural) classes. The former are taken to refer to classes that exist before
human investigation, chemical elements being an obvious example. Root
(1997) speaks of the various types of sexual deviance defined by late
nineteenth century psychoanalysts as an example of artificial classes. Note
that scientists can be confident of their artificiality precisely because
scientists now doubt that they exist. If not for this, it might be conjectured
that these classes too were “out there” long before their discovery by science.
Strict naturalists limit the term “natural class” to things that existed before
any sort of human intervention: marital status, religion, and perhaps even
gender can thus not qualify as natural classes. Root (1997) favors a less
strict definition such that human decisions that precede scientific
investigation should also count as natural classes: marital status, occupation,
gender, and a host of other socially constituted classes would thus qualify.
He worries, nevertheless, that too much credibility may be attributed to
socially created classes: he argues, for example, that prostitutes were not
considered a separate class before nineteenth century concerns with public
health.15 In this, he follows Foucault, who worried that scientific
classifications of people as insane or criminal serve as mechanisms of social
control. However, this observation need not render the categories
meaningless: instead one could analyze the phenomenon of “crime” or
“mental disorder” in terms of multiple causal links, including the personality
characteristics of those involved, and the cultural attitudes and political and
social considerations at work (Szostak, 2003).16 But neither Root’s extension
of the definition of natural class, nor any defence of this extension, aids the
present enterprise. Classifications of science can hardly precede scientific
analysis itself, and thus the classifications provided in this book must surely
be non-natural. Does this matter? The distinction, it seems, is made by those
who wish to limit the opportunity for reification of classes that do not
reflect an external reality. Yet scientists cannot be sure that natural classes
reflect reality. As Root notes, the argument that “vertebrates” is a useful class
of animals rests on scientific consensus that this is a more useful

15
He surely exaggerates, for prostitution must at least have been recognized as an
occupation. But his point is that prostitutes could more easily move across occupations
and marital status before they became branded for life.
16
Root (1997, 166) worries that if a class such as “homosexual” is created for purposes of
social control, scientific analysis of that class must reflect that bias. But if Root is right
that before the invention of homosexuality bisexuality was the norm, then scientific
investigation of the genetic causes of homosexuality would find that there is no difference.
Root then worries that analysis of marital status must support the legal preferences
afforded heterosexual couples; but surely analysis of the effects of such preferences might
yield the opposite conclusion?
1. Classifying Science 13

zoological classification than “animals eaten by the English”. The

natural/non-natural distinction does not do away with the need for scientific
analysis and consensus. And this in turn depends on a system of
classification being “justified”. As noted above, the classifications
developed in this book will be justified in many ways.17
Bryant (2000) identifies an “essentialist” attitude that the world divides
naturally into classes, and strives to discredit this. She argues that there is
always more than one way that any set of entities could be distinguished, and
none of these merits priority (as do Bowker and Star, 1999, 322-3). While
most of her examples are from biology, she does make passing reference to
the periodic table, arguing that scientists could conceivably classify in terms
of isotope number rather than atomic number (2000, 89-90). This brief
argument seems highly questionable: there are good theoretical reasons for
preferring atomic number, and classifying by isotope could be seen merely as
unpacking elements into their isotopes. Nevertheless, Bryant is likely correct
that much of the time there is no one unique system by which particular
entities might be classified.18 This raises the possibility that various biases
might influence the choices made by any scientist: “Each standard and each
category valorizes some point of view and silences another. This is not
inherently a bad thing – indeed it is inescapable” (Bowker and Star, 1999, 5).
Bryant suggests two criteria by which the best classifications might be
chosen in such situations: the “what is useful” constraint should be heeded by
classifying in terms of distinctions that actually exist,19 and the “suitably
scientific” constraint should be heeded by providing theoretical justification
(2000, 111-6). In what follows, all classifications will be justified both
theoretically and empirically, and their advantages over alternatives
explicated. In the case of unidimensional classifications, it will be especially
important to provide evidence that these reflect an empirical reality.
As Bailey notes, the static nature of most classification systems can be a
drawback for some uses. As society evolves, the classification system may
need to change as well.20 It is of course possible that scientists will discover
new theories, methods, or phenomena, and philosophers develop new

17
Ziman (2000, 120) follows Kant in arguing that humans have an inherent capacity for
pattern recognition (and thus classification). Considerable scientific evidence supports this
conjecture.
18
This creates problems for systems of document classification. For a system of
classification to be mutually exclusive, each step in disaggregation must occur in terms of
only one characteristic. This is generally impractical (Langridge, 1992, 15). The logical
structure of the classifications in this book should minimize this problem (see chapter 7).
19
Fabian (1975) also urges scholars to reduce bias by classifying with respect to the real
world rather than just with respect to other classes.
20
Bailey (1994, 8)also distinguishes a synchronic classification based on analysis at one
point in time from diachronic analysis of classes of phenomena that evolve over time.
14 Chapter 1

analyses of both science and ethics. Indeed one purpose of interdisciplinary

scholarship is the identification of theories, methods, and perhaps phenomena
that have been ignored by disciplines. In such a case, it would be important
to find a place for these in the system of classification. A classification of the
phenomena of interest to human scientists was developed in Szostak (2003)
that applied equally well to the study of traditional and modern societies.
The classification was, nevertheless, flexible enough that new phenomena
could be added as discovered. A vast array of ethical analyses from across
diverse philosophical traditions and a time span of millenia was classified
into five broad groups by Szostak (2004). This book will show that the
methods and types of theories used by scientists evolve so slowly that a static
classification is well suited to the task. It is probable that any new
discoveries can be added without upsetting the classification scheme to be
developed. Moreover, to the extent that the full range of possibilities is
identified in these typologies, they will necessarily account for all future
discoveries, just as the periodic table has provided places for elements not yet
created.
Bailey also speaks of problems in selecting the appropriate dimensions
for a typology. The 5W approach has been found in many (scientific and
everyday) applications to yield a comprehensive set of dimensions for
analysis. By choosing one simple approach, the possibility that authorial bias
influences the results is substantially reduced. Note, though, that there is still
some scope for interpreting the precise nature of the five questions in
particular circumstances.

3.2 Classifying Scientific Disciplines

Both in ancient Greece and turn of the twentieth century America,

academic specialization by discipline within educational institutions raised
the obvious questions of how disciplines could be both delimited and
justified. Was there some logical division of the scientific enterprise? Did
some disciplines depend on others for inputs? Could one determine logically
how many disciplines there should be? How, in other words, could academic
institutions decide whether or not to create a new discipline? This research
enterprise largely faded away as the twentieth century proceeded (but see
Whitley, 2000). The further proliferation of disciplines (much less
interdisciplinary fields) discouraged efforts to define an ideal set of these, but
encouraged instead a tendency to see disciplines as somewhat arbitrary
divisions of the scholarly enterprise, especially in human science. The
boundaries of disciplines evolve historically. The result at any point in time
is that disciplines overlap in places and leave unexplored territory between
themselves in others (Messer-Davidow et al., 1993; Klein, 1996).
1. Classifying Science 15

Any scholarly enterprise that attracts hundreds of scholars over a period of

decades is unlikely to be entirely misguided. Peirce felt that all scientists
should and would like to know how their discipline was related to others
(Kent, 1987, 52). This insight can be redrafted to remove the disciplinary
focus: all scientists should want to know how their research relates to the
research of other scientists. The classifications to be developed in this book
will allow all scientists to situate their work within the broader scientific
enterprise. Peirce emphasized three types of relationship (Kent, 122; see
Bechtel, 1986, for a similar analysis). First, disciplines might overlap with
respect to a certain topic. The classifications developed in this book will
allow scientists in different disciplines to ascertain when they are addressing
the same topic. Second, one discipline might generate a problem that others
could solve. Classifications, especially of phenomena, but also of theory
types and methods, will allow scientists to better see where they might look
for solutions to their problems. Third, one discipline might borrow principles
from another. Classifications of theories and methods will allow scientists to
borrow from the full range of scientific research possibilities; while the
typology of (critiques of) scientific practice could help scientists learn ways
of reducing bias and error. Peirce recognized that any classification should
reflect both induction – from studying how science works in practice – and
deduction – the search for logical classes (Kent, 48-51). Both induction and
deduction will be employed in what follows. Peirce recognized a potential
conflict between classifying “sciences as they are” and “sciences as they
should be”. The approach taken in this book allows science to be classified as
it is, while suggesting various ways in which science could be improved.
Indeed, much improvement will flow from the simple act of classification.
Peirce appreciated that science could advance without classification, but with
“unspeakable waste” (Kent, 1987, 55). Hopefully, by shifting the focus
from disciplines, and drawing lessons from these century-old efforts, this
book can achieve a much more useful set of classifications.
Otto Neurath, Rudolph Carnap, and others, heroically attempted an
“Encyclopedia of Unified Science” in the 1930s. They despaired of the
dangers of over-specialization, and hoped to show how various disciplines
were inter-related. In the end, only some 19 of 260 projected contributions
were assembled (Neurath et al., 1970). Neurath noted on the first page that
“The classification of sciences may be interesting in itself and sometimes
also stimulating, but the premature drawing of boundary lines does not help
us much in research work, not even in arranging its results, as long as we are
not able to combine the classification of simple items with the presentation
of interrelations between them.” Neurath and his colleagues were unable to
meet this goal by building up inductively from an exploration of disciplines.
Arguably, this goal can only be accomplished by (both deductively and
16 Chapter 1

inductively) developing discipline-spanning classifications of the key

elements of scientific practice.
This book, then, hopes to provide for the first time a simple, detailed, and
exhaustive classification of science itself that will allow all scholars to see
where their own research fits within the scientific enterprise as a whole. We
have discussed above how science is best classified in terms of phenomena,
data, theory, method, and practice. In each of the next four chapters the
limitations of any previous efforts to classify these will be detailed: the most
common observation will be that simple, detailed, and exhaustive
classifications have not previously been attempted.

3.3 The Golden Mean

At several points in this book “extreme” philosophical arguments that

would cast doubt upon elements of this project will need to be confronted:
for example, the argument that science does not exist. While specific
counter-arguments will be employed in each case, it is useful to first make
the simple but oft-ignored point that there is almost always some validity to
opposing positions in any debate. Thus scholars should always be wary of
extreme positions. Notably it is not necessary to back very far away from
any extreme point of view in order to justify the present undertaking.
The Golden Mean is widely applauded as a guide to ethical behavior (see
Szostak, 2004). Aristotle – and indeed philosophers across most, if not all,
philosophical traditions – had noted, for example, that courage taken to an
extreme became foolhardiness. The best practice, then, was not an extreme
but a balance between courage and caution. The same guideline can usefully
be applied to a host of other behaviors and attitudes. Some seventy ethical
statements were identified in Szostak (2004) that received support from each
of the five main types of ethical analysis (see chapter 5); the vast majority of
these referred to some sort of Golden Mean. The application of the Golden
Mean was thus extended to a much wider set of cases than it had been by
previous philosophers.
A “kernel of truth” belief was also advocated in that book: any argument
that has been accepted by many people over a long period of time likely has
some kernel of truth in it. A logical extension of this argument is that when
faced with two opposing viewpoints on an issue, each of which has struck a
cord with many people over time, it is likely that the correct answer lies
somewhere in between. One could, for example, be sceptical of a statement
that no such thing as a “correct answer” exists, while accepting supporting
arguments in support of that statement which suggest limits to human
understanding. More generally, in philosophy of science (and often in
science itself) as in ethics, one should seek a Golden Mean rather than cling
1. Classifying Science 17

to extreme viewpoints. In this book, the Golden Mean will be applied to

philosophical “what is true?” questions as well as to “what is good?”
questions. It is important to note, though, that the term “Golden Mean” can
be misleading: while the truth lies somewhere between extremes, it may lie
much closer to one than the other.
What, though, of arguments such as that the sun revolves around the
Earth? Surely scholars would want to stick to the extreme view that the Earth
revolves around the sun? Note first that the Golden Mean will be applied
mostly to philosophical rather than scientific debates in what follows. Note
also that in the face of modern evidence vanishingly few people hold to an
Earth-centred view. In science in particular, the compilation of enough
argument and evidence in favor of a particular point of view such that it
becomes virtually unanimous is a valid goal, and obviates the need for
Golden Mean analysis (though the possibility that unanimity is grounded in
bias or error, rather than understanding of reality, needs to be interrogated).
While scientists can thus ignore the Earth-centred alternative, they could if
they wished seek supporting arguments within an Earth-centred view that
were acceptable to them. Ptolemy had, after all, based his Earth-centred
conjecture on some reasonable premises. What, though, if some cult gained
a wide following on the basis of an Earth-centred view grounded in bizarre
premises? This scenario suggests that the statement “always seek the Golden
Mean” itself deserves some Golden Mean softening.
Does it matter? There are a host of philosophical questions that will of
necessity be touched upon in what follows.21 In each case this book will
show how an appreciation of the different types of science points away
from extremes, and in turn how a relaxation of extreme views guides an
appreciation of the classifications developed here. Note that it is generally
straightforward to discredit the most extreme claims; Bunge (1998, xii)
argues that one can overturn extreme scepticism of the possibility of
scientific advance by merely pointing to any one success in science.22 It is

21
“Prominent among the targets of my criticism are certain radical views: holism (or
collectivism) and individualism (or atomism), spiritualism (idealism) and physicalism,
irrationalism and hyperrationalism (apriorism), positivism and obscurantist anti-positivism,
zealotry and insensitivity to the moral aspect of social issues, bogus rigour and the cult of
data, grand theory and ideology, moralizing ignorant of social science, and the
compartmentalization of social studies” (Bunge, 1998, xi-xii).
22
Of course, no philosophical argument is perfect, and thus one can always question whether
any particular piece of theory takes us closer to a true understanding of the world. But the
vast majority of scholars and casual observers would accept that there is evidence of
advanced understanding in both natural and social science. With respect to the former,
Gower (1997, 4) gloats that, “The facts that keep aircraft aloft are, fortunately for
passengers, not socially determined but ascertained on the basis of evidence by logical
reasoning.”
18 Chapter 1

generally necessary only to back ever so slightly away from an extreme

position for the present project to be supported. Note in this regard that any
scientific or philosophical argument will offend extreme proponents of some
philosophical (or social, cultural, or political) perspective.
Arguments regarding the possibility of scientific advance deserve special
attention. The entire endeavor of this book would be of little use if, as some
argue, science is a profitless enterprise, incapable of providing humanity with
enhanced understanding of the world it inhabits. Of course, if this were truly
the case, this endeavor would still be exactly as useful as any other piece of
science. The utility of this enterprise would also be questionable if an
opposing extreme view of science were accepted: that it is a perfectly rational
and objective enterprise that inevitably leads to enhanced understanding.
Why bother classifying science if it is either impossible or perfect? But the
truth, in line with the Golden Mean, lies somewhere in between: science is
imperfect but contains elements that cause some tendency toward enhanced
understanding (Kincaid, 1996, 8-9, makes a similar argument). This being
the case, classifying science can be a project of great utility: it can identify
scientific practices that enhance or inhibit the pursuit of enhanced
understanding. This will be done throughout the book, but especially in
chapter 5 where (critiques of) scientific practice will be classified. In that
chapter diverse illustrations of the argument that science is neither perfect
nor useless will also be provided.
Ziman (2000) devotes much of his book to outlining the deficiencies of
both of these extreme lines of argument. Notably, he does so by looking in
detail at how scientists actually operate on a daily basis. Much of this book,
and particularly chapter 5, will follow a similar stratagem. But Ziman is all
too aware that both extremes exert a “strange attraction” on scholars. “I am
continually aware of the danger of leaning too far towards one side or the
other, and falling into the corresponding black hole” (2000, 233). He
recognizes that many of his colleagues, faced with his Golden Mean
conclusion, would urge him to “get off the fence.” Scientists understandably
find it difficult to work in an environment of chronic doubt, and are thus
tempted to either embrace despair or to ignore the potential imperfections in
scientific practice (Seale, 1999, 19). Kitcher (1982, 34) fears that scientists
reminded of their fallibility can too easily become sceptics and treat science
like religion: he notes that between the two extreme positions there is a
continuum where scientists can and do amass better and better evidence in
support of their theories. Fortunately, casual empiricism nevertheless
suggests that while extremists are vocal they are a distinct minority.
Scientists in practice recognize to some degree that science is fallible and yet
still useful. Happily, Moser, Mulder, and Trout (1998) confirm that this is
also the case among epistemologists, or philosophers of knowledge. A few
1. Classifying Science 19

argue that humans can know everything or nothing, but most accept that
humans have an imperfect capacity to enhance understanding.23
Polkinghorne (1996) notes that from the perspective of a practicing
physicist, the strong program in the sociology of knowledge, which contends
that developments in scientific theory reflect only sociocultural influences, is
nonsense. The behavior of the natural world, as revealed through
experiments and other methods, places severe constraints on the range of
theorizing possible. Yet Polkinghorne appreciates that social and cultural
influences matter. It would be a mistake, then, for practicing scientists to
completely reject all insights of the strong program. By advocating a Golden
Mean approach to this and other questions, this book hopes to speak to both
those who practice and those who study science (who are, notably, scientists
themselves within the definition of science used in this book).

3.4 Interdisciplinarity

This book grew out of the author’s interest in the practice of

interdisciplinarity. The book nevertheless speaks to both specialized and
interdisciplinary scholars. Since one of the key messages of the book is that
scholarly communities can and should be open to a wider range of
phenomena, data, theory, and method, and since theories and methods are
often embedded in particular disciplines, the book naturally informs
interdisciplinary practice. Various implications of particular classifications
for interdisciplinarity will be discussed in the next chapters. Chapter 8 will
outline a twelve-step process for interdisciplinary analysis that relies on each
of the classifications developed in the book. Interdisciplinary studies are
occasionally accused of being ad hoc and non-rigorous. The process to be
outlined provides a credible counter-argument, showing how inter-
disciplinary analysis can be systematic and cumulative.

23
Epistemologists ask questions such as “What does knowledge consist of?”, “How do
individuals acquire it?”, “How can it be distinguished from opinion?”, “How much should
one rely on other people?”, and “What is the role of scepticism?” (Moser et al., 1998).
With respect to science, this book will build toward answers to the first three questions in
chapter 5, when science is defined and distinguished from non-science. The question of
authority will be discussed briefly in chapters 2 and 3. Various types of scepticism will be
classified, and their insights discussed, in chapter 5. Some will worry that the consensus
identified by Moser et al. applies only to the English-speaking world. Norris (2000)
suggests that there is a greater degree of consensus between Anglophone and Continental
philosophers than is often appreciated. He also suggests that extreme scepticism is often
grounded in a misperception of certain problems in quantum physics.
20 Chapter 1

As Klein (1996) notes, the increasingly common practice of borrowing

theories and methods from other disciplines need not foster
interdisciplinarity. The receiving discipline may simply make the borrowed
theory or method its own. This may result in the same broad theory or
method evolving in different directions in different disciplines. Even
scholars who do not consider themselves interdisciplinary can thus benefit
from the identification of the key strengths and weaknesses of different
theories and methods. In chapters 5 and 8 in particular, suggestions will be
made as to how specialized researchers might usefully broaden their
perspective. Often, though, interdisciplinary fields emerge, such as cognitive
science, that continue to explicitly borrow theories and methods from
multiple disciplines (Klein, 1996, 18-9).
As noted by Salter and Hearn (1996) and Klein (1996), the word
“interdisciplinary” has multiple and conflicting definitions. For some
scholars, interdisciplinarity means the critique of disciplines. Some
interdisciplinary scholars question the very possibility of knowledge (see
Moran, 2002. Interdisciplinarity is pursued here as an effort to gain broader
and more accurate understanding of particular themes or questions (and the
interrelations among these) than can be found through the lens of any one
community of scholars. It can be contrasted with the (important) research of
specialized scholars, who apply a subset of scientific theories and methods to
a constrained set of phenomena. Integration does not just mean adding
together disciplinary insights, but involves contending with disagreements
and filling in missing pieces of analysis. It will be shown that the
classifications developed here aid all types of integration.
The discourse of interdisciplinarity is essential if scholars wish to hold
on to the idea of the unity of science in an increasingly specialized world
(Weingart, 2000, 41). The unity of science has been sought in the past in
terms of some grand unifying theory or method. It is proposed here that
science be understood as the interconnected study of a complex but finite list
of phenomena through the application of a finite set of theory types and
methods, and involving a finite set of philosophical perspectives, errors, and
biases (where “finite” should not be taken to mean that the full set is
necessarily already known). Galison (1996, 4-6) argues that the popularity
of the idea of the unity of science waned in large part because of broader
cultural movements: at the turn of the last century unity was seen as the
answer to intolerance and prejudice of all types; a century later diversity is
celebrated and many fear the homogenizing influence of a unified science.
The view of unity proffered here allows scholars and communities of
scholars to continue to choose a mix of phenomena, theory, method, and
philosophical perspective. The classifications provide guidance on the
advantages and limitations of different combinations, but do not exclude any
1. Classifying Science 21

from consideration. The benefits of unity can thus be gained without the
costs. Only by developing the classifications in this book can both
specialized and interdisciplinary scholars see how each scientific community
fits within the broader scientific enterprise. Bechtel (1986) has described
interdisciplinarity as an ongoing process of discovery, not an attempt to
systematize. Klein (1996a, 21) and many others counter that it can be both:
the classifications in this book encourage both scientific discovery and the
integration of diverse insights into a broader whole.
Weingart (2000) explores the apparent paradox of constant pleas for
interdisciplinarity, but the continuing strength of disciplines. He concludes
that scientific specialization within disciplines and interdisciplinarity “both
are crucially important; they are complementary rather than contradictory”24
He notes in particular that interdisciplinarity aids discovery by juxtaposing
different perspectives, theories, methods, and so on. He draws upon
Hubenthal (1994), who had in turn cited various famous scientists on how
their major discoveries had reflected such juxtaposition (see also Palmer,
1996, and Bromne, 2000; the latter argues that cognitive development of any
type requires the confrontation of different perspectives). Yet Weingart
appreciates the advantages of specialized research too, in allowing the
rigorous study of narrowly defined questions within a community of scholars
with a shared understanding, vocabulary, expertise, and purpose. Dogan and
Pahre (1990) coined the term “hybridisation” to reflect the fact that both
increased specialization and the joining of specializations to form new fields
are essential processes in the evolution of science. By identifying the place
of every scientific specialization within the scientific enterprise, the
classifications developed in this book should aid the essential process of
developing links across specializations.25 As Dogan (1996, 296) notes, “the
process of hybridisation consists, first of all, of borrowing and lending
concepts, methods, theories, and praxis”
Vague or implicit classifications are a poor basis for science of any
sort, but especially for interdisciplinary science. The specialist can learn
to appreciate what is studied in their field, albeit slowly, no matter how
poorly articulated this might be. The interdisciplinary scholar, trying to draw
connections across fields, will struggle to understand what scholars are

24
Newell (1998) reaches a similar conclusion. He notes that while all interdisciplinarians
recognize that disciplines are imperfect, most also feel that disciplines have important
contributions to make to interdisciplinary analysis. He urges those with a more critical
stance to appreciate that interdisciplinary analysis can and does transform disciplines
themselves.
25
Stanley Fish had criticized interdisciplinary maps of some or all of the scientific enterprise
for ignoring what goes on within disciplines (Moran, 2002, 146). The classifications
developed in this book embrace both what happens within and across disciplines.
22 Chapter 1

talking about. If agreement on classifications of phenomena, theories,

methods, and perspectives could be achieved, this common problem facing
interdisciplinary scholars (Salter and Hearn, 1996) would be greatly reduced.
Fish (1998) famously argued that interdisciplinarity was effectively
impossible, because any scholarly community needs a set of favored
theories, methods, and so on. An interdisciplinary community of scholars
would inevitably develop such standards, and become in time just another
discipline. While his argument runs against historical experience, the
question of scholarly standards is important. Chapter 5 will address in detail
the costs for all scholars of narrow standards. Interdisciplinarians can and
should appreciate the value of a shared conversation among specialized
researchers. They also see the need to draw connections across communities
of scholars; in this effort they cannot arbitrarily constrain themselves in
advance to look at only certain theories or methods (see chapter 8).
Some scholars argue for incommensurability. For example, it is
suggested that one cannot compare the Newtonian and Einsteinian
paradigms. Once relativity theory had triumphed, Newtonian laws were
viewed through the prism of relativity. Scholars who think they are
comparing the two are in fact dealing with a bastardised version of the
Newtonian paradigm. The argument can be applied to interdisciplinarity, to
suggest that conversation across scholarly communities is impossible, for
these can only be understood from within. Similar arguments, not
coincidentally, are often applied to cultural groups; this view flies in the face
of common experience that people can communicate across cultural
boundaries, and scientists can comprehend (much of) the internal workings
of cultures different from their own (Szostak, 2003). Likewise in science,
Galison (1996, 13-14) argues that conversation is always possible across
scholarly fields: “pidgin” languages may need to be constructed, but fruitful
conversation is widely observed. A Golden Mean conclusion – that
translation problems exist but are not impenetrable – suggests itself.
This book, in attempting to classify different types of science, must
assume that they are not completely incommensurable. It need not ignore
translation problems. Indeed, as noted above, one main benefit of the
classifications proffered in this volume is that they can reduce translation
difficulties. Bromne (2000, 119-29) builds on common ground theory in
psychology to argue that some common ground is required for
communication across disciplines, and that in particular common
terminology and accurate stereotypes of the other are conducive to
communication. A shared classification of types of science would provide a
common ground, common terminology, and encourage accurate stereotypes.
Chapter 2
CLASSIFYING PHENOMENA AND DATA

1. CLASSIFYING PHENOMENA

The value of classifying the phenomena that scientists study may seem
particularly obvious. If scientists do not comprehend the identity of the
phenomena they study, they are unlikely to proceed very far in understanding
how particular phenomena evolve or influence each other. “In the [natural]
sciences the development of ‘pure’ elements and homogenous, stable
phenomena was a key point in the emergence of modern chemistry and other
natural sciences …” (Whitley, 2000, 122). As Singleton and Strait note
(1999, 19-20), “science makes a great fuss about language; observations
must be precisely and reliably reported; terms must be carefully defined,
with clear referents; and the phenomena to which a scientific discipline
addresses itself must be organized and classified in a meaningful way.”
Sadly, though, while the need for classification may be obvious, the
difficulties inherent in the task have prevented most disciplines from
achieving “meaningful classification.” At the level of the modern scientific
enterprise as a whole, the task has not even been attempted. Among the
unfortunate side effects of this situation is the arbitrary nature of disciplinary
boundaries: without a logical classification of phenomena to guide them it is
hardly surprising that disciplines do not pursue a logical division of the
terrain. In some cases several disciplines may share a phenomenon (often
without paying heed to the efforts of other disciplines to comprehend it); in
other cases a phenomenon may be ignored by all.
The task of classifying all of the phenomena that scientists study must
seem particularly daunting. Surely there are thousands upon thousands of
such phenomena? While this is true, these phenomena can be organized
hierarchically within a small number of larger categories. Krieger (1997, 31-
24 Chapter 2

2) speaks in this respect of a general scientific commitment to the belief that

things are made up of constituent parts. This task is performed for human
science in Table 2.1. The starting point for this classification is an attempt to
divide the subject matter of human science into logical categories. These
categories must cope both with individual characteristics and with societal
characteristics. At the level of individuals two categories of phenomena can
be identified. The first is “genetic predisposition.” As a species, humans
share a gene pool that gives all a set of basic abilities, motivations, and
emotions. While this common gene pool guarantees a certain set of
characteristics that defines the species, differences in the precise genes that
individuals possess, in concert with differences in environment, serve to
guarantee that individuals differ from each other both physically and
psychologically. This yields a second category of “individual differences.”
All humans are necessarily part of a larger community, especially for the
first few years of life. That is, one of the shared genetic characteristics noted
above is that humans are born needing the help of others. Several distinct
categories of collective behavior can be identified:
x how humans interact with the non-human environment in order to create
(and distribute) food, shelter, and other items of practical utility: “the
economy”
x how humans interact with the non-human environment to create items
desired primarily for their aesthetic appeal rather than their utility: “art”.
Note that works of art, through their aesthetic appeal, may serve further
purposes, such as encouraging religious belief; such effects would be
captured in causal links. Art is often viewed as a subset of “culture”; it is
treated separately here because works of art, while they contain cultural
elements, are defined in terms of an aesthetic effect that transcends
cultural boundaries.
x how the various sub-groups of society interact: “social structure”. There
are always at least two types of sub-group, for the family is ubiquitous,
albeit in different forms, and genders have never yet been treated in
precisely the same way.
x how power is distributed and exercised: “politics”.
x It is obvious that hierarchical economic, social, and political structures
generate beliefs in the correctness of those structures, or at least attempts
are made by those at the top to do so. Such beliefs thus logically belong
to those categories. Yet societies have a host of religious beliefs,
customs, habits, and so on whose connection to these other realms is (at
least potentially) tenuous: these can be termed “culture”. Attitudes
toward all categories except economy, politics, and social structure are
thus part of culture. Following common usage, languages are treated
2. Classifying Phenomena and Data 25

here as a subset of culture. The precise definition of culture becomes

clearer as the category is unpacked.
x Humans also develop knowledge of how they can best manipulate the
non-human environment to suit various ends: “science and technology”.
x The list may seem complete, but humanity must also perpetuate itself as a
species, and thus “population” must be considered. Ability to reproduce
depends in turn on ability to survive. The related matter of “health” must
also, then, be considered; this deserves more attention than it receives
from human scientists.
x The “non-human environment” has been mentioned more than once
above. Since it both shapes and is shaped by humanity, it deserves its
own role in the classification as another category. This category (and that
of genetic predisposition) would provide a link between this classification
and a classification of natural science phenomena.

This list of ten – eleven if health and population are treated separately –
logically distinct categories is also arguably exhaustive, for the ten categories
seem to subsume all human activities and characteristics. As Table 2.1
illustrates, it is straightforward to place all subsidiary phenomena within
these categories. In several cases, care must be taken to establish the
boundaries between categories: as noted above, art can be distinguished
from culture by defining art as that which has an aesthetic appeal not limited
to members of particular groups. These precise boundaries become clearer
as the categories are further disaggregated.
Table 2.1 reflects an extensive exercise in disaggeregation undertaken in
Szostak (2000, 2003). Both deduction and induction were used to develop
lists of second and then third-order phenomena (Note that these titles refer
only to the level of aggregation and imply no value judgment): induction in
the sense of finding a place in the classification for all phenomena discussed
in a wide variety of works consulted, and deduction in the sense of thinking
about how phenomena could logically be broken into their constituent parts.
The result should be nearly exhaustive, though some phenomena may have
been missed.1 It seems a reasonable conjecture that all phenomena that affect
human lives will have been noticed by someone sometime: the scholarly
community can thus aspire to an exhaustive list, at least at higher levels of
aggregation: sub-atomic particles are an obvious example of important
lower-level phenomena only observable with advanced scientific

1
One entry has, indeed, been added to the table. Though it was noted above that culture
will include attitudes toward most other categories, an entry for “attitudes toward healing”
was missing from previous versions of Table 2.1. This oversight (which was pointed out
to me) was especially egregious since Wissler, in his efforts decades ago to classify
phenomena, had grouped healing with religion.
26 Chapter 2

methods. Importantly, the table is inherently flexible so that new phenomena

can be added. As for the organization of the table, this too is flexible: if it
were found empirically, for example, that “language” was more strongly
related to phenomena in a category other than “culture” the table could be
adjusted to reflect that. More often than not, though, there was considerable
scholarly consensus on how a particular phenomenon should be unpacked.
Scholars of personality agree that there are a handful of key personality
dimensions, and that these can be unpacked into traits, though they disagree
about precise boundaries at both levels. Scholars of social division agree on
most but not quite all of these. While scholars of culture celebrate the
diversity of definitions of that word, few would doubt that values or stories
or religion belong there. For the purposes of this book it is not necessary to
review the detailed makeup of the table, nor to develop a natural science
counterpart. It is important to appreciate that such a classification is feasible:
where scholarly disagreement exists about where a phenomenon best fits,
this can potentially be solved empirically by establishing what other
phenomena it is most tightly linked to.
While Table 2.1 undoubtedly appears daunting at first, it is organized in
terms of a manageably small set of categories and second-level phenomena.
A student or scholar with some familiarity with this table should be able to
quickly locate any particular phenomenon within it. While a similar exercise
has not been performed for natural science, there is no reason to doubt that
this could be done. Indeed, disciplinary boundaries in natural science reflect
different levels of aggregation to a much greater degree than is the case in
human science (Szostak, 2003). This means that it should be much easier to
draw on disciplines in designing a hierarchical list of the phenomena of
interest to natural scientists. Key elements of such a list can clearly be
identified: lists of subatomic particles, the periodic table of chemical
elements, biological taxonomies of species, classifications of soil and rock
types (It is notable that those elements which first leap to mind are each
associated with a particular discipline). As with human science, some of
these elements reflect a great deal of scholarly consensus. In other cases,
there is still room for dispute. Chemists and philosophers debate how to
determine when two substances are the same or different (Bhushan and
Rosenfeld, 2000, 8). Biologists have long debated their very ability to define
species, given that evolutionary theory implies that these change through
time. The answer seems to lie in the identification of key characteristics of a
particular species; if it evolves too much it then becomes a new species (see
Mahner and Bunge, 1997, ch. 7, Hull and Ruse, 1998, part V). As noted
above, there is a greater risk of the natural scientific community not having
yet identified all phenomena that should be of scientific interest. Dogan and
2. Classifying Phenomena and Data 27

Table 2-1. The Phenomena of Interest to Human Scientists

Categories Second Level Phenomena Third Level Phenomena
Genetic Abilities Consciousness, subconsciousness,
Predisposition vocalization, perception (five
senses), decision-making,
toolmaking, learning, other
physical attributes (locomotion,
eating, etc.)
Motivations Food, clothing, shelter, safety, sex,
betterment, aggression, altruism,
fairness, identification with group
Emotions Love, anger, fear, jealousy, guilt,
empathy, anxiety, fatigue, humor,
joy, grief, disgust, aesthetic
sense, emotional display
Time Preference

Individual (Abilities:)
Differences
Physical Abilities Speed, strength, endurance
Physical Appearance Height, weight, symmetry
Energy Level Physical, mental
Intelligences Musical, spatial, mathematical,
verbal, kinesthetic, interpersonal
(Personality:)

Emotionality Contentment, composure, vs.

(Stable/Moody) anxiety, self-pity
Conscientiousness Thoroughness, precision, foresight,
organization, perseverance vs.
carelessness, disorderly, frivolous
Affection Sympathetic, appreciative, kind,
(Selfish/agreeable) generous, vs. cruel, quarrelsome,
faultfinding
Intellectual Orientation Openness, imagination, curiosity,
(Holistic/analytical) sensitivity vs. closemindedness
Other dimensions? Dominant/submissive,
strong/weak, in/dependant, humor,
aggression future/present oriented,
happiness
Disorders? Schizophrenia, psychoticism, ...?

Sexual Orientation

Schemas View of self, others, causal

relationships
28 Chapter 2

Categories Second Level Phenomena Third Level Phenomena

Interpersonal Relationships Parent/child, sibling, employee/r,
romance, friendship, casual

Economy Total Output Price level, unemployment,

individual goods and services
Income Distribution
Economic Ideology
Economic Institutions Ownership, production, exchange,
trade, finance, labor relations,
organizations

Art Non-reproducible Painting, sculpture, architecture,

prose, poetry
Reproducible Theater, film, photography, music,
Dance

Politics Political Institutions Decision-making systems, rules,

organizations
Political Ideology
Nationalism
Public Opinion Issues (various)

Crime Versus Persons/Property

Culture Languages By Descent?

Religions Providence, revelation, salvation,
miracles, doctrine
Stories Myths, fairy tales, legends, family
sagas, fables, jokes and riddles
Expressions of culture Rituals, dance, song, cuisine, attire,
ornamentation of buildings, games
Values (Goals:) Ambition, optimism, attitudes to
wealth, power, prestige, beauty,
honor, recognition, love,
friendship, sex, incest, marriage,
time preference, physical and
psychological wellbeing
(Means:) Honesty, ethics, righteousness,
fate?, work valued intrinsically,
violence, vengeance, curiosity,
innovation, nature, healing
(Community:) Identity, family versus community,
openness to outsiders, trust,
egalitarianism, attitude to young
and old , responsibility,
authoritarianism, respect for
individuals
2. Classifying Phenomena and Data 29

Categories Second Level Phenomena Third Level Phenomena

(Everyday Norms:) Courtesy, manners, proxemics,
tidiness, cleanliness, punctuality,
conversational rules, locomotion
rules, tipping
Social Structure Gender
Family Types, Kinship Nuclear, extended, single parent
Classes (various typologies) Occupations (various)
Ethnic/racial Divisions
Social Ideology

Technology and Fields (various) Innovations (various)

Science
Recognizing the Problem
Setting the Stage
Act of Insight
Critical Revision
Diffusion/transmission Communication, adoption

Health Nutrition Diverse nutritional needs

Disease Viral, bacterial, environmental
Population Fertility Fecundity, deviation from
maximum
Mortality Causes of death (various)
Migration Distance,international?,temporary?
Age Distribution

Non-Human Soil Soil Types (various)

Environment Topography Land forms (various)
Climate Climate Patterns (various)
Flora Species (various)
Fauna Species (various)
Resource Availability Various Resources
Water Availability
Natural Disasters Flood, tornado, hurricane,
earthquake, volcano
Day and Night
Transport Infrastructure Mode (various)
Built Environments Offices, houses, fences, etc
Population Density

Pahre (1990) argue that the discovery of new phenomena has been a key
component of advance in natural science in recent decades: they refer to
quarks, quasars, and DNA (and the greenhouse effect, which is a causal
link); for the most part, these new phenomena result from the unpacking of
previously recognized phenomena. It is thus still quite useful to classify
known phenomena, and add new phenomena as these are discovered.
30 Chapter 2

In chapter 1, it was argued that both specialization and integration are

essential components of the scientific enterprise. In identifying the
phenomena of interest to scientists, and placing these in a logical
classification, specialized scientists are guided to precisely identify their
subject area, and closely related phenomena. Yet causal links exist between
virtually any pair of these phenomena:2 the classification of phenomena thus
also aids integrative work by identifying a large but finite set of causal links
that might be explored. Note that boundary work, the linking of two or more
areas of scientific specialization, will proceed best if the boundaries
themselves are precisely defined.

1.1 Defining Terms

Alfred Kuhn’s (1974) definition of the elements of his system could serve
as a definition of phenomena as used here: any identifiable entities, concrete
or abstract, individual or collective.3 Kuhn speaks, though, of events.
Individual events would not be considered as phenomena here, for
phenomena are defined as ongoing, indeed timeless, characteristics of human
society. Individuals may differ in terms of personality, but a set of
personality dimensions are always with us. Genetic evolution may change
human aggressiveness, but attitude to aggression remains a phenomenon
worthy of study. Economic and political institutions evolve, and differ
across societies, but the classification of possible types of institution need not
change to capture this diversity (societies lacking a particular institution fit as
comfortably within the classification as those who differ in the details of
institutional structure).4 “The classes that are important in science are

2
Dogan (1996, 296) hypothesizes that if one developed a 12 by 12 grid of social science
disciplines, virtually every cell would contain some hybrid field. Certainly, if one were to
develop a grid of categories, or even second-level phenomena, one would find it
straightforward to identify causal linkages in most/all cells that deserved sustained
scholarly investigation.
3
The word phenomena has been given precise and conflicting definitions by philosophers
over the years. For Husserl, it referred to the essence of things, or “things in themselves”
which humans with limited sensual apparatus could only approach understanding of (see
below). Conversely, philosophers both before and after him have used the word to refer
instead to sensory perceptions of things. McKeon (1994) notes in this regard that the
Greek base of the word phenomenon is the same as the Latin base of the word appearance.
In this book, individual phenomena will be defined as precisely as possible, but it is
appreciated that human understanding is everywhere imperfect; the use of the word
phenomena here is thus distinct from either of these philosophical usages.
4
As society becomes more complex, scientists may become aware of new phenomena that
deserve to be classified. Note, though, that logically the classification is designed so that it
encompasses societies at all yet-realized levels of complexity. McKeon (1994, 330)
suggests that in science, as in philosophy, definitions of the phenomena being studied
2. Classifying Phenomena and Data 31

those that are spatio-temporally unrestricted, so that they can function in

laws [theories] of nature” (Hull, 1998, 297). Neither the French Revolution
nor the Mona Lisa are phenomena; still Table 2.1 guides scholars to discuss
both the causes and effects of changes in political structure and acts of
artistic genius, and the value of the classification can be tested by how well it
allows scientists to understand particular events.5
The most precise definition of the word “phenomena” is provided by
Table 2.1 (and the discussion of natural science phenomena), wherein a
lengthy list of phenomena is provided. What characteristics do these
hundreds of phenomena share? They are, as noted above, enduring aspects of
human existence. They are also of interest to scientists and susceptible of
scientific description and explanation. Indeed, Table 2.1 comprises (ideally)
the complete set of enduring aspects of human existence that are (or should
be) of interest to human scientists.
This definition of phenomena has important implications for the
definition of causal link. In examining the link between two phenomena,
scientists ask either how a particular realization within one phenomenon
effects the realization within another, or how a change in the one realization
induces a change in the other. In looking at the link between “Attitude
toward honesty” and “Economic output”, scientists could wonder whether or
not – and if so, why – greater honesty tends to facilitate economic
transactions and thus a higher rate of growth in output.6 It is thus not the
phenomena themselves that change – they are defined so that they do not
change – but rather particular realizations of those phenomena that change.
Scholars in different disciplines interpret the word causal differently.7
For some, to say that a change in X causes a change in Y implies that only a

change through time. It would be better to speak of changes in realizations of phenomena

within an unchanging classification.
5
Elster (1989, 3) notes that it is more fundamental to explain facts (that is, realizations in
terms of particular phenomena) in terms of (facts and) events than the reverse.
6
A distinction can be made between what Barber (1993) calls changes in type versus
changes within type. Another element of culture is standards of personal attire. One type
of such standard is modern western fashion, wherein styles are expected to change. If this
type were to give way to a relatively unchanging standard of attire, this would be a change
in type. Yearly changes in styles of clothing are instead changes within type. These are
usually of less interest to the human scientist, though they may still be at least in part
causally related to changing artistic tastes or gender roles or a host of other phenomena.
7
Nicolescu (2002, 11) complains that modern science ignores the formal and final causes
posited by Aristotle, in order to focus on local causation. He worries that scientists thus
encouraged an unfortunate belief that societies could be transformed overnight, without
regard to their history. However, history does not in fact act causally at a distance, but
locally through its effects on culture, institutions and other phenomena. This book will
largely follow scientific practice in emphasizing local causation. A final cause was “that
for the sake of which” something was done (Velkley, 2000, 314-7); the place of functional
causation is addressed in chapter 3.
32 Chapter 2

change in X does so. Yet one key result of Szostak (2003) is that all
phenomena are influenced by many others. As the word influence captures
this idea of multiple causal links better than the word cause, it is generally
used here, though causation will from time to time be spoken of. This is
never meant to imply that only one phenomenon influences any other. The
phrase causal link has been retained because there is no obvious alternative,
and it is poor practice to create jargon unnecessarily. Both cause or
influence, then, refer merely to the various ways in which a realization of one
phenomenon induces a change in the realization of another.
Some scholars feel that it is inappropriate for human scientists to study
causation; they should instead deal only with interpretation of the meaning
individuals attach to situations. In both chapters 3 and 4 it will be shown that
such a distinction is both unnecessary and inadvisable. Interpretation can be
embraced without rejecting causation. Moreover interpretivists in practice
make causal arguments. We can thus join Singleton and Strait (1999, 24-6)
and many others in claiming that scientific understanding necessarily
involves the study of causal links.8
Embracing the study of causal links does not imply a lack of appreciation
of the multiple causation and feedback effects that characterize the world.
That is, phenomena often exert influence in combination with others, and the
phenomena affected by a causal link often exert some reactive influence.
These types of causation can be captured readily: imagine arrows joining and
moving in multiple directions among sets of phenomena in Table 2.1.9

1.2 Advantages of a Classification of Phenomena

The many advantages for both research and teaching in human science of
a classification of phenomena such as that outlined in Table 2.1 were
discussed at length in Szostak (2000, 2003). They are summarized here:
x It can change the “gestalt” of teachers and researchers such that they
more fully appreciate that the questions they concentrate upon are parts of
a much more complex “big picture”. It is all too easy for scholars to

8
Potter (2000, ch.8) celebrates realist philosophy. If one were to ask what the world would
have to look like for science to be successful, it would have to consist of things
[phenomena] that exert influences on each other.
9
Stump (1996, 449) is critical of science studies for looking only at socio-cultural
influences on science, and ignoring feedbacks. Dupr- (2001) is critical of both
evolutionary psychology and neoclassical economics for tending to assume that the
phenomena they emphasize exert exclusive influence on many others (and notes that
simplistic policy prescriptions follow in both cases). Similar critiques can be made of
other fields.
2. Classifying Phenomena and Data 33

denigrate through ignorance links other than those they study, even when
these are closely related to their own research. Note in this regard
Bunge’s (1998, 17) feeling that perhaps the major failing of human
science theory is that it tends to involve too few relevant variables: he
appreciates that it is important not only to identify those phenomena
which exert an important influence, but those which must be held
constant for a posited causal link to operate (see chapter 3).10
x It should encourage a more balanced approach to research and teaching,
and also to public policy advice. At present some causal links are
virtually ignored by scholars, despite being of considerable intrinsic
importance. In particular, policy advice tends to reflect disciplinary
biases: economists focus on economic phenomena when judging trade
deals, and political scientists on political phenomena. Scholars providing
policy advice all too often ignore relevant links with respect to which
they lack expertise. Decision-makers are thus led to believe that they are
being given a more complete analysis than is the case.
x It provides a powerful illustration of the interconnected nature of human
science, and thus of the danger of barriers to interdisciplinary
communication. Take crime as an example: a moment’s reflection
indicates that a host of factors – genetic, individual differences, political,
social, cultural, and more – will influence crime, and thus that a
comprehensive analysis of crime must span many disciplines. The
classification points to (at least) the redefinition of the scope of existing
disciplines.
x By identifying which links researchers are investigating, it can allow
teachers and researchers to recognize when research in other disciplines
deals with the same link. The research performed for Szostak (2003)
uncovered many examples of scholars being completely unaware of
related research in other disciplines. Yet there are also times when

10
Kincaid (1996, 64-9) notes that scientists commonly make ceteris parabis assumptions:
they assume that all relevant variables outside those being studied do not change. In the
real world, especially in human science, this assumption is almost always violated. So
what kind of explanation do scientists provide? They explain certain causal links [Kincaid
speaks of “tendencies”] in isolation from those they regularly interact with. Can
confirmation be provided for a posited causal link that exists within a complex web of such
links? Yes. There may be times when the ceteris parabis world is approximated (naturally
or experimentally), the effect may be visible anyway (especially as the ceteris parabis
world is approached), it may be possible to identify other key influences and jointly
establish their relative importance, indirect evidence may be gained if the posited link
complements others that evidence exists for, and it may be possible to identify
intermediate variables. Blalock (1969), noting that it is infeasible to examine large
numbers of links simultaneously, recommended a block-recursive approach where sets of
links were analyzed. The danger here would be that scientists exaggerate the importance
of links within such blocks and downplay the importance of links between blocks.
34 Chapter 2

scholars engage in a debate without realizing that they are in fact talking
about different links. Economists may thus be befuddled by opposition
by political scientists to a trade deal that enhances economic outcomes;
an appreciation that political scientists stress links to, say, nationalism,
would encourage a more constructive exchange.
x The table can also provide a superior cataloguing system for human
science insights, and thus reduce the scholarly tendency to reinvent the
wheel (see chapter 7).
x It provides very precise definitions of phenomena in terms of the
phenomena into which they are unpacked. For example, culture equals
precisely the phenomena listed under “culture” in Table 2.1. It can thus
provide a solution to one of the greatest barriers to (especially
interdisciplinary) communication: the fact that different groups of
scholars tend to define the same word differently (and are often vague
about these definitions as well). At present, any reference to “culture” by
a scholar is necessarily ambiguous, given literally hundreds of competing
definitions. Scholars will be encouraged to specify which cultural
phenomena are particularly germane to their studies, rather than
appealing to the vague aggregate of culture.
x It provides a positive response to postmodern concerns with complexity,
while avoiding postmodern despair that enhanced understanding is
unattainable. A glance at Table 2.1 brings home powerfully the simple
point that human science must engage thousands of links among
thousands of phenomena. Yet it also suggests that the number of links is
finite and that understanding can be accumulated link by link.
x Grand theories can be mapped onto the list of phenomena, to show that
each is better suited to some links than others (see chapter 3). Scientists
should not leap from a conclusion that Marxian theory provides insight
into one link to an assumption that it provides insight into all; nor should
its failure to illuminate a particular link be taken as cause to reject the
theory for all links. The classification serves as a powerful antidote to the
siren song of grand theory: these are attractive because humans yearn for
a means of simplifying complexity. The classification of phenomena
allows scientists to cope with complexity while embracing theoretical
(and methodological) flexibility.

The previous bullets dealt primarily with the advantages of the

classification for research. The next bullets focus instead on the value of the
classification for teaching:
x Most centrally, it provides a handy means for students to connect
material both within courses and across diverse courses, by comparing
the links covered in different courses or parts of courses. In the same
2. Classifying Phenomena and Data 35

fashion it provides a simple means for administrators of interdisciplinary

programs to measure the breadth of their coursework. At present, students
are often required to take courses across a range of disciplines, without
being given any sort of “map” that could facilitate their drawing
connections across diverse material.11 Graff (1992) worries that students
compartmentalize knowledge course by course, and are thus often
blissfully unaware that different instructors have reached different
conclusions about the same topic. If instructors instead regularly referred
to a common list of phenomena, students would be forced to confront
these intellectual conflicts (and classifications of theory and method
would help them to identify the source of these). Even when instructors
address different topics, students will be better able to recall this diverse
information if it is readily organized into a coherent whole.
x Students can also easily ascertain which links received little or no
attention in their formal studies. This should both encourage and
facilitate lifelong learning. Instructors often give students a misleading
impression of the completeness of their studies – since they tend naturally
to emphasize phenomena and links that they understand well.
x The classification of phenomena facilitates the acquisition not only of
content but of various skills identified by interdisciplinary scholars (see
Klein, 1996). It provides a mechanism for coping with complexity. By
highlighting the limitations of grand theory, and the danger of ignoring
relevant links, it encourages an appreciation of diverse viewpoints,
awareness of bias, and (especially) suspicion of authority. It guides
students to recognize and overcome ignorance. It guides students to
analyze relationships and assess their importance. It helps students to
appreciate the need for – and achieve – a balance between depth and
breadth. And it is well suited to class presentations, discussions, and
web-based pedagogy.
x The classification provides a powerful response to recent criticisms of the
Liberal Arts. It suggests that there is a core of knowledge that should be
imparted to students. It thus provides coherence to undergraduate
education (and even greater coherence can be achieved with overviews of
theory, method, and philosophical perspective as well). Yet it
simultaneously puts issues of diversity front and centre. It shows that

11
Table 2.1, like the alphabet, can seem daunting on one’s first acquaintance with it. But it
soon proves to be an extremely handy device for categorizing human science
understanding. Like a phone book, one need not memorize every entry in order to be able
to utilize it. Students should not be asked or expected to memorize the hundreds of entries,
but should over time gain an appreciation of the ten major categories and some familiarity
with at least second-level phenomena. If instructors, when introducing new topics, remind
students of where these fit schematically, the student should fairly effortlessly master the
table along with other material.
36 Chapter 2

phenomena within both the Social Structure (see Szostak, 2001) and
Culture (see Szostak, 2003) categories are causally related (in both
directions) to phenomena in all other categories. It thus simultaneously
provides responses to critiques from the “right” and “left” (see chapter 8,
Gaff, 1997). The classification also provides a response to those who
worry that universities and colleges do not adequately facilitate lifelong
learning (see above).

Table 2.1 is useful for the study of any subject in human science. Szostak
(2003) took culture as its focus. That work hopefully established both the
validity and value of the classification, by showing that it can easily organize
a complex literature into discussions of particular links (and indeed can
easily cope with multiple causation and feedback effects). Six broad lessons
emerged for the study of culture, which apply by extension to other
categories:
x Analysis is best performed in terms of subsidiary phenomena, rather than
at the vague level of culture itself. To understand the role of culture in
human societies, hundreds (thousands) of causal links must be
investigated.
x Attempts to unify this diverse literature in terms of a single theory or
method must necessarily severely oversimplify the complexity of culture.
x There are strong links both to and from cultural phenomena. It is thus a
mistake to view culture as either simply the cause or result of phenomena
in other categories (This means, among other things, that it is a mistake to
see either individuals or societal aggregates as uniquely cause or effect;
see Bunge, 1998, 64).
x The siren song of grand theory (or method) is hard to resist in the absence
of some other mechanism for integrating insights into diverse linkages.
Prus (1996), for example, argues that culture can only be understood
through the single lens of his favored theory and method. The
classification of phenomena serves as an antidote to grand theory by
providing a means of organizing diverse causal analyses into a coherent
whole.
x The classification serves an invaluable pedagogical role. The student can
be exposed to a wide range of links (The book treated those generally
considered most important plus many that receive less or sometimes no
attention). Instructors will naturally focus on a subset of these, without
giving students a misleading sense of completeness. Students can be
encouraged to make presentations, write short papers, and so on, on links
not covered by the instructor.
x Notably the classification guides scholars to avoid three problems
identified by Cornwell and Stoddard (1999) as commonly afflicting
2. Classifying Phenomena and Data 37

interdisciplinary courses on culture: treating cultures as monolithic rather

than as an aggregate of diverse attitudes and practices, treating “our”
culture as normal, and ignoring the internal divisions that exist in any
society. Cornwell and Stoddard indeed recommend the exploration of
causal links.

1.3 Applying the 5W Questions to Phenomena

The value of the 5W questions tends to be inversely related to our ability

to first provide a unidimensional classification of a particular key element of
science. Since an exhaustive and coherent list of phenomena can be
provided, the 5W typology is less critical here than it is with respect to
theory. This is fortunate, for the 5W questions do not point toward a
typology but serve instead to provide several insights into the classification.
Who do scientists study? As will be seen in chapter 3, science should
embrace non-intentional and intentional agents, whether individuals,
relationships, or groups. A glance at Table 2.1 shows that it embraces
phenomena that represent each type of agency. Two categories capture
individual-level phenomena. Relationships are an important component of
one of these. The non-human environment category captures all types of
non-intentional agents, plus animals. The seven societal-level categories
capture a mix of group-level and non-intentional agents.
At a particular point in time, a scientist inevitably studies some subset of
the phenomena in Table 2.1. The scientific enterprise as a whole should
nevertheless embrace all of these. Thus one important role for the
classification is to identify phenomena or causal links that at present receive
little or no scientific attention, so that research can be encouraged.
It should be stressed here that scientists only examine phenomena and
causal links. The vast bulk of scientific research deals with the latter:
scientists ask how, if, or how much one or more phenomena influence one or
more others. Some scientists instead seek to identify, define, or measure
realizations of individual phenomena. To be sure, some scientists disdain
words such as “causal” or even “influence” but these scholars nevertheless
make causal arguments (see chapter 4). And some scientists are so vague
about what they are studying that it is hard to identify recognizable
phenomena in their efforts; chapters 2, 3 and 5 will discuss whether such
efforts are best viewed as bad science or non-science.
What exactly are these phenomena? As stated above, they are enduring
characteristics of human societies.12 Some scientists and philosophers would

12
That, at least, was the intention. In some cases, such as the unpacking of occupations, it
was particularly difficult to identify universals. But generally this was not problematic.
38 Chapter 2

doubt the very existence of reality,13 however, and would suggest that any
phenomenon is merely a social construct (or perhaps a scientific construct).
That is, there was no such thing as “attitude toward ambition” or “income
distribution” or “personality dimensions” or even “climate” out there until
humans thought of such things. Some who hold to this view doubt the very
possibility of science: they suspect that every individual creates their own
reality and thus that there is no truth for scientists to attempt to uncover (see
chapter 5). Others are less pessimistic, still embrace the study of phenomena
and causal links, and ask only that scientists be particularly careful to avoid
scientific bias (again, see chapter 5). This book relies on a Golden Mean
perspective that there is an external reality, but limits to human abilities to
perceive and comprehend this.14
Seale (1999, 88) notes that those who deny the existence of an external
reality could argue that distinctions between data, theory, and research [or
method or phenomena] are just elements of a language game;15 “I would
argue that at the very least it is a game that we must choose whether or not to
play.” Seale goes on to stress that if an external reality is accepted, these
distinctions are both valid and important.
A list of phenomena such as that in Table 2.1 inevitably raises the
question of reductionism. However, Szostak (2003) found that explanation
should not, and cannot, always be reduced to the lowest possible level of
aggregation, for higher-level phenomena exhibit emergent properties.
Bhushan and Rosenfeld (2000, 5) make a similar argument that chemistry
cannot be reduced to physics, as do Mahner and Bunge (1997, 197) with
respect to biology and physics.
What do phenomena do? Wherever possible, phenomena in Table 2.1
have been defined in terms of the role they play in human societies. Thus,
each phenomenon is functionally unique. This in turn is why subsidiary

13
“It is one of my premises that most people operate as though reality does exist. It is only
academics who make a living arguing the opposite” (Oakley, 2000, 20): Oakley goes on to
suggest that even these academic sceptics generally treat the world as real as they perform
their research.
14
Weingart (2000) suggests that most interdisciplinarians assume an external reality that
interdisciplinary analysis can enhance understanding of. He worries, though, that any
organization of knowledge is inherently arbitrary: environmental studies may be shaped
more by opportunities for funding than any external reality. While the boundaries of
disciplines and interdisciplinary fields are arbitrary, this book is based on the premise that
objective classifications of phenomena, data, theory, and method are possible.
15
Woolgar (1988) argues that there are no objects before our representation of them; this
argument would seem to fly in the face of everyday experience. Prus (1996) does not
doubt that there are real phenomena “out there,” but argues that these resist efforts to
comprehend them. Understanding is thus subjective, and each individual’s perceptions are
shaped by their cultural environment. Nevertheless, a considerable degree of consensus
can be achieved on a list of phenomena such as that in Table 2.1.
2. Classifying Phenomena and Data 39

phenomena (once properly defined) belong in only one place in the

classification. Phenomena exist in a web of influence whereby each
phenomenon influences (virtually) every other phenomenon. An
understanding of what a phenomenon does must thus embrace both the
influences it exerts on others and the influences they exert on it.
The goal of science should be to understand what happens along every
causal link. As shall be seen in chapter 3, understanding of what happens
along one link should inform understanding of what happens along others.
Nevertheless, there is a danger in only studying systems of links, for theories
(and especially theoretical models; see Dupré, 2001, 13) of systems tend
naturally to exaggerate the importance of links within the system and ignore
links to phenomena outside the system. In terms of method, McKim (1997,
10) bemoans the fact that multiple regressions tend to ignore the particular
links that exist between every pair of variables involved in the equations.
Exploration of each pair-wise relationship is difficult, but is becoming easier
with new statistical packages. “However, the rationale for thinking of the
problem in these terms should be clear. Each positive or negative partial
correlation coefficient should mark a path of direct causal influence between
two variables” (McKim, 10).
Why do scientists study phenomena? There are severe limits to human
ability to understand the world holistically. In both their everyday lives and
in science, people must break the whole into parts and attempt to understand
how these interact. So understanding requires that both phenomena and
causal links be grappled with.
How do scientists study phenomena? Humans only have access to their
perceptions of phenomena. As thousands of philosophers across a wide
range of traditions have noted, these perceptions are not identical to the
essence of the phenomena, the “thing in itself”. [The philosophical debate
often focused on physical objects, but can be extended to all phenomena, at
least for present purposes.] In other words, humans must worry about the
accuracy of their perceptions. Philosophers have disagreed as to how
trustworthy these may be. It might be anticipated that the problem is greater
with respect to some phenomena than others. This is especially so, given
that the classification includes many phenomena – such as attitudes – that are
not directly observable.
Scientists do not examine phenomena but data. That is, they study
evidence that is taken to represent (some aspects of) certain phenomena, or
perhaps certain links among these. Philosophers worry about the degree of
correspondence between data and the phenomena referred to in a particular
theory. In drawing causal inferences from data, the scientist must use their
judgment as to how well the data truly reflects the phenomena or links in
question. Some phenomena – such as per capita income – are fairly easily
40 Chapter 2

and accurately operationalized (though there are important disputes about

what to include in even this calculation), while others, such as various
attitudes, are harder. Some phenomena (say, stress) may be operationalized
by a quite different datum (corticosterone level in the blood). A continuum of
operationalizability can be conceived. “Only low-level empirical
generalizations, such as “all birds have feathers” can be directly confronted
with empirical data;” all others require “independent hypotheses” that
maintain that a particular datum represent a particular phenomenon or causal
link (Mahner and Bunge, 1997, 95). Note, though, that the development of
new scientific techniques may alter the relative operationalizability of
phenomena. Advances in neurological analysis, for example, may allow
scientists to “see” a person’s thoughts. Bhushan and Rosenfeld (2000, 5)
celebrate the ability of the scanning tunneling microscope to “see” molecules
and molecular events. Rothbart (2000, 81) reviews arguments that the
readouts from new instruments can be viewed as direct observations of data,
or alternatively as entirely constructed by the researchers involved, and
reaches a Golden Mean conclusion that instruments provide scientists with
improved but still imperfect and subject-to-bias information (see also
Bechtel, 2000).16 Even advances in theory may alter operationability: it was
difficult to observe diffraction before wave theory told scientists what to look
for (Gower, 1997, 241). Of course, there is a danger that data dependent on
theory provide biased evidence toward that theory; data of this type often
prove invariant to theoretical change, however (Mahner and Bunge, 1997,
74).
Where and When? These two questions guide scientists to remember that
all causal links take place in time and space. Scientists should be careful not
to casually assume that a link observed in one time and place could occur in
the same way in a different time and place. In other words they should be
careful about generalizability (see chapter 3). That is, scientists should be
interested in the range of realizations of all other phenomena which
allow/facilitate a particular causal link (a point made by Ragin, 2000, 108).
Chemists habitually indicate the range of temperature and pressure over
which a particular reaction will occur; social scientists are much less likely to
identify the often more diverse set of circumstances in which a posited causal
relationship will hold.

16
Baird (2000) notes that instruments are often hailed for “objectivity” because they produce
quantitative readouts. Judgment is still required as to whether readouts reflect reality.
This narrower sense of objectivity should not be confused with the broader and ubiquitous
need for scientific judgment. Baird worries that the advent of sophisticated instruments
has contributed to the devaluation of the importance of scientific judgment.
2. Classifying Phenomena and Data 41

1.4 A Note on Concepts

One inductive check on the exhaustiveness of the broadest

classification in this book – the equation in chapter 1 of types of science with
types of phenomena, data, theory, method, and practice – is to look at
attempts to define the essence of scientific disciplines in general. As
summarized in Szostak (2003a), both Klein (1996) and Salter and Hearn
(1996) generally identify disciplines under five broad headings: subject
matter [that is, phenomena], favored theories, favored methods [which can
included favored data sources], disciplinary worldview, and disciplinary
“rules of the game” for evaluating research. The last two of these will be
discussed when scientific practice is investigated in chapter 5. Klein often
refers, though, to “concepts”, and Dogan (1996, 296) defines disciplines in
terms of “concepts, methods, theories, and praxis.” What are concepts? In
Wittgensteinian fashion, Dogan (1996) provides a few examples. Many of
these, such as culture, are clearly phenomena. Others, such as
authoritarianism, are realizations of phenomena (in this case, political
decision-making systems). Some concepts, such as ergativity of language,
reflect efforts at unpacking. Still others – revolution, lag, convergence –
describe processes of change within or between phenomena. All of the
examples provided in Dogan (1996), and many more, can thus be precisely
defined in terms of phenomena or causal links. Often, the classification of
phenomena would clarify what a particular concept means.
Halperin (1997, 71-2) argues that scientists must grapple with six types
of relationships between concepts. One of these, “X causes Y,” would be
reflected in causal links. Another, “X is part of Y,” would be captured in the
unpacking of phenomena. Two more, “X is a characteristic of Y,” and “X is
an example of Y,” would be important in the definition of particular
phenomena and the characterization of causal links. A fifth, “X is like Y,”
might capture certain similarities among phenomena – if Z is unpacked into
X and Y they will share certain similarities but also important differences –
but more commonly refers to analogies that might be drawn between
different causal links; this issue will be addressed when generalizability of
theory is discussed in chapter 3. Finally, “X is evidence of Y,” refers to the
relationship between data and phenomena/links (see above). Note how
defining these six relationships in terms of phenomena, causal links, data,
and theories helps to clarify what they entail.17
17
Similarly Meltzoff (1998, ch.2) argues that research will be weak unless research
questions are framed clearly. He speaks first of existence questions: does X exist?
Second, what is X like, and how should it be classified? Third, what are the components
of X? These three questions underpin the list of phenomena in Table 2.1. He also
identifies three types of causal question: does X affect Y; is this effect greater than Z’s
42 Chapter 2

There are, though, other concepts. In chapter 3, Comte’s concept of an

“advanced theological society” will be criticized. This, it would seem, is a
description of realizations of diverse cultural, social, political, and other
phenomena. Unless these realizations are carefully specified, the scientist
cannot know exactly when such a society is encountered. Unless the
scientific community attempts to understand the links among the relevant
phenomena, scientists will be unable to understand how such societies
emerge or disappear. Unless links with other phenomena are delineated, the
causes and effects of “advanced theological society” will remain obscure.
Since all of these paths of exploration involve breaking Comte’s concept into
its constituent parts, it is an open question as to whether the concept itself is
useful. More importantly, it is only potentially useful if it is defined
carefully in terms of realizations of phenomena (and perhaps theorized links
between these). Similar remarks could be made about more modern
concepts such as “patriarchy”: this term can only be useful scientifically if it
is carefully specified in terms of realizations of phenomena. If patriarchy is
defined as a particular set of cultural, social, political, and other outcomes,
scientists can then investigate the links within and beyond these phenomena.
Dogan and Pahre (1990, 123-8) note that while such concepts may stimulate
interesting questions, their ambiguity hinders the scientific conversation;18
they hail the efforts of Sartori (1984) and others to clarify scientific concepts,
while noting that such efforts have had little success.
While no scholar has attempted an exhaustive survey of scholarly
concepts, a handful of authors have identified key concepts within particular
scholarly areas. Thus Sandler (2001) outlines “economic concepts for the
social sciences”: some of these are phenomena (institutions, property rights),
and some are theoretical terms (game theory, general equilibrium theory,
rational expectations), but most deal with particular occurrences along causal
links: market failure, invisible hand, free-rider problem, returns to scale. Bal
(2002) discusses several key concepts in the humanities – such as

impact?; and under what circumstances does this effect occur? These are the sorts of
questions to be answered in identifying causal links. His other two questions deal with
whether X is related to, and/or similar to Y; as noted with respect to Halperin, such
questions could be asked of both phenomena and links.
18
A handful of scholars were invited to address “Disciplinary keywords and their evolved
meanings” in The Chronicle of Higher Education 49:26. The term “Identity” is used by
scholars to speak to at least two distinct phenomena: visions of self within individual
schemas, and identification with groups; in uses such as “identity politics” it refers to a
causal link between such groups and political phenomena. “Postmodernism” is most often
a critique of scientific practice, but in some hands a theoretical position or even a statement
about realizations of phenomena. Various words such as “art”, “tradition”, and “text” have
come to encompass a wider range of phenomena over time. These examples highlight both
the possibility of and advantages of defining concepts more precisely.
2. Classifying Phenomena and Data 43

focalization, gaze, look, image, framing, and performance – most of which

can best be understood as illustrating elements of causal links (particularly
the link between art and human appreciation). Most famously, Williams
(1985) listed well over 100 “Keywords” for the study of culture and society.
About a third of these are adjectives such as bourgeois, collective, dramatic,
empirical, and pragmatic. The rest of Williams’ keywords bear greater
similarity to the “concepts” proffered by others. Over 25 of his keywords are
phenomena, including art, class, culture, family, industry, institution,
literature, myth, personality, science, and unemployment. Over a dozen refer
to causal links, including alienation, charity, criticism, exploitation,
management, mediation, and reform. Another dozen reflect realizations of
(especially political) phenomena, including anarchism, democracy,
monopoly, and socialist. Ten keywords reflect combinations of realizations
of phenomena: capitalism, civilization, commercialism, consumerism,
ecology, hegemony, imperialism, materialism, naturalism, and nature. Three
keywords refer to processes of change within (sets of) phenomena:
development, improvement, and evolution. Some twenty keywords cannot
be easily defined in terms of phenomena or causal links: some, like “theory”
itself refer to theories, and some, like “philosophy” itself refer to
philosophical evaluation, but some – such as man, individual, interest,
humanity, and society – lend themselves to vague general usage.
Each of these authors were motivated by a belief that clarity with respect
to concepts (or keywords) would enhance scholarly communication and
understanding. Bal thus argues that concepts “need to be clear, explicit, and
defined.”19 She worries, though, that this is difficult, because concepts are
part of “a systematic set of distinctions, not oppositions” (2002, 22). Her
point is that concepts are not independent, but rather one concept can only be
defined in terms of what another concept is not. In other words,
classification is the key to careful definition of concepts. The classifications
developed in the book can greatly facilitate the definition of those concepts
that refer to phenomena, realizations of phenomena, and types of theory. For
those concepts that refer to realizations of multiple phenomena, these too can
be defined in terms of each of these (though it may in some cases be better to
simply eschew use of the general concept). Even for the many concepts that
refer to causal links, the classification of phenomena suggests that these can
and should be carefully defined link by link.
The foregoing analysis suggests an important hypothesis: that concepts
that are not or cannot be defined in terms of (especially) phenomena, causal
links, theory, or method (or perhaps data or scientific practice) are of little or

19
To be sure, Bal is at times quite skeptical of the possibility of firm definitions, and even
celebrates the fact that different scholars will necessarily interpret concepts differently.
Nevertheless, she urges scholars to attempt to agree on what particular concepts mean.
44 Chapter 2

no scientific utility. Having followed an inductive approach to this point, the

possibility must be admitted that some concepts might be identified that are
both useful and non-reducible in this way. The logic of the classification of
science pursued in chapter 1 provides a deductive rationale. Further
justification comes from Dahlberg (1994), who attempts an exhaustive
classification of concepts with the goal of library classification in mind. She
identifies “object-related concepts”, which are equivalent to phenomena,
“concepts related to specialties of objects”, which involves realizations or
descriptions of phenomena, “activity related concepts” which deal with
causal processes (like convergence) and states (like revolution), and finally
“general and theoretical concepts”, which embraces theory, method, and
scientific practice.
A few subsidiary hypotheses can also be derived from the above analysis.
First, scholars must be careful to distinguish concepts that refer to
phenomena from those that refer to realizations of phenomena. If this is not
done, scholars can too easily confuse the eternal (there are always genders
and families) with the experience of particular societies (certain family forms
and types of gender relations). With respect to causal links, scholars must be
wary of terms that apply to multiple causal links. While most (but not all;
think of collective action) of Sandler’s causal concepts have fairly precise
meanings, most of Williams’ causal concepts can mean quite different things
along different links. While there may, of course, be value in drawing
analogies across quite different cases of “exploitation”, a scholar speaking of
“exploitation” needs to carefully define what is meant by that term. When a
term like “behavior” is applied at both individual and societal levels, special
care must be taken to identify what it means in these quite different cases.20
The same argument applies to the process terms such as “evolution”: chapter
3 will outline different types of evolutionary process. Finally, it is important
to distinguish phenomena from causal links. Bal (2002, 22) worries that
describing an object as an “image” distorts the object. But of course the
image reflects how the object is perceived (where every object or individual
reflects the intersection of multiple causal links), and thus speaking of the
image should not detract from other causal links in which the object may be
involved (as, say, when it falls on one’s foot).
It deserves emphasis that any one concept can only serve one role. Bal
(2002, 23) suggests that “concepts can work as shorthand theories.” But if a

20
The term “meaning” itself is often equated with “authorial/artist intent”, but not always
(Bal, 2002, 23). So authors need to be careful in defining “meaning” itself. Bal (p.27)
suggests that students are required to choose a definition of “meaning”; why, though,
should students not be urged to embrace all possible definitions (and the causal links they
imply): artist intent, audience reaction, and so on? – they could still choose to specialize
with respect to one of these (or not) without needing to assume the others away.
2. Classifying Phenomena and Data 45

concept defines a phenomenon or a realization of phenomenon, it cannot

simultaneously be a theory, for scholars would then be constrained to view
the particular (realization of) phenomenon through the lens of that theory. A
concept that refers to one or more causal links can imply a particular theory.
As we shall see in the next chapter, though, care must be taken to distinguish
the different constituent elements of a theory.
While clarity in the use of concepts will be of value to all scholars, it will
be especially valuable to interdisciplinary scholars. Interdisciplinary analysis
will be incoherent if scholars cannot agree on what particular concepts mean
(Bal, 2002, 28).

2. CLASSIFYING DATA

Data are by definition that which can be observed.21 Data might thus be
classified in terms of humanity’s five perceptual senses: sight, hearing, taste,
touch, and smell. In actuality, science relies heavily on the first two of these.
Since none of these senses are perfect, greater use might perhaps be made of
the latter three senses: it could be that errors in these are poorly correlated
with errors in sight or hearing (this is an argument that will be made with
respect to different methods in chapter 4). Bhushan and Rosenfeld (2000,
11-12) note, for example, that while scientists are aware of visual illusions,
there may be no such thing as smell illusion, since the sense of smell is
triggered by direct contact with molecules. More generally, they note that
chemists, like other scientists, have emphasized the visual; thus in the area of
odor research there has been a tendency to borrow inappropriate research
techniques from visual research. While classifications of visual stimuli
(color, shape, size) have long existed, classification of odors has hardly
begun. In any case, scientists should certainly be aware of the limitations of
the human perceptual apparatus (see Jackendoff, 1992).
Nevertheless, it seems more useful to classify data in terms of their nature
than in terms of perceptual apparatus. Three broad categories of data can be
observed:
x inanimate objects
x living things, including people
x events (interactions among beings and objects)

21
Mahner and Bunge (1997, 74) would define a datum as a statement about a fact, such as
“W exists” or “X is in state Y”. Such statements can be more or less true. The approach
taken here equates data with the facts themselves. The perceptual and cognitive errors that
can affect a scientist’s perception of the data will be treated in chapter 5.
46 Chapter 2

Special note should be made here of the human ability to communicate.

This gives individuals access to the observations of others, though doing so
raises concerns about errors in transmission (which may be either purposeful
or accidental). Errors in transmission are especially common when
individuals come from different cultural or disciplinary backgrounds.
Communication also provides access to others’ attitudes, and through
introspection (communication with oneself) to one’s own. Attitudes deserve
treatment as a fourth type of data, though observation of these is less direct
than is the case with the first three.
Table 2.1 gives some idea of how these four types of data might be
unpacked in terms of various examples of object, being, event, and attitude.
Lengthy lists might thus be developed of each. There are several key points
to be made about the relationship between classifications of phenomena (as
in Table 2.1) and such a list of the types of data:
Most events are not phenomena. Many would be captured in causal links.
Unique historical events are neither phenomena nor links, though scientists
should strive to explain these in terms of causal links (Szostak, 2003).
Tashakkori and Teddlie (1998) stress that some data serve as proxies for
phenomena and others for causal links.
The categories in Table 2.1 embrace different types of data. Culture is
composed mostly of attitudes, but cultural expressions embrace both objects
and events. The non-human environment is mostly objects, but contains
classes of events such as tornados. Other categories are even more diverse.
The fact that particular categories of phenomena embrace different types of
data provides the first powerful indication in this book that (most)
communities of scholars should be open to diverse types of data.
Many phenomena have no direct data counterpart. This is because
humans cannot directly observe, say, personality dimensions (or photons).
Nor can these be accessed through communication or introspection. Scholars
can nevertheless learn about them indirectly. A fair bit of consensus
regarding personality dimensions has been achieved by studying behavior
and attitudes. Ian Hacking posited that scientists can know that a
phenomenon exists if they can manipulate it; Vicedo (2000) counters that
manipulations can be consistent with different theories that posit different
phenomena. For example, the same particle may be called a gamma ray or a
photon (Krieger, 1997, 20). Nevertheless, non-observable phenomena have
played a key role in scientific advance.
Mahner and Bunge (1997, 35) define phenomena as “perceptual
appearance to senses”; they thus preclude by definition “unobservable
phenomena.” This forces them later to recognize that scientific theories will
often need to contain “non-observable constructions” (134) which must be
justified theoretically and/or empirically. By including such constructions
2. Classifying Phenomena and Data 47

among phenomena here, it can be argued in chapter 3 that scientific theories

should only be about phenomena and the links among them.
Even when a phenomenon accords with a particular type of data, there
may be other (even better) methods of observing it. Much, for example, can
be learned about attitudes by studying people and events.
Recall that most scientific investigations involve causal links among
phenomena. In investigating, say, why a particular group of people uses a
particular object in a particular way, scientists could easily have recourse to
all four types of data.
It is important not to leave the impression that data only represent
realizations of phenomena. Mokyr (2002) describes four types of useful
“epistemic knowledge”: classifications of phenomena, patterns in occurrence
[of realizations of these], regularities that govern their behavior, and
principles that guide them. The last of these is a theoretical concern. Data
can, however, usefully be gathered on the first three (which indirectly inform
theory). Mokyr notes that a regularity or relationship may at times be
identified for which there is not (yet) a theoretical explanation: why, for
example, does Planck’s constant have that particular value? (see Sklar,
1992). Such empirical regularities must nevertheless be counted as an
important scientific product; they are treated as a subset of data in this book.
A final point: much but not all of what is observed can be counted. Note
in particular that it is difficult to quantify that which is observed through the
senses of smell, taste, and touch. Valiela (2001, 29-31) modifies a
classification of the quantifiability of data first suggested by the psychologist
S.S. Steiner decades ago. He identifies three types of data. Nominal data
cannot be assigned quantitative properties for it deals with attributes such as
hair color or country of origin; nevertheless frequencies with which these
attributes are observed can be calculated (such frequencies can be termed
enumerative data). Ranked (ordinal) data can be ordered: one can say that
they like X more than Y but not by how much. Measurement data – which
Valiela believes to be the most common type – allow the full range of
mathematical manipulation. Such data may be continuous or discrete.
Valiela notes that Steiner’s classification has been controversial. In
particular, it has been suggested that his types are not precise or exhaustive.
Moreover some have worried that he classifies not data themselves but how
they are used.22 The point to emphasize is that some data cannot legitimately
be used in certain ways. Thus data can sensibly be categorized in terms of
the uses to which they might be put. And, as was seen in the previous

22
Tashakkori and Teddlie (1998, ch.6) distinguish between quantitative and qualitative data.
They note, though, that one can quantify qualitative data (by for example measuring what
proportion of people display a particular emotion in a particular situation) and alternatively
treat quantitative data in a qualitative way (using words like “very” for example).
48 Chapter 2

chapter, recognition of the possibility of some overlap among types need not
destroy the utility of a classification.

2.1 Applying the 5W Questions to Data

As with phenomena themselves, the 5W questions point to several

important considerations regarding data.
Why do scientists examine particular data? Ideally, they do so in order to
learn about particular phenomena or causal links. This simple point is too
easily forgotten. Faced with a voluminous data set, scientists will be tempted
to analyze it. They will also be tempted to reify the data, and downplay or
even ignore the necessarily imperfect correspondence between data and the
phenomenon or link for which it proxies.
How (When and Where) do scientists examine data? Relevant distinctions
were discussed above, and five broad possibilities can be generated:
x direct observation by the researcher
x direct observation by others
x introspection by the researcher
x introspection by others
x indirectly, when some combination of the above is used to create a proxy
measure. This would include the use of instruments, which some
community of scholars agrees can tell scientists about phenomena that
cannot be observed directly.23

It is important to stress that observation is never perfect. While these five

distinct possibilities can be identified, any particular correspondence between
phenomenon/link and data can be placed along a continuum of observability.
Maxwell, as early as 1962, had appreciated that there was such a continuum
rather than a sharp distinction between observable and unobservable (in
Klee, 1999). Note that the five possibilities outlined above will overlap
along this continuum, though the first four, at least, have been listed in the
order in which on average they will achieve correspondence.
Data themselves differ in the degree to which they can be observed
objectively. If scientists want to count how many cars pass a particular
intersection during a particular time period, they can potentially reach a very
objective and accurate result. If they want to evaluate how happy the drivers
look they will face greater difficulties. Different raters may generate quite

23
Pahre (1996, 206) notes that data can cross disciplinary boundaries easily, whereas a
lengthy conversation is generally required with respect to transmission of theory and
method. If data need interpretation, interdisciplinary transmission becomes more difficult.
2. Classifying Phenomena and Data 49

different results. This may especially be the case if a continuous scale is

utilized rather than a simple yes/no scale (Meltzoff, 1998, ch.7).
In addition to problems in observation, it should be noted that data,
whether qualitative or quantitative in nature, usually must be summarized in
some way (Langley et al., 1987, 19). A mass of quantitative data may be
summarized statistically, a detailed qualitative examination of a particular
case may be summarized descriptively. When a summary is necessary,
further opportunities for the entry of a subjective component naturally arise.
What is the data? Since the concern here is with the relationship between
data and phenomena or links, this question can be rephrased as follows:
What is the relationship between the data and a particular phenomenon or
link? There are three broad possibilities:
x Direct proxy. Scientists are interested in the phenomenon “total economic
output”, and thus study statistics for Gross Domestic Product. These may
be flawed, but ideally represent the phenomenon in question precisely.
x Proxy for some component of the phenomenon or link. Scientists wish to
study the phenomenon “aggression,” and perform an experiment in which
they measure how often people respond aggressively to a particular
stimulus. They recognize that this at best captures a component of the
broader phenomenon. One concern that the researcher should have is
whether the phenomenon should be unpacked and analysis performed at
lower levels of aggregation where direct data proxies may be available.
Note that scientists often have recourse to different sorts of data to
capture different characteristics of a phenomenon: for example, using a
thermometer to measure temperature and a barometer for pressure.
x Proxy for some combination of phenomena or links. Data exist on
frequency of extramarital affairs; these reflect, among other phenomena,
attitudes toward love, marriage, sex, and honesty. Scientists must be
careful not to assume that their data is related only to the phenomenon
they wish to study.

Palys (1997, ch.4) notes that scientists cannot avoid the question of
operationalizing the phenomena they wish to study. This is as true for
qualitative as for quantitative research. It is also true for inductive research,
though in this case the process works in reverse: data is first gathered and in
the process of theorizing it is associated with the phenomena that is theorized
about. Since operationalization is essential but necessarily imperfect,
scientists should be open both about the precise nature of the
operationalization that they choose and the justification that they can provide
for this (which may hinge largely on data availability), and alert to potential
50 Chapter 2

biases and errors.24 Scientific communities may agree on best practices with
respect to operationalization; this reduces opportunities for individual
researcher bias but increases the risk of reification.
The scientific community as a whole is guided to seek multiple sources of
data that might shed light on a particular phenomenon or link (Tashakkori
and Teddlie, 1998, ch. 6). In later chapters similar arguments will be made
with respect to theory and method, and 5W typologies used as guides to what
sorts of theories and methods are best suited to particular research
questions.25 The 5W approach is less helpful here, though it is likely true
that if similar results are achieved using different types of data great
confidence can be placed in these results. If different data do not agree, the
temptation to simply ignore inconvenient results needs to be avoided. Note
also that the choice of method – and to a lesser extent theory – constrains the
type of data utilized, and thus later chapters will provide more precise, albeit
indirect, advice on what types of data to embrace.
Who is analyzing the data? This, as with the ‘Who?’ question in chapter
one, raises questions of scientific bias that will be addressed in chapter 5.

2.2 Advantages of Classifying Data

Hopefully, classifying data types, and noting that there is no one-to-one

correspondence between this classification and that of phenomena, will
simultaneously highlight the imperfect operationalization of every
phenomenon (though some more than others) and alert researchers to the
advantages of utilizing more than one type of data. In particular, scientists
should be wary of the danger of reifying data. The classification serves also
to alert scientists to the possibilities: what types of data there are. This can
aid scientists (as will later discussions of theory and method) in seeking
multiple data sources. Finally, the discussion of the three types of data
proxies alerts scientists to the twin dangers of insufficient unpacking on the
one hand, and insufficient disentangling of multiple effects on the other.

24
“All terms used in descriptions or explanations must be specifically and realistically
defined. This makes it possible for someone else to understand explicitly what the facts
are, and perhaps carry out further tests of the facts or explanations. Operational definitions
make ideas testable” (Valiela, 2001, 6).
25
When evaluating methods in chapter 4, the criteria of reliability and validity will be
discussed. Palys (1997) applies these criteria to operationalization. Operationalization is
reliable if similar studies using this operationalization, especially with different
researchers, provide similar results. With respect to validity, a researcher must show that
the operationalization is associated with the phenomenon in question (this is termed
convergent validity) and not with others (divergent validity).
Chapter 3
CLASSIFYING THEORY

The focus in this chapter is on “scientific” theory, which is concerned

with how the natural and/or human world works. Later in this chapter, a
more precise definition of scientific theory will be given. Scientific theories
can be distinguished from a range of philosophical theories. Some
philosophical theories deal with questions of epistemology: how much and
how can humans understand? Theories about the nature of mathematics and
mathematical relationships can be included here. Epistemological questions
have been touched upon at various points in this book: in chapter 1 it was
argued that understanding is possible but limited, and science possible but
imperfect. Much of this book is designed to guide superior scientific
practice. Theories about how science should proceed will be engaged in
chapter 5.1 Chapter 5 will also discuss ethical theories about how the world
should work; these can color scientist’s perceptions of how it does work.2
Sadly, some theoretical texts are unclear as to whether they are addressing
“is” or “should.” Some theoretical approaches involve more than one type
of theory: feminist theory, for example, encompasses ideas about
epistemology, how the world should work, and how it does work. So does

1
Moran (2002, 87) suggests that much recent theory in the humanities has been about how
scholarship should proceed. He discusses queer theory, and notes that this is as much a
critique of whether “homosexual” is a meaningful category as an investigation of causal
relationships. As noted in chapter 1, scientific analysis of links can inform questions of the
legitimacy of certain phenomena. Moran notes that queer theory has difficulty coping with
arguments that there may be a genetic basis for homosexuality.
2
Bruce Hunter, in private correspondence, has encouraged careful use of the word “should”
here. A scientist might reasonably wonder how an institution “should” work if official
rules were followed to the letter (or how an organism should function in the environment it
was selected for); an ethicist would instead wonder how institutional rules should be
constructed ideally.
52 Chapter 3

Marxian theory; S. Turner (1996, 7) suggests that the Marxian package is no

longer coherent but each of the three elements can usefully be engaged. This
book operates from the premise that these different types of theory are best
analysed separately. Among other things, this helps solve a problem
identified by Moran (2002, 112-3) that compound theories are of forbidding
complexity. Chapter 5 will explore the dangers of a scientific community
operating within the constraint of interconnected scientific and philosophical
theories.
Throughout this chapter, the word “theory” should be taken to mean
“scientific” theory. Theories can also usefully be distinguished from
descriptive generalizations, though in practice one shades into the other
(Langley et al.., 1987, 19). As noted in the preceding chapter, researchers
may find that, say, “particle X emits field Y,” or “event X always precedes
event Y.” These are examples of descriptive generalizations. Theories try to
explain such generalizations.3 Indeed, as we shall see, theories should
provide answers to each of the 5W questions.
In the case of theory, unlike the cases of phenomena, method, and ethical
evaluation, it is not possible to begin by providing an exhaustive or nearly
exhaustive list of these. Casual inspection suggests that there are thousands
upon thousands of theories floating around in science. Nor is there any one
obvious dimension along which these can be arrayed and then grouped into a
manageable number of theory types. This being the case, development of a
typology of theory will be particularly important.
Some theories, generally termed grand theories (see list below), attempt
to encompass a wide array of causal links – some might claim to cover all.
Other theories are much more narrowly focussed, dealing with only one
phenomenon or link. Between grand and narrow theory, there is naturally a
continuum of middle-range theory designed to deal with some set of
phenomena and causal links. These are often designed with empirical
analysis in mind, and thus focus on phenomena that it is believed can be
operationalized readily.4

3
Care should be taken that a theory crafted to explain certain empirical regularities is not
tested only with respect to these. Karl Popper thus emphasized the importance of “novel
facts.” While he meant these in a temporal sense, Imre Lakatos argued that it was
important only that some supporting evidence not have guided the original development of
the theory (Motterlini, 1999). This insight supports indirectly the conjectures of this book
that evidence from multiple methods and data types is desirable.
4
Merton is generally credited with coining the phrase “middle-range theory.” Goldenberg
(1992) discusses how Merton and Mills wished to avoid the extremes of grand theory and
empiricism, by developing theories whose scope encouraged the interplay between theory
and empirics essential to scientific advance. Our usage of the term is a bit looser than
theirs.
3. Classifying Theory 53

One fairly clumsy way by which one might classify narrow theories is in
terms of the phenomenon or link to which they apply. Middle-range theories
would be a yet trickier subject, though one could hope to map the extent of
such theories onto a list of phenomena such as that provided in Table 2.1
(that is, draw arrows between phenomena to reflect the direction(s) of
causality implied by the theory). Such a mapping can be useful in providing
a quick and memorable guide to what a particular theory is about. Given the
number, and tendency to overlap, of middle-range theories, such mapping
would provide a messy device for remembering the place of all such theories.
It might, though, provide a useful means for cataloguing such theories so that
researchers interested in a particular set of phenomena could readily access
theories old and new which dealt with similar sets of phenomena (see chapter
7). In developing a typology later in this chapter, the subject matter of a
theory will naturally be a key concern.
In the case of grand theory, a list can be attempted of at least those
theories that have received significant scholarly attention. Yet scholarly
attention itself can be a problem, for theories evolve through time, and may
well split or coalesce: it may thus be far from clear where one theory ends
and another begins. Collins (1998) argues persuasively that a community of
scholars can only cope with a handful of competing theories at any point in
time. There is thus a constant pressure for some theoretical perspectives to
be merged, as others are split. While his focus was on philosophical
argumentation, a similar logic likely applies in science. Specialization
allows the scientific community as a whole to embrace an almost infinite
number of more narrowly focussed theories, but at the level of grand theory
there can be room for only a few. Even here, though, there is clearly room
for more than the handful posited by Collins. Different grand theories
emphasize different causal forces (and often thus different subsets of
phenomena), and thus encourage quite different research agendas. While
some grand theories do compete with others, in other cases scholars
interested in a particular grand theory may find that various other grand
theories deal with an entirely different set of questions. There is thus room
for different groups of scholars to focus on different sets of grand theory.
It might be thought that zoologists can hope to provide systems of
classification that do not require constant revision due to the sluggish pace of
genetic evolution. However, as only some 10% of the world’s biota is
classified, discovery of new species is a regular occurrence. These provide
little difficulty for zoologists if there is a logical place for them in existing
classifications. The emergence of new theories is more problematic for
classifications of theory, precisely because existing classifications are ad
hoc. Moreover, care must be taken in examining the work of past
generations of scientists: they likely saw the theoretical possibilities before
54 Chapter 3

them quite differently from what we do: Durkheim was thus viewed as
having been a social theorist in the 1920s but was categorized as a
functionalist in the 1970s (Freidheim, 1982, 4). Until a typology of theory,
defined in terms of an exhaustive set of criteria, is developed, it is impossible
to develop a timeless list of theories; only a snapshot of the theories in vogue
at a particular point in time is possible. The development of such a typology
will be the primary focus of this chapter.
With this important caveat in mind, a list of “grand theories in human
science that are currently embraced by large numbers of researchers” can be
attempted. The chapter headings in The Blackwell Companion to Social
Theory (B. Turner, 2000), provides a useful starting point:
x Action theory
x Systems theory, and functionalism
x Psychoanalysis
x Structuralism
x Symbolic Interactionism
x Rational Choice theory
x Phenomenology
x Feminism
x Cultural Studies
x Postmodernism

Not all of these, to be sure, are theories in the sense the word is used in
this chapter. Postmodernism, for example, is more a critique of the
possibilities of knowledge, and an advocacy of a turn away from grand
theory, than a grand theory in its own right. Likewise, feminism
encompasses a dizzying array of perspectives on scientific practice. Note,
though, that most of the rest do indeed serve the role of grand theory. And
they generally do this by guiding scientists to look for the same thing across
a wide range of links. Rational choice theory guides scientists to ask what
rational calculating individuals would do in the circumstance in question.
Functionalism guides scientists to ask what purpose an institution or attitude
serves. Psychoanalysis guides scientists to look at the conflicts raging within
subconscious minds.
Narrower theories are often a combination of aspects of grand theories,
applied to a particular research question. A scientist may posit that the link
between two phenomena reflects a particular interaction among rational
calculation, symbolism, and the role that one phenomenon plays in
supporting the other. Categorizing grand theory, then, can provide insight
into the nature of at least some narrower theories as well.
Is the above list exhaustive? The absence of Marxian theory or theories
(which is discussed briefly in several of the chapters), which still motivates a
3. Classifying Theory 55

range of scholars, suggests that it is not. Nor does the list mention
evolutionary psychology, which motivates large numbers of researchers. But
it nevertheless likely captures the majority of grand theories in widespread
use today. Without any objective criteria for judging what qualifies as a
distinct grand theory, an exhaustive list would be controversial. The
scholarly community as a whole might nevertheless be able to achieve a fair
degree of consensus.

1. TYPOLOGY OF THEORY

As suggested in chapter 1, a typology of theory can be constructed by

asking the 5W questions. This typology, it should be emphasized, is in terms
of the content of theories. Previous attempts to classify theories have,
notably, tended not to focus on content, but instead looked at the origins of
particular theories. Such classifications are easier, in the sense that they do
not depend on an understanding of the key attributes of theories, but are also
less useful in guiding scientists as to which theories might be appropriate to a
particuilar inquiry.5
E. Klein (1998), for example, types economic theory in terms of two
questions. First, did theorists intend for the theory to generate models that
have empirically testable elements? Second, was the goal of theorists to
understand/explain, or were they instead concerned with controlling and/or
influencing the economy? It will be argued over the course of this and the
next two chapters that all theories that deserve to be called scientific should
be subject to exploration in terms of some scientific method. Klein’s first
distinction thus disappears, though it might be useful to compare theories in
terms of how amenable to empirical investigation they are. Even this,
though, is a criterion for evaluating theories, not describing them. As for
Klein’s second distinction, economists can only influence – at least with
some hope of getting the desired result – if they understand; likewise
understanding inevitably sets the stage for action. There is a further

5
Scholars such as Wallace, Ritzer, and Freidheim have bemoaned the tendency to classify
theories only in terms of their philosophical or sociological origins, rather than in terms of
their internal properties. Moreover, when internal properties are discussed, there is
disagreement as to whether this should be in terms of philosophical principles, some model
of science, some view of the world, or the subject content of a theory (Freidheim, 1982, 4-
7). Freidheim (1982) herself proposed that theories could be distinguished in terms of
whether their goal was explanation or prediction; however scientists should not classify
theories in terms of the goals of the theorist, and should only trust a theory’s predictions if
it has explanatory capability (see below).
56 Chapter 3

problem: theories tend to evolve through time, and thus typing these in terms
of the intentions of the original theorist is quite problematic.

1.1 Who?

Most theories deal with how one or more phenomena influence one or
more others. Less commonly, but importantly, theories discuss the nature of
or internally generated transformations in one phenomenon. In either case
scientists investigate how changes, perhaps just relative to some
hypothesized alternative, in realizations of one (set of) phenomenon will
affect either itself or other phenomena. In asking ‘Who?’, then, scientists are
asking “Who is effecting change?” Any theory must grapple with agency at
this level. If it is said that “racist attitudes cause X,” it will be a vague theory
indeed unless the type of agent that holds such attitudes is specified. If
scholars speak of the effects of aggressive behavior they should specify
whose behavior. In most instances, the type of agent that responds to the
initiating change must also be specified.
Two important distinctions must be made here. The first involves
whether the agent is capable of intentional behavior. Non-Intentional Agents
provide most of the subject matter of the natural sciences. Atoms and
chemicals and even tornados cannot have intentions. Biologists, at least,
often investigate intentionality. Social scientists most often deal with
intentional agents, but far from always. Note in particular that institutions
and organizations, though created through human intentionality, can then
exert influences non-intentionally. Table 2.1 lists several other non-
intentional agents, such as level of income, population distribution, and
gender divisions.6 Within both intentionality and non-intentionality, it is
useful to distinguish three different types of agent:
x Individual Agents. Intentional theories here commonly focus on a typical
individual. They may also explore how different types of individual will
act in a particular situation. Theories may emphasize the effects of
individual acts, though these theories should engage the possibility of
feedback effects, such that the individual’s actions or beliefs are
|themselves conditioned by the effects these are deemed likely to have.
To understand why an individual behaved or believed as they did, it is
necessary to look within the individual at genetic predispositions,
abilities, or personality dimensions. Theories of individual agency can
embrace both subconscious motives of which the agent is consciously

6
Pahre (1996, 206) finds it useful to stress that natural science and social science cannot be
distinguished on the basis of intentionality.
3. Classifying Theory 57

unaware and also “accidents” in which neither the agent’s conscious nor
subconscious minds imagined the outcome.
Newtonian mechanics provides an example of a non-intentional theory
focused upon individual agency. Here the scholar analyses the behavior
of individual particles or bodies that are presumed to act as a single unit.
x Groups of Individuals: Talcott Parsons hypothesized four types of social
group worthy of scientific analysis (see Freidheim, 1982). The first is the
primary group or family. This has deservedly been the focus of a great
deal of scientific analysis. While some theories may reduce familial
behavior to some weighted sum of the actions and wishes of individual
family members, many theories posit that families are not just the sum of
their parts. Parsons’ second type is the organization. As hinted above, an
understanding of organizations may require a mix of non-intentional and
intentional theorizing; again, organizations are not just the sum of their
parts. Parsons also spoke of community and society. The first of these
refers to groups operating in a non-formal fashion, as in developing
cultural guidelines. The study of society, which operates through formal
institutions, requires a mix of intentional and non-intentional insights.
Statistical mechanics deals with distributions of velocities among
groups of particles. The magnitudes that are important are group
averages. Natural scientists have long distinguished these sorts of
theories from the individual-agent analysis discussed above. A perennial
question in physics is how these two types of theories might be tied
together, and whether analysis at the group level reflects either a physical
reality or ignorance about how individual particles interact. Krieger
(1997, 20, 57) notes that sometimes the behavior of a group of particles is
the average or sum of individual particle behavior, and sometimes not;
there are emergent properties of families of particles that are not apparent
at the level of individual particles.
x Relationships. Some “interactionist” scholars argue that what happens
between people is more important than what happens within them:
individuals are reconstituted through their interaction with others
(Barnes, 1995).7 Interactionists thus theorize about the effects that
various types of human relationship have. While there are obvious

7
Unfortunately, Barnes argues that only relational theory is valid. He suggests that
individual-level theory is wrong because tastes are not fixed, and functional theory is
wrong because norms are not fixed (they are manipulated rather than taken). As with all
who argue for one type of causal agent, he ignores the simple fact that the world is a
complex web of causal links: the fact that an agent is acted upon does not preclude them
from exercising causal influence. His suggestion that individuals always try to maximize
pride and minimize shame in relationships (which would allow him to ignore how
individuals shape relationships) is a dangerous oversimplification of human motivation.
58 Chapter 3

overlaps with the study of groups, the focus is on personal interactions

rather than group processes.
The analysis of fields in physics provides an example of the relational
study of non-intentional agents. Fields mediate the interaction between
the sources of the fields. It might be thought that fields could be
understood as the effects of particles, but fields inhabit the spaces
between particles, and exert a causal influence on particles (Krieger,
1997, 23-7). Some scholars focus on the fields rather than the sources.
Krieger (1997, 1) speaks of a division of labor in the natural world
between particles and fields (and the walls that separate one phenomenon
from another), and notes that these are subject to different types of
change.

It should be obvious that these six broad categories of agency can be

disaggregated. Moreover, they can for the most part be disaggregated in
terms of the list of phenomena identified in Table 2.1, and a similar list of
natural science phenomena. It is noteworthy that this disaggregation does
not proceed uniformly in terms of the categories in Table 2.1. The study of
intentional individuals does, to be sure, mostly comprise analysis of the
“Genetic Tendency” and “Individual Differences” categories, though even
here “Interpersonal Relationships” involve a different form of agency. All
elements of “Non-Human Environment”, “Culture”, and “Social Structure,”
except fauna, qualify as non-intentional agents: some are individuals and
others act as groups. In the case of “Economy” and “Health and Population”,
non-intentional and individual-level intentional agency are important.
Within the category of “Politics”, institutions call for non-intentional
individual agency, public opinion and nationalism for intentional group
agency, and crime for individual intentional agency (in each case, other
forms of analysis are possible). Art studied in the aggregate deals with non-
intentional or perhaps intentional group agency, while the study of particular
works or artists generally engages individual agency.
As noted above, scientists can often only fully comprehend the effects of
a phenomenon if they also understand its causes. Since all phenomena are
causally inter-related, this opens up possibilities of applying different types
of agency to the study of every phenomenon. While “Technology and
Science”, for example, is non-intentional in its effects, all forms of
intentional causation need to be embraced if the path of innovation is to be
understood.
If this is so, and if it is accepted for the moment that every theory
emphasizes one type of agency, a strong conclusion can be drawn: The study
of every phenomenon should involve the use of multiple types of theory. This
argument will only gain strength as the four other dimensions of theory are
3. Classifying Theory 59

discussed. Note that this argument is even stronger at the level of

disciplines, for most disciplines encompass phenomena that exhibit different
types of agency.
Some scholars might question this result. They might do this at the level
of their favored phenomenon of study, arguing that it is uniquely
comprehensible in terms of one form of agency. The research performed in
Szostak (2003) suggests that human science phenomena, at least, are
generally implicated in multiple causal links, and thus this result is highly
unlikely. The exception to this rule is “genetic predisposition” which
evolves so slowly as to be impervious to the effects of changes in most other
phenomena, at least in the days before genetic engineering was possible, and
thus may be adequately analysed in terms of individual agency. Even, here,
though, there are theories of group selection. This example suggests that
natural science phenomena can also often be beneficially analysed from the
perspective of multiple theories
A more common objection is to claim that one or more of the types of
agency listed above do not, in fact, exist. The goal of this book is to classify
all possible scientific theories: it is important not to exclude theories from
classification unless they fail some reasonably objective scientific standard.
This is a situation where the Golden Mean is a handy guide. When
advocates of only individual (intentional) agency argue that social aggregates
are created by individuals, they are right. When they go on to argue that all
such social aggregates can thus be reduced to the individual agency that
created them, they are wrong. Individuals create organizations, cultural
attitudes, and technologies that have effects quite different from any they
imagined. It is nonsensical to argue that these unforeseen effects lack causal
significance or empirical importance (see Kincaid, 1996, ch.3). Likewise, it
is correct to note that individuals operate within a complex web of societal
influences. But it is not correct to then argue that individuals are nothing
more than creatures of these influences, for each has personalities and
abilities that influence actions and beliefs. While it makes a great deal of
sense for individual scholars to emphasize one sort of agency or another
within particular research projects, they need not decree that the entire
scientific enterprise should do likewise (see Szostak, 2003).8
“Sociological theories of whatever kind can be usefully understood as
standing in direct opposition to these kind of [individual agency]

8
Ritzer had hypothesized three different types of social theory centred on social facts
(structures and institutions), social behaviors (how individuals respond to situations), and
social definitions (which at least in part dealt with group processes). He noted that writers
of all three types spoke as if their view embraced all, but argued that different data (for
which I would read “phenomena”) required different approaches (in Freidheim, 1982). I
have argued that even individual phenomena will benefit from different approaches.
60 Chapter 3

approaches”; if Jones (1985, 1) were right in this, and note that he wrote
before rational choice theory (see below) gained widespread popularity in
sociology, it could still be hoped that sociologists would be open to insights
from other disciplines which provided a complementary individual-agency
analysis. However, Jones continues that sociologists should look at such
approaches and discover (only) “what is wrong or incomplete about them”.
He then criticizes arguments from genetic tendencies on the grounds that
there are exceptions to many commonly posited links, and arguments from
personality on the grounds that personality is influenced by environment.
But these are not arguments against individual-level explanation but only for
enhancing scientific understanding of the nature of individuals and the
influences upon them. Little (1998, viii) argues in turn that the types of
theory which Jones favors require “at least a stylized conception of how
individuals within social contexts bring about these relationships,” because
there are “no supra-individual actors in social causation.” But Little cannot
have it both ways: claiming some limited possibility for theory at the
aggregate level while denying any form of non-individual agency. Either
everything can be understood in terms only of individual behavior, or the
causal power of social aggregates must be recognized.9 As Bunge (1998)
emphasizes, social aggregates have emergent properties not inherent in the
individuals that comprise them.
To recapitulate: the most common argument against the conclusion that
multiple theories are necessary is an argument that only one type of agency
exists, and this argument fails on several grounds. First, it fails the Golden
Mean test in that legions of scientists over the years have found analysis of
different types of agency valuable. Second, it requires that many of the
phenomena identified in chapter 2 be ignored; yet scientists have found all of
these phenomena useful for analytical purposes. Third, arguments in
defence of the position are faulty: humans live in a complex world of
multiple causation wherein all phenomena are both causes and effects of
others; arguing that X causes Y does not preclude the possibility that Y also
exerts influences upon other phenomena; indeed the existence of Y implies
this.

9
Little nevertheless accepts that multiple theories are necessary, at least in human science.
“Instead, the social sciences inherently consist of a messy, cross-cutting, overlapping set of
theories, hypotheses, causal models, idealized facts, interpretive principles, and bodies of
empirical findings that eliminate but do not reduce.” He likely exaggerates the orderly
nature of natural science.
3. Classifying Theory 61

1.2 What?

Having identified the agents of change, a theory must then cope with the
question of “What do change agents do?” In the case of non-intentional
agency, agents usually cannot actively do anything (tornados providing an
obvious exception). They are thus restricted to a passive form of action
wherein they provide constraints and/or incentives. It should be emphasized
that these effects can be of huge importance despite their passive nature.
Intentional individuals and especially groups can also act passively. They
also share with natural phenomena such as tornados the capacity to act
actively: to do something in common parlance. They also have the ability to
act in a way that by their nature non-intentional agents cannot: they can form
beliefs or attitudes or intentions. That is, they can have causal effects
through thoughts as well as deeds. And these thoughts can be analysed at the
level of individuals and at the level of groups; in the latter case, one can
speak of culture or public opinion or ideology or nationalism.
There are, then, three types of “what” which map imperfectly onto the six
types of “who” identified above. Most non-intentional agents are only
capable of passive causation. Some are also capable of active causation
(note that the existence of tornados places constraints and incentives on
human activity, quite apart from the active behavior of any particular
tornado). Intentional individuals are capable of all three types of causation.
Intentional groups likely have greater powers at passive causation than
individuals, and can also act and form collective beliefs. Intentional
relationships are often posited to affect mostly attitudes, but can also induce
actions and provide constraints.
One might also make a distinction between whether agents are “acting”
or “reacting”. It has been argued, though, that the world is a complex web of
multiple causation. Every phenomenon influences and is influenced by
hundreds of others. Every act, whether passive or active, action or attitude,
is necessarily conditioned by the numerous influences upon the phenomenon
in question. A scientist studying reaction is explicitly taking these into
account. A scientist studying action is treating these as exogenous for the
moment and focussing upon effects. A scientist who imagines that there are
no antecedent causes commits a metaphysical as well as scientific error.
More commonly such scientists are merely pursuing scientific specialization.
Depending on the nature of feedback effects, their investigations may yield
valuable insights. As was argued in chapter 2, scientists should study
individual causal links, but understand that any one link must be placed
within the broader context of related links.
As with agency, it is possible to disaggregate the types of action
described above. Table 2.1 provides much less guidance here: it suggests
62 Chapter 3

some of the types of attitude scientists might study. A detailed classification

of the types of action that agents might undertake could be especially useful
in distinguishing particular theories within a particular theory type. While
some deduction is possible, information scientists might usefully develop
such a list inductively while classifying works in terms of theory (see chapter
7).
The distinction between actions and attitudes has been much noted.
Indeed, the former is often associated with a “positivistic” approach to
science10 which emphasizes the analysis of only that which can be observed,
and the latter with an “interpretive” perspective which focuses upon how
agents interpret their situations. The critical role of passive causation is
ignored in this dichotomy, at least explicitly, though advocates of both
approaches may examine various constraints and incentives. As in almost
every scientific debate among opposites, there are legions of positivist
scientists and philosophers who condemn the practice of interpretive science,
and legions of interpretivists who argue the opposite. As in the preceding
section, the goal here is to classify all theories, and it is straightforward to
argue for the Golden Mean legitimacy of both types of analysis.
Indeed, Potter (2000) celebrates the belief that this longstanding debate is
disappearing in the face of recent developments, including the emergence of
critical realist philosophy that advocates a mix of interpretivist and positivist
approaches. He notes that the positivist position reflected a desire that
human science be like natural science (recall that natural objects exert active
and passive influence, but do not have attitudes). Interpretivists instead
embraced an ideal of two solitudes, and argued that human science required a
quite different set of theories and methods than natural science. The critical
realist position in turn suggests that the types of theory and method
employed in natural science are useful to human science, but must be
supplemented by theories and methods designed to deal with intentionality.
[Potter also mentions a fourth possible view, the extreme postmodernist
claim that both natural and human science are impossible; this was touched
upon in chapter 1.]
Potter goes on to show that both positivist and interpretivist extremes are
problematic, and thus that some sort of critical realist compromise is
desirable. Positivists claim that they can test their theories by looking at
observables, and only by looking at observables. They embrace the principle
of falsification whereby real world data could and would show misguided
theories to be mistaken. It is now widely appreciated, at least by

10
Seale (1999, 22) can be joined in decrying the oversimplification involved in terms like
“positivist,” which tend to lump together scholars with diverse perspectives on many
issues. Nevertheless, certain common or average tendencies can be identified in what
follows.
3. Classifying Theory 63

philosophers, though many scientists are blissfully unaware, that strict

falsificationism is impossible, even in natural science (see Motterlini, 1999,
Klee, 1999, or any modern text in philosophy of science). Human ingenuity
can always devise an excuse to explain away any unfortunate disagreement
between theory and evidence; such excuses can involve either flaws in
methodology or the effects of extraneous variables that one would wish to
hold constant. While some evidence is more damning (or, conversely,
supportive) than others, there is always scope for scientific judgment about
the validity of a theory. Say that a theory posits a particular chemical
reaction, and laboratory experiments fail to observe this: one can potentially
blame the experimental setup, or un-noted contaminants, or a host of other
problems. If numerous experiments produce the same result, the scientific
community may become confident that the theory was misguided, but only
after exercising collective judgment that all possible excuses are irrelevant.
At best then, the difference in scientific credentials could be only of degree:
scientists must interpret evidence of any type. And positivist human
scientists in any case often faced great difficulty in operationalizing the
variables in their theories: any mismatch between phenomena and data
naturally weakens the ability of evidence to test theory. Positivists had
dismissed alternative approaches for failing to match standards that they
themselves could not achieve. Still, positivists should be applauded for
attempting to identify evidence that might discredit a particular theory.
Positivists also celebrate objectivity. Again, philosophers have come to
appreciate that objectivity is a goal to be strived for but never achieved. No
scientific approach ensures perfect objectivity. Observers are more likely to
see what they are looking for, and what they look for will be determined by
their theory (and a host of social and cultural influences). Positivists also
promised that their theories would have predictive capability;11 this has rarely
been the case in human science, likely because of the complexity of the real
world. Positivists also emphasize quantification; the next chapter will
discuss the complementary roles of quantitative and qualitative analysis.
Positivists limit the range of questions that they feel can be addressed
scientifically, and even question the value of entire disciplines, including
History. Interpretivists, in turn, dismiss the entire positivist program. They
argue that human actions are inherently unpredictable, and thus it is useless

11
The economist Milton Friedman famously argued in the 1950s that the test of economic
theory should be its ability to predict: scientists should not mind if assumptions are
unrealistic as long as the results are accurate. But this argument is rarely voiced
(consciously) today; it is widely recognized that people can have little faith in predictions
if they do not really believe the model that generates them. A theory’s ability to predict
thus depends on the theory’s ability to explain, as well as the degree of stochasticity (see
below) in the theory.
64 Chapter 3

to try to establish any sort of behavioral regularities. But everyone survives

from day to day on the basis of such regularities: that most people drive on a
certain side of the road, that we will not be knifed for our seat in class, that
the cashier will give us change. Exceptions to these regularities occur, but
then individuals can try to explain these in terms of unusual personalities or
circumstances. Interpretivists are on much firmer ground in arguing that
humans are naturally self-reflective, and that human intentionality can hardly
be ignored. Rather than making human science somehow lesser than natural
science this opens up the possibility of asking agents why they did what they
did. Of course the ability to ask that question need not and should not drive
scientists to ignore other questions, such as what actions an intentional agent
performed, or how a non-intentional agent exerts influence.
Interpretivists argue that intentionality necessarily means that scientists
cannot establish regularities in human science as in natural science. Natural
science is justifiably famous for a set of laws, which seem to always hold, at
least over a wide range of circumstances (note that the laws of Newtonian
mechanics were found to be limited in applicability within the theory of
relativity), though many natural science theories have probabilistic elements.
Since human agency plays a role in most causal sequences in human science,
and humans are idiosyncratic, scientists cannot expect such lawlike
relationships to be found. There is a kernel of truth in this, to be sure.
Think, though, of the Law of Demand, whereby economists argue that
people buy more of a product when the price falls. Various conditions can
be identified in which the rule does not hold: when prices of other goods are
changing in particular ways, when price is taken as a signal of quality [in
which case scientists must investigate how individuals interpret prices], and
so on. But over a wide range of circumstances the relationship likely holds
(though it has proven surprisingly difficult to amass the sort of quantitative
evidence that positivists like, since it is hard to distinguish one effect
empirically from others). The Law of Demand may be less trustworthy than
the Laws of Thermodynamics, but it should hardly be ignored.
Interpretivists go on to suggest that the purpose of human science should
be understanding rather than explanation. In doing so, they seem to accept
certain positivist arguments about the nature of science, and then suggest that
human science cannot meet this standard. Again the difference between
human and natural science is one of degree rather than kind. While
intentionality creates some difficulties, it also opens the door to a host of
theories and methods unavailable to the natural scientist [at least when
analyzing natural objects; it will be suggested in chapter 5 that all scientists
should reflect on why they do what they do, and thus even natural scientists
will have some use for interpretive approaches]. If scientists observe that
3. Classifying Theory 65

people buy less coffee when the price rises, can they not aspire to explain
this result?12
Bunge (1998) provides numerous examples of the productive mixing of
positivist and interpretive approaches. Most branches of psychology blend
analysis of thought and action. Bunge emphasizes in particular how
neuroscience – the analysis of physical changes in the brain – has and will
aid scientific understanding of how people think and act. He notes that some
aspects of thought clearly require positivist explanations, as when brain
injuries interfere with the language faculty. Moreover, human fallibility
means that scientists cannot attribute the results of intentional actions
entirely to intentions, but must pay heed to unforeseen consequences. Nor
can scientists ignore those intentions, and treat actions as the unmediated
responses to certain stimuli. Even in cases such as the rising price of coffee,
there will be scope for interpretation (for example, of whether one’s social
standing depends on coffee consumption), and in other cases this may
swamp other considerations.13 Like Potter, Bunge appreciates that a
compromise is desirable in terms of both method and theory.
It was noted in the preceding section that scientists should be open to
combinations of theories embracing different types of agency. The same
result has now emerged with respect to “what”: scientists should embrace
positivist (action-based), interpretivist (attitude-based) and passive
(constraint and incentive based) theories. Similar results will emerge in the
next three sections as well (and will be italicized each time).

12
Peter Winch and other interpretivists have claimed that human motives are distinct from
causes, and thus interpretation must be different from explanation (B. Turner, 2000). But
surely it makes sense to speak of a person’s anger causing them to strike out. Kincaid
(1996, ch.6) argues that interpretive science is difficult but possible, provides examples of
good and bad interpretive practice (notably a tendency toward tautological arguments, as
when a norm is imagined that would generate observed behavior), argues that
interpretivism does not preclude other types of human science inquiry, and suggests that
interpretivism does not need special methods or evidence. We will see, though, that
interpretivists can have access to different methods and data.
13
Bunge then argues that in accepting an extreme criticism of objectivity and scientism,
interpretivists have accomplished little and done much damage. Likewise, B. Turner
(2000) bemoans the cultural turn in modern sociology which he credits with being
speculative rather than empirical, idealist rather than realist, concerned with images rather
than causal links, focussed on particularities rather than generalities, and thus engaged in
“the narcissistic study of its own textual traditions” (xv); he too urges a combination of
approaches. Similar arguments were made about positivist narrowmindedness in Szostak
(1999).
66 Chapter 3

1.3 Why?

“Why did the agent(s) act (react) as they did?” Whether scientists are
primarily concerned with causes or effects, they must engage this question.
In the case of a non-intentional agent, analysis must proceed in terms of the
constraints and incentives inherent in its nature. In the case of intentional
agents the decision-making process involved must be investigated. There are
five ways in which agents might make any decision. As will be discussed in
chapter 5, this list was first derived inductively, and underpins the five types
of ethical analysis. It can also be derived deductively, by asking what
possible alternatives there are to these processes.
x Rational/Consequentialist. Agents focus on desirable consequences, and
try to rationally calculate the best way to achieve those goals. An agent
deciding whether to continue in school would thus try to carefully
evaluate the costs and benefits of doing so in terms of the goals they hope
to achieve. Their ability to do so depends on their ability to accurately
perceive how the world works, and thus on their schemas (see Table 2.1).
Agents may be selfish or altruistic, and focussed on a diverse set of
genetic drives or personal desires, and these latter may reflect
sociocultural influences to varying degrees.
x Intuitive: Agents act on what their “gut” tells them, without much
conscious thought. An agent might thus drop out of school because it
feels good. An agent’s intuition may guide them toward satisfying
various genetic drives. It may act upon various subconscious schemas
about how the world works, the accuracy of which depends on the agent’s
lifetime experience, and particularly of traumatic events. It may thus
guide agents toward acts they would not rationally (consciously)
countenance.
x Traditional/Cultural. Agents are guided to do “what is done around
here.” For example, if everyone in their peer group is staying in school
they will too. This sort of decision-making at times occur intuitively, if
agents have internalized their group’s attitudes and will feel guilt for
breaking them. However, it generally has some conscious element,
where an agent’s desire to fit in overwhelms any other consequences.
x Rules. Agents follow a set of rules, which they have previously
determined to be socially desirable. These may or not accord with
cultural guidelines. Agents have devoted some thought to them (which in
turn may be influenced by any of the five decision-making processes).
Common examples are the Golden Rule, Kantian Imperative, or a belief
in Rights. But a host of more specific rules are also possible: agents may
for example believe that premarital sex is either a wonderful idea or a
3. Classifying Theory 67

horrible idea. Decisions regarding schooling might depend on rules such

as “respect authority”, “seek knowledge”, or “maintain a daily routine.”
x Process/Virtues. Agents again act consciously, but are concerned more
with how they act (the process) than what they achieve. Agents will
generally evaluate their behavior in terms of what are commonly called
virtues: they may for example be determined to behave honestly, or
responsibly, or punctually at all times. An agent might thus judge staying
in school, or alternatively quitting to support their family, to be the
responsible thing to do.

These decision-making methods are complementary. In judging what is

responsible behavior, an agent may have recourse to cultural guidelines,
evaluation of likely consequences, the Golden Rule, and gut feelings.
Theorists focussing on any one of them should be aware of (and ideally
speak to) the possibility of other influences. The five-way classification
system captures a variety of distinctions made in the theoretical literature,
such as conscious versus subconscious, and process versus outcome
oriented.14 Most importantly, it has room for both individual-level and
societal-level influences. It leaves theorists with the scope to emphasize the
internal influences on an individual (both intuitive and conscious) and/or the
societal influences upon them (again both intuitive and conscious). And
while each of the five broad types can be disaggregated –the first in terms of
goals, the second in terms of sources of intuition, the third and fourth in
terms of rule types, and the fifth with respect to virtues – each is nevertheless
characterized by a well-defined decision rule.
With respect to group or relationship agency there are additional
concerns. Is decision-making democratic, or do some individuals exert
disproportionate influence? Are there formal or informal guidelines for
decision-making? Is decision-making consensual, or are there dissatisfied
minorities? Even when dealing with individual agency, the simultaneous
acts of multiple agents must sometimes be analyzed: in some cases (such as
voting) individual decisions may simply be added, in others (such as market
exchange) individuals may compete.

14
Dupré (2001, 117) suggests that philosophers believe that an action can only be explained
in terms of goals and reasons to believe that actions further those goals. He then
recognizes culture as an alternative, and later refers to “principles” that might guide action.
His discussion suggests that greater scholarly familiarity with the five types of decision-
making is advisable.
68 Chapter 3

1.4 Where?

“Where does whatever the agent does (and the reactions to this) occur?”
This question can be interpreted both literally and figuratively. Literally, one
can wonder where in the real world “it” happens. There are two broad
possibilities: that “it” can happen anywhere, or that “it” can only happen in
certain specified situations. Figuratively, one can think of Table 2.1, and a
similar construct for natural science, as a map of the potential set of causal
links, and ask where on that map “it” occurs (one could, that is, draw arrows
connecting the phenomena embraced by the theory). Again, a theory that
speaks to a very small set of causal links can be distinguished from one that
implies that the same effect is felt along a wide array of causal links. Since
different sorts of phenomena exhibit different sorts of agency, no theory
could posit that any one sort of agency was directly reflected along all links;
other links might nevertheless encompass reactions to the particular type of
agency being studied.
Both in the literal and figurative cases, a search for general regularities
can be distinguished from a search for particularities. Philosophers have
since the turn of the last century used the words “nomothetic” and
“idiographic” to refer to these two types of analysis. As noted in the first
chapter, there has been a fair bit of confusion about what exactly these words
mean, what it is that they refer to, and where the boundary between the two
might lie. The apparent answer to the second query is that they represent
different sorts of answer to the question of “Where?” the causal relationship
under question occurs.
As for the first query, the literal and figurative readings of the question
provide different interpretations. The literal interpretation refers to
realizations of phenomena other than those specified within the causal
relationship itself. That is, if it is said that “it” can only happen in the
mountains, a particular realization of the phenomenon “topography” is being
specified. And if a scientist specifies that it can only happen in home or
office, the scientist needs to carefully specify what aspects of relationships,
family structure, occupation, technology, or several other phenomena are in
fact crucial. Nomothetic theory can be defined as that which posits a general
relationship among two or more phenomena, and idiographic theory as that
which posits a relationship only under specified conditions.
The figurative reading refers instead to causal links. A nomothetic theory
is one that suggests a relationship that operates along many causal links: a
perfectly nomothetic theory would apply directly or indirectly to all links. It
is thus a theory that applies over different sets of phenomena. Idiographic
theory refers instead to a causal relationship that affects a small number of
causal links: a perfectly idiographic theory would embrace only one causal
3. Classifying Theory 69

link. Idiographic theory would thus in the extreme apply to only one set of
phenomena.
Defining a boundary between nomothetic and idiographic theory is
problematic in both cases. There is instead a continuum: in the first case
running from specifying no external conditions to specifying these in terms
of all other phenomena,15 in the second case running from application to one
set of phenomena to application to all sets. The only way to avoid a random
cutoff is to define one of these terms with respect to an extreme position. To
do that with nomothetic theory would be ridiculous, for some set of
phenomena or external conditions for which a posited relationship fails to
hold can always be found.16 Nor can idiographic theory be held to the
impossible standard of always identifying realizations of every phenomenon
whether these are inside or outside the causal relationship in question.
As noted in the introductory chapter, well-defined boundaries are not
essential for a classification to be useful. Nomothetic and idiographic can be
helpful classes, even if every theory cannot flawlessly be placed into one
class or another. Not only is there no obvious cutoff, but a given theory may
seem to be nomothetic along one continuum and idiographic along the other.
Scientists should be aware that there is a role for both types of theory within
the scientific enterprise. Schaffner (2000), for example, suggests that
biology would benefit from theories at different levels of generality.
Theorists should struggle to be as explicit as possible about what can be
termed the “range of applicability” of theories: over what particular links
does the theory apply, and under what precise conditions (thus while
generalizability is measured along two continua, range of applicability is
defined precisely in terms of particular links and phenomena; empirical
research should strive to identify how strong the theory is in each of its uses
as well). Indeed, this is one of the primary goals of science, and one area in
which natural science excels over human science.
Scientists often employ the ceteris parabis (all other things remaining the
same) assumption. This can be a useful nomothetic tool. When an
economist theorizes that the increased price of a good will cause decreases in
consumption, ceteris parabis, they are positing a general relationship that
may fail to be observed empirically if overwhelmed by other changes (such
as an increase in income). But even here ideally the theorist (or at least the
community of scientists as a whole) would identify the widest range of

15
Freidheim (1982, 19) distinguishes between interactive types of theory, which describe the
effects of all possible combinations of variables, and ideal types, in which certain
combinations are stressed. While ideal types can be valuable, their limits must carefully be
identified, and what happens outside of this range of applicability investigated.
16
Bunge (1998, 26) notes that social regularities usually hold only for particular times and
places, and that chemical regularities only hold for a range of temperature and pressure.
70 Chapter 3

possible countervailing factors. And the general use of ceteris parabis often
clouds the nomothetic/idiographic distinction: scientists not only ignore the
possible changes in other phenomena (including sometimes feedback effects
from the changes they analyze) that would overturn their results, but even
static realizations of other phenomena that would do so (such as a belief that
price signals quality).17
The careful specification of ceteris parabis assumptions is just one area
in which nomothetic and idiographic theory can usefully be blended.
Economists tend to apply rational choice theory (see below), but are open to
a recognition that certain factors – particularly imperfect access to
information – can interfere with rational behavior. Their response to
arguments that rationality is generally “bounded” can be characterized as “So
what; we’ll just specify the relevant constraints.” This is a compelling
viewpoint, except when it is appreciated how often these constraints simply
overwhelm rational decision-making (see Fuller, 1993, 180). The upshot is
that rational choice analysis often generates the wrong result, and that it can
only be constructively applied in conjunction with a careful consideration of
the factors impeding rational behavior in a particular circumstance (Szostak,
1999). This is an example of a wider truth: scientists should seek a mixture
of nomothetic and idiographic theory.
Previous sections have discussed how answers to the “what” and “why”
questions were related to answers to the “who” question. Are nomothetic
and idiographic theory of different relative importance depending on who
the change agent is? Within the original formulation by Rickert, with its
emphasis on covering laws, the answer would surely be yes. For as Little
(1998) forcefully argues, covering laws are best suited to non-intentional
causation. A covering law stipulates a necessary relationship: if X then Y.
This can be observed in the natural world, as in the Laws of
Thermodynamics. However, as noted above, if human agency is involved
any where in the causal process, there will almost certainly be exceptions to
any rule, and thus it can only be said that “X usually or almost always results
in Y.”18 Such a regularity would still qualify as nomothetic under any but
17
“Laws with wide scope are likely to pick out only very partial causes” (Kincaid, 1996, 97).
Idiographic theory is generally needed to explain complex causal interactions.
18
Bunge (1998, 21-33) argues that some relationships are universal, such as that between
anomie and (increased) crime (or technological innovation and social mobility). Note,
though, that this relationship does not hold for all individuals, and thus an argument that it
always holds at the level of societies is questionable: it depends on individual
idiosyncrasies not overwhelming a general tendency. With enough observations, a case
might be observed where the result did not hold. Bunge recognizes that there are few
lawlike regularities in human science, and that this allowed philosophers such as Dilthey
to associate nomothetic with natural science and idiographic with human science. He
emphasizes, though, that both types of science mix the two: they look for patterns, and
3. Classifying Theory 71

the most narrow of definitions. Alternatively, Bunge (1998) notes that there
are idiosyncrasies even among atoms, though much less than among complex
organisms. There is thus a place for both types of theory across types of
agent, though admittedly the scope for idiographic analysis will tend to
increase with the role of intentional agency.
Little notes a further difference: while nomothetic analysis involving only
non-intentional agents can refer to the nature of the cause agent(s) (as when
explaining the reaction of sodium and chlorine), “Social regularities emerge
rather than govern” (1998, 242). For example common property institutions
do not cause overuse of the resource because of their intrinsic nature but
because of how people respond to this particular incentive structure: an
explanation then must involve a discussion of how people respond. But this
simply means, as has been argued above, that explanations involving
intentionality must describe what agents do.
The Golden Mean ideal can be employed again to appreciate that science
should encompass both nomothetic and idiographic theory. Almost nobody
claims that science should deal only with the general. As Aristotle said,
“Practical wisdom is not concerned with universals only; it must also
recognize particulars, for it is practical, and practice concerns particulars”
(he referred to ethics, but the principle can be extended; Hawthorn, 1991,
34). Some would argue that human science at least is inherently idiographic,
but it has been seen that nomothetic theory has a place there as well.

1.5 When?

“When does ‘it’ happen?” Interpreted literally, this question can yield the
same answer as the ‘where’ question, for it would guide scientists to look at
what conjunction of realizations of other phenomena must be in place for a
given relationship to hold.19 Instead, the “time path” of the process needs to
be investigated. How, in other words, does the causal relationship in
question work itself out through time? There are four broad possibilities:
x There are negative feedback effects such that the original impulse is
exactly negated and the system as a whole is unchanged.

also exceptions, and thus any difference is merely of degree. Potter (2000, 207-8) urges
human scientists to look for tendencies rather than laws; one advantage is that they will
then be more careful in making predictions.
19
If a causal relationship only holds during certain seasons, or at a certain time of day, these
constraints too can be defined in terms of realizations of natural phenomena (time of day).
While it may seem counterintuitive that ‘where’ and ‘when’ questions yield the same
answer, in fact scientists and philosophers generally tend to address questions of time and
space in concert (e.g. Urry, 2000).
72 Chapter 3

x The original change induces changes in a variety of other phenomena, but

the system of phenomena moves toward a new equilibrium where the
system will only change in response to further shocks.
x There are positive feedback effects such that certain phenomena continue
to change in a particular direction.
x The effects are stochastic, and thus quite different outcomes are possible
(one common subset is termed chaos theory).

It could be argued at the level of a comprehensive “theory cluster” which

encompassed all possible causal links that some at least of these types of
theory must be wrong: a world of positive feedbacks would spin out of
control, a theory which posits no change seems inherently unrealistic, and
even a series of equilibria may seem an unrealistic description of reality.
However, individual theories almost inevitably deal with a limited system of
phenomena. Even at that level certain types of theory might seem
suspicious. Indeed, all but the last time path can arguably only be posited by
ignoring links between the system in question and all other phenomena
(Szostak, 2003). Still, all types of time path can likely shed some light on
particular questions. That is, even if they are misguided in some respects,
they may yield insight about the general direction of change. Sklar (1992,
112), for example, celebrates the healthy tension between theories that posit
that some equilibrium will be achieved in the distribution of velocities of
molecules in a substance at constant temperature, and others that insist that
no equilibrium will be achieved. Scientists should, yet again, be open to
insights from more than one type of theory.
Some theories explain stability within a system of phenomena, and others
explain dynamic change. Still others discuss how a system might return to
stability after a period of change. Since many natural and human systems
exhibit stability in some circumstances, and change in others, all of these
types of theory can provide valuable insights. Likewise, scientists may
sometimes wish to understand gradual changes, and at other times more
dramatic transformations. One scientist might theorize about water as it is
slowly heated, and another about its behavior at the boiling point.
There are, naturally, a variety of causal mechanisms that can be posited
within each of these four types. Kontopolous (1993), for example, describes
some twenty “logics”, such as the s-shaped diffusion curve for innovations
(slow adoption, followed by faster, and then slower again as saturation is
approached), that might be used to distinguish particular theories within a
theory type. Moreover, many causal mechanisms can be structured so as to
fit more than one type. Theorists positing a particular causal mechanism
should carefully specify which type of time path they envision. Freidheim
(1982), for example, discusses eight processes that occur within intentional
3. Classifying Theory 73

groups. Some of these, such as socialization and coordination, appear to fit

best within the first type of time path, but only if it is thought that these
processes are successful. Conversely, differentiation within the group and
stratification may be essential to maintaining the group as is, or could lead to
group fragmentation, which may in turn be perceived as a new equilibrium,
or part of a dynamic or even chaotic process. Group formation, innovation,
and social change are processes that would fit poorly within the first time
path, but might characterize any of the remaining three.
While theorists should specify which of the four types of time path they
intend, they should not be forbidden from appealing to combinations of
these. As noted in previous sections, human scientists at least are rarely able
to posit with certainly any causal relationship. A theorist might think that
one of the first three time paths is the likely outcome, but admit that the more
uncertain fourth type is a possibility. Or they might imagine circumstances
in which multiple time paths are possible. Of course, they should strive to
identify the circumstances in question.
To claim that a certain result will inevitably follow from a set of causes is
to assert that this set of causes is sufficient. Chemists can thus claim that
sodium and chlorine under appropriate conditions are sufficient to create salt.
Ragin (2000, 103) urges scientists to seek sufficiency: “This analytic strategy
– the examination of the sufficiency of combinations of causal conditions –
is the preferred tactic for the study of causal complexity.” This he suggests
(229) is true whether the posited result is discrete (such as a movement to a
particular equilibrium) or continuous (such as movement in a particular
direction). Ragin (108) recognizes that especially in human science
sufficiency is an elusive quarry: “It is virtually impossible to construct social
scientific models that take account of every possible factor that might
influence some action or outcome.” Given the number of realizations of
other phenomena that may bear upon a causal relationship, it is all too easy
to find exceptions to any hypothesized set of sufficient conditions (examples,
that is, where the posited result did not occur). Ragin suggests that scientists
speak of “almost sufficient” in cases where only a small number of
exceptions occur but many successes are observed; scientists could and
should still seek to identify circumstances that distinguish the exceptions (or
identify data problems that may have generated false exceptions). Ragin
notes that the strategy for identifying sufficiency is to look across a wide
range of cases in which the posited set of causes was in place, and ask
whether the result always followed; he regrets that scientists often ignore this
simple strategy.
Ragin may underestimate the degree of stochasticity in human affairs.
That is, even if the true model of how the world works were known, the
future would still be unpredictable (see Szostak, 2003). Thus, he ignores the
74 Chapter 3

role that theories of the fourth type may play. He is, though, correct in
suggesting that those who posit a theory of the first three types should seek
sufficiency. The next chapter will explore how this can and should be done.

1.6 How?

As mentioned in chapter 1, the ‘How?’ question can be used as a check

on the exhaustiveness of the 5W questions. How does something happen?
In answering such a question, scientists would wonder who caused it to
happen, and what they did to cause it to happen. Scientists might also
wonder why any intentional agents did what they did. Scientists should
certainly wonder what other conditions had to be in place for “it” to have
happened the way that it did. Finally, scientists would care about how
precisely “it” occurred through time and space. It would seem, then, that the
5W questions have indeed captured all elements of the ‘How?’ question.

2. WHAT IS A THEORY?

The vexed question of defining (scientific) theory has been avoided to

this point. “Trying to give a basic comprehensive account of the concept of a
‘theory’ is an invigorating but fruitless walkabout in metaphysics” (Ziman,
2000, 117). Ziman goes on to assert that scientists generally know a theory
when they see one, and that theories are social institutions that conform to
the standards of scholarly communities. While true, such a definition hardly
helps scholars understand what science is (chapter 5), nor indicates what
elements a theory should contain. By identifying above the various
dimensions along which theories can differ, a more precise definition of
theory can be given. Theories, in other words, are attempts to specify in a
logically consistent manner agency, action, decision-making processes,
location, and time path, with respect to any (posited) causal relationship(s).20
The key defining characteristic of a causal relationship here is that it refers to
something that happens (at least potentially) multiple times rather than a
unique historical event.21

20
Kantorovich (1993, 24-5) concurs that theories necessarily embrace causal relationships.
He also stresses that each theory must have some element of novelty. Unfortunately, he
suggests that each theory invents new phenomena, when it is causal mechanisms that
display novelty.
21
As Roy Bhaskar, the realist philosopher, has noted, events are the conjunction of many
causal links, and thus unpredictable. Scientists can, though, attempt to understand each of
the (replicable) causal links involved. With respect to these, prediction may be possible,
but only if all possible conditioning influences of other phenomena are known (Freidheim,
1982). Bunge makes a similar argument about individuals and societies: these too are
3. Classifying Theory 75

The role of causal links deserves emphasis. The purpose of science is to

understand causal links, and thus each theory needs to carefully express
which causal links it addresses, and under what circumstances the theory
holds along those links: that is, its range of applicability. In other words, the
exploration of the ‘Where?’ question should underline the importance for the
scientific enterprise of the classification of phenomena, as in chapter 2.
Krieger (1997, xiv) maintains that the key to science is to identify “ways a
physical system might change or be different than it is now;” Krieger’s
program would clearly necessitate the identification of changes along all
relevant links. Just as theories will differ in how many links they address,
causal links will differ in how many theories are applicable to them. Krieger
(1997, 61) discusses the number of different forces that exist among
elemental particles – gravitational, weak electromagnetic, strong – and how
each of these may require different theoretical treatment. In some cases
different theories can be added together, in other cases different theories will
operate under different circumstances. Krieger, in particular, urges scientists
to cope with rare occurrences by recognizing that a variety of causal
mechanisms may operate along a particular causal link, some of which may
be uncommon. [Rare events may also be explained stochastically, of course.]
Grand theories especially, but many narrower theories as well, combine
theoretical analysis of different types of causal relationship. That is, they
combine theories that give different answers to at least some of the 5W
questions. These might usefully be thought of as theory clusters. It has been
argued above, and in Szostak (2003), that there are obvious advantages to
evaluating theories causal link by causal link. An analogous and
complementary argument can be made here: that theory clusters should be
evaluated theory by theory. There are several advantages to this approach:
x The similarities and differences between theory clusters can more readily
be appreciated at the level of individual theories.
x Identifying which theories within a theory cluster accord best with
empirical reality, or involve arguments that are hard to operationalize, is
also possible.
x It may also be possible to identify theories within a theory cluster that are
poorly specified (that is, for which not all of the 5 W questions have been
addressed). Mayer (1993) decries the common tendency of scientists to
emphasize certain links in a causal chain, and gloss over those that they
have little to say about.

combinations of properties, and scientists can only hope to comprehend these through
analysis of their properties and the influences these exercise (1998, 24-5).
76 Chapter 3

Indeed, the advantages are so great that scientific communities might be

better off to focus almost exclusively at the level of theory rather than theory
cluster. This is especially so given that it is advisable when evaluating any
causal relationship to consider its relationship to related processes.22 Though
theory clusters hint at how particular theories may fit into the big picture,
they do so by linking theories that need not be interdependent. One should
not reject the entire Marxian theory cluster simply because one component
theory appears misguided.
Freidheim (1982) draws a distinction between “holistic” and “relational
analytic” theory. The latter of these seems to represent the various sorts of
causal analysis discussed above. As an example of the former, Freidheim
refers to Comte’s characterization of fourteenth century France as an
“advanced theological society.” Arguably, though, one cannot analyze the
operation and effects of such a society (and certainly not operationalize the
concept so that the theory can be exposed to reality) without breaking the
idea of an “advanced theological society” into its parts: what institutions and
attitudes does the concept embrace, and how are these causally related to
each other as well as to other phenomena? (see chapter 2). In Ragin’s
terminology, scientists need to state precisely their set(s) of causal conditions
and whatever results these are sufficient to generate. Holistic theories are
poorly specified theory clusters, where the component theories are relational
analytic. Freidheim’s distinction is thus a chimera.23
If a vague aggregate concept such as “advanced theological society” is
not broken into its constituent parts, what type of theory do we have then?
Vague aggregate concepts could have been, but were not, included as a

22
Ragin (2000, ch.4) suggests that democracies arise in different ways and that this affects
how they operate. This implies that scientists need to understand how a particular
democracy emerged in order to understand how it operates. Strictly speaking, this is
wrong. The way in which a democracy emerges will influence its precise institutional
form, the attitudes of citizens, and diverse other phenomena. Scientists can thus
potentially understand why a particular democracy functions as it does with reference only
to the modern realizations of phenomena. Given our ignorance, though, we can be guided
by considerations of history to look for the outcomes of historical processes and ask how
these in turn might operate.
23
There are those who would argue that concepts like “culture” are inherently indivisible into
their constituent parts. I have argued strenuously against this point of view in Szostak
(2003). I do recognize, of course, some important kernels of truth: there are indeed strong
causal links among various elements of culture. But if these were as strong as holists
suggest than it is hard to see how cultures could evolve gradually, as they clearly do, nor
why neighbouring cultures are often similar with respect to many characteristics but quite
different with respect to others. If holists wish to maintain that some realizations of
cultural phenomena are always sufficient to generate certain realizations in other cultural
phenomena, such conjectures need to be stated precisely and bolstered empirically, not
casually assumed.
3. Classifying Theory 77

change agent above. Holistic theorists thus have no answer to give to the
‘who’ question. Nor would it seem that they would find the ‘what’ question
terribly easy to engage, for without knowing what sort of change agent is
involved one will be hard pressed to specify what it does. The same logic
applies to the decision-making processes outlined under ‘why’. The theory
of the unrepentant holist is thus no scientific theory at all, for it fails to
answer at least three of the five questions that any scientific theory should
engage.24 It could be called “bad theory” rather than “no theory”: such an
appellation may be deservedly applied to theories in which answers to one or
two questions were simply omitted through ignorance – a practice that will
be found to be all too common below. But “theories” which cannot by their
nature answer these questions do not deserve to be called scientific, whether
they exist in the natural sciences, social sciences, or humanities.
There is a long and troubled philosophical literature, at least in the
English-speaking world, which attempts to differentiate science from non-
science. The popularity of this discourse suggests that many scientists and
philosophers wish to distinguish what scientists do from what non-scientists
do. It is thus no bad thing that this definition of a scientific theory is
exclusive. Indeed, in general a definition that embraces everything is
useless. For a definition of scientific theory to be meaningful, in other
words, it must reject some forms of argument as unscientific. Some
scientists and philosophers may wish that a somewhat different line had been
drawn in the sand. However, this definition did not flow from a desire to
establish a particular boundary, but rather emerges as a logical outcome of
the 5W typology. Analysis of causal links in chapter 2 and in Szostak (2003)
led to the conclusion that holistic theories and concepts needed to be
disaggregated. The analysis here has reinforced that view.25
Having spoken of lines in the sand above, the longstanding philosophical
tradition of defining theories in terms of covering laws (especially popular
among positivists) should be noted. A theory, it was maintained, consisted
of one or more generalizable laws, as well as correspondences between
theoretical and observational terms. Such definitions have an inherent
nomothetic bias and thus exclude theories toward the idiographic end of the
continuum. Not surprisingly, the covering law approach has been repeatedly

24
Bunge (1998, 26) attributes the predictive failure of holistic Marxian analysis to its
reliance on laws of historical transformation, and failure to encompass the particular
constraints and circumstances that affect causal outcomes.
25
I have, to be sure, never liked vagueness, and thus am suspicious of any theory that uses
words that seem to lack a reasonable range of definition (see the discussion of concepts in
chapter 2). Of course, those who question the very possibility of science (and thus
celebrate vagueness, and style over substance) can likely imagine a range of personality
defects that drive me toward the ridiculous goal of organizing human understanding.
78 Chapter 3

attacked on exactly this point: that scientists can reasonably posit theories
that lack a generalizable law. The definition of theory posited here, in
contrast, does not arbitrarily exclude legitimate sorts of scientific theorizing.

3. EVALUATING THEORIES

The identification of a huge, albeit finite, set of possible theory types

could support arguments that science is impossible or entirely subjective,
unless scientists have some clear criteria for preferring one (set of) theory to
another in answering particular questions. It cannot be emphasized too much
that while this book recommends theoretical flexibility, its goal is
nevertheless to further the scientific project and this must involve identifying
the correct theory or set of theories for each causal link. Scientists should
not arbitrarily exclude certain theories from consideration, but nor should
they accept theories without question. How should theories be evaluated?
Much of the answer to that question must wait: the key test of a theory is
whether it helps scientists understand the natural or human world. This in
turn depends on whether scientific investigation using the methods to be
outlined in the next chapter provides evidence in favor of the theory.26 In
other words, the role of scientific methods is to allow scientists to distinguish
among competing theories. And this in turn will depend on theories being
stated in terms that can be operationalized for empirical analysis (see chapter
2). “Scientific theories … rest on indirect arguments from the observational
evidence” (Kitcher, 1982, 33).
It is generally only possible or desirable to test certain aspects of a theory
at a time. Even when the theory as a whole can be subjected to empirical
evaluation, the next chapter will argue that evidence should still be sought
for components of the theory. This will often mean investigating just one
causal link. Scientific hypotheses are statements of all or part of a theory in
a form that facilitates empirical investigation. In chapter 1 it was argued that
science could be understood in terms of phenomena, data, theory, method,
and practice; just as empirical generalizations were treated as a subset of
data in chapter 2, hypotheses are recognized as an important adjunct to
scientific theory here. Hypotheses should not be held to the constraint of

26
Ziman (2000, 257) notes that there is less theoretical consensus in human science. He
attributes this to the subjectivity inherent in the study of meaning. A greater appreciation
of the strengths and weaknesses of theories and methods, and recognition that every theory
has a range of applicability in terms of phenomena, would help human science recognize
the complementarity of competing theories. Szostak (2003) argued against those (e.g.
Prus) who would maintain that positivist, interpretivist, and postmodern theories cannot be
integrated.
3. Classifying Theory 79

engaging all 5W questions, though in general at least one would be engaged.

Hypotheses should satisfy three conditions. First they should be logically
derived from a theory. Even those supportive hypotheses identified by
Mahner and Bunge (1997, 78), of the form “phenomenon X can be
operationalized in terms of datum Y,” are indirectly derived from theory(ies).
Recall that any empirical evidence can be potentially dismissed on the
grounds that it was not really a test of the theory: the closer the
correspondence between theory and hypothesis the easier it will be for a
scientific community to reject this excuse. The second criterion follows: the
hypothesis should be operationalizable. So too does the third: the hypothesis
should be clearly stated.
The Duhem-Quine thesis maintains that it will always be possible to
devise multiple theories that fit equally well with any set of empirical
evidence. While true, in practice scientific communities are generally able to
achieve consensus on a preferred theory. Given the scientific incentives to
develop competing theories, this result is hard to attribute to the social
construction of reality by scientists themselves (Polkinghorne, 1996, 19).
Quine himself has pointed to many criteria that might be used to choose
among competing theories: simplicity, consistency with other theories,
generalizability, and the fruitfulness of a theory in suggesting interesting
research questions. To be sure, each of these is problematic: reality is
complex and thus scientists may need complicated theories, other theories
may be wrong, generalizability is not always desirable, and scientists may be
excited by questions that allow publications without adding to collective
understanding of the world. Moreover, it could be that different theories are
applicable to different facets of the question at hand. The point remains that
scientists can use their collective judgment to evaluate theories, and will do
so in terms of both the evidence and the internal characteristics of the theory
itself. The Golden Mean is once again applied here, this time in the
longstanding debate between empiricists who emphasize judging theories by
their accord with evidence, and rationalists who wish to judge theories in
terms of coherence (chapter 4). Happily the vast majority of modern
philosophers appreciate that there is a place for both (Klee, 1999).
What internal characteristics of a theory are particularly important in
evaluation? One critical desideratum for any theory has been developed
above: a good theory should be able to provide clear answers to each of the
5W questions. While it is likely that the answers to some of these questions
are more important for particular theories than the answers to others, a theory
is not completely specified unless answers are provided to each. With
respect to ‘where’ a theory should identify what causal links are implicated,
and what realizations of external phenomena are necessary for the posited
causal relationship to hold. That is, the theory’s range of applicability
80 Chapter 3

should be carefully specified. Having argued that answers should be

provided to each of the 5W questions, it then follows that logical consistency
is crucial. The various components of a theory must be complementary.
Consistency with other theories widely held to be correct is also desirable.
But this should not be a requirement: scientific advance often involves
overturning widely held theories, and this can hardly happen if theories
which disagree with these are forbidden.27 Interdisciplinary scholars,
especially in human science, will often confront situations in which related
theories are contradictory, and thus consistency with all unattainable.
A final plea can be made here for clarity. Theories should make clear
what are their assumptions, arguments, and conclusions. All of these should
be as explicit as possible. If scientists are to judge whether a theory provides
answers to all 5W questions, whether it is logically consistent, and how
operationalizable it is, they will need to know precisely what the theory
means. As well, clarity should also contribute to the identification of
interesting research questions – admittedly a more questionable criterion by
which a theory might be judged – though ambiguous theories often stimulate
considerable scholarly interest. Singleton and Strait (1999, 50) advise
scientists to break theories into constituent parts, and then search for implicit
assumptions and arguments. A well-stated theory should facilitate this task
as well.
It was noted at the start of this section that much of the task of evaluating
theories must involve their being tested against real world evidence, using
the methods outlined in the next chapter. The criteria of clarity, logical
consistency, and answers to the 5W questions are critical for empirical
evaluation. Dogan (1996, 300) worries indeed that sociological theory is
often too vague and broad to be testable empirically. A similar result was
derived with respect to holistic theory above. Only a well-defined theory
generates well-defined hypotheses.

4. TYPES OF THEORY AND THE FIVE TYPES OF

CAUSATION

Szostak (2003) argued that all causal links could be understood in terms
of five types of causation:
x Strict causation, as when sodium and chlorine react.
x Intentional causation, as when a person or group acts purposefully.

27
Szostak (1999) and Szostak (2003) argue in favor of various philosophical orientations –
pancritical rationalism, web of belief, and so on – which maintain that people should be
sceptical of all beliefs, but act in accord with those they have reason to believe unless/until
they compile evidence in opposition to these. See chapter 6.
3. Classifying Theory 81

x Hermeneutic causation, as when meaning is transmitted.

x Functional/structural causation, when a particular realization of one
phenomenon is necessitated by a particular realization of another
phenomenon.
x Evolutionary causation, involving some sort of mutation, transmission
mechanism, and selection criteria.

The list had been developed inductively: these are the five types of
causation that philosophers engage, and encompass the scientific analyses
addressed in Szostak (2003). It was noted in Szostak (2003) that the
relationship between these was unclear: some but not all scholars treat
intentional as a subset of strict causation, or hermeneutic as a subset of
intentional, and some question whether functional explanations are not
entirely reducible to intentional arguments.
The 5W typology of theory allows these five types of causation to be
identified much more precisely. Strict causation applies to non-intentional
agents. Intentional and hermeneutic causation apply to intentional agents:
the former deals with their actions, and the latter with their thoughts. The
first three types of causation, then, can be defined in terms of answers to the
‘who’ and ‘what’ questions. It would, in this formulation, clearly be a
mistake to view any one of these as a subset of the others.
An example of functional causation is the argument that “centralized
states need some sort of army or police force to maintain order.” As Elster
(1989) and others have noted, it is all too easy to assume such relationships.
The theorist must go a step further, and ask how the functional relationship
emerges: how, for example, do agents of the state establish an army? This
must necessarily engage scientists in intentional analysis. The question is
thus raised as to whether the only sensible part of a functional relationship is
the intentional part. But why is an army necessary for the maintenance of a
state? This can only be so if an army is the only way in which attitudes and
actions supportive (or at least non-destructive) of the state can be
encouraged.28 This must mean that the army creates appropriate constraints
and incentives. Functional arguments, then, must combine the passive action
of a non-intentional agent with the intentional acts of intentional agents (note

28
Note that arguments for necessity work in the opposite direction from those for
sufficiency. In this case, the posited result (the state) must be identified, and the question
asked whether this has/can ever exist (for long) without the posited necessary conditions
(an army). Ragin (2000) suggests that scientists are as ignorant of the implications of
necessity as they are of sufficiency. Since necessity need not imply sufficiency, scientists
might wish to identify why the necessary factors do not always guarantee the posited
result. Kincaid (1996) notes that evidence for the functional relationship is often easier to
find than evidence regarding the emergence of the necessary conditions.
82 Chapter 3

that there is also scope for intentional analysis of why the constraints
inherent in an army work).
What sorts of agents evolve? The 5W typology suggests this unusual
question. And the answer is that only non-intentional agents evolve.
Intentional individuals and groups do not evolve, though elements of these,
such as genes and beliefs, may evolve. But note that evolutionary theories
focus not on what particular nonintentional agents do but on how they
become what they are. The analysis proceeds in terms of the inherent nature
of the individual agent itself (the possibilities for mutation) and the effects of
a selection environment that comprises other agents. Theories of group
selection exist, and thus some analysis at the level of groups is possible,
though it is generally appreciated that these must explain how group
processes interact with individual-level selection. Selection criteria may be
non-intentional (as when genetic mutations are selected by the physical
environment) or intentional (as when technologies are selected by individuals
or groups). Since a theory of evolution should specify the relevant selection
environment, evolutionary causation may involve intentional as well as non-
intentional arguments.

5. CLASSIFYING INDIVIDUAL THEORIES

Of any theory five questions can be asked: Who are the agents? What sort
of action is involved? What decision-making process is employed? Where
does the causal relationship occur, both in reality and in terms of causal
links? (That is, how generalizable is the theory?) What sort of time path
does the causal process follow? Diverse theories will be placed within the
typology in this section; the results are summarized in Table 3.1 below.
Most theories in natural science, outside of biological science, involve
non-intentional agents, and passive reaction, and thus no active decision-
making process. As discussed above, these theories can be distinguished in
terms of individual, group, or relationship agency. They can also be
distinguished in terms of where they occur (in both senses of where) and
what sort of time path is involved.
Evolutionary theory in biology, as discussed above, also involves
nonintentional – generally individual – agency and no active decision-
making. Here, though, there is non-passive action, for mutation is an
inherent quality of the organisms under study. Evolution applies to all
organisms. While nominally nomothetic, evolutionary theory has
idiographic elements in that every mutation and selection environment is
necessarily unique. Evolution is a dynamic process, with organisms
changing through time (though in unpredictable directions); change likely
3. Classifying Theory 83

occurs slowly enough that some semblance of ecosystem equilibrium is

achieved. Evolutionary theory can also be applied to various elements of
human science – culture, institutions, technology, art (see Szostak, 2003).
Intentional agency thus becomes important, and considerations of decision-
making more complex. As well, since acquired characteristics can be
transmitted (evolution is Lamarckian rather than Darwinian), mutations are
much more frequent, and thus equilibrium a less likely outcome.
Complexity theory (used here to embrace catastrophe, chaos, and other
related theories) is applied across natural and human science. Newell (2001)
has argued that the questions addressed by interdisciplinary scholars are
inherently complex in this sense (though others in that volume were
sceptical). Complexity theory describes systems of interactions among
phenomena. These theories are primarily concerned with time path, and
variously generate equilibria or instability. They are generally nomothetic in
orientation, though particular applications may be highly specific. They are
focused on actions, and thus tend not to directly address attitudes (though
they could, conceivably). Decision-making cannot be strictly rational, but
must involve some adaptive elements where the actions of some agents are
conditioned by the previous actions of others (Rosser, 1999). All types of
agent can play some role within complexity theories. Notably, complexity
theories can easily be combined with others: these can describe how
particular agents behave, and complexity theory can engage the interactions
among agents.
Theories in human science (and biology) must grapple not only with non-
intentional agents, but also with the various types of intentional agency, and
thus attitudes as well as actions, and the five types of decision-making.
Indeed, some theory in human science could likely be found that would fit
every possible combination of agent, action, decision-making process, and
time-path. With respect to the ‘where?’ question, though, there are as many
possibilities as there are phenomena and causal links (a finite but huge
number). It is thus highly unlikely that every type of theory has been applied
to every possible causal link, much less every combination of relevant
phenomena along each link. As noted above the typology can thus guide
scientists to ask new questions. Note in this regard that in investigating a
link XĺY, it is useful to know something about WĺX and YĺZ, and thus
scientists can add to their understanding of a link involving one sort of
agency with appeals to theories focussed upon other sorts of agency.
The next chapter will classify every scientific method. Given the infinite
number of possible theories, a similar task cannot be performed here.
However a selection of major human science theories can be classified. To
make the task both more manageable and less subject to authorial bias, this
84 Chapter 3

section engages those theories surveyed in B. Turner (2000) and the analyses
of those theories contained therein.29
It would be desirable to begin the discussion of each of these theories
with a brief definition and/or some typical examples. Generally, though, a
wide variety of analyses are embraced by these theories. One advantage of
the typology is that it identifies key areas of agreement and disagreement
within theories. Brief examples are provided to illustrate analyses that
particular theories embrace.

5.1 Action Theory (Including Theories of Praxis)

Who?: The focus of these theories is on individual-level intentional

behavior, though the analysis of interpersonal relationships has come to be
emphasized by many recent action theories. As Cohen (2000) suggests, this
could easily be viewed as a different type of theory.
What?: The very title of “action theory” would suggest that these
theories are primarily concerned with action. Yet in distinguishing theories
of action from related theories of praxis, Cohen (2000) identifies the latter as
concerned with “what we do when we act” and the former as analyzing the
subjective meaning of human actions. Thus action theory proper is
concerned with the intersection of attitude and action. Weber, for example,
had talked about how agents would strive to act in accord with certain
values; Talcott Parsons would later suggest that agents would pursue a chain
of actions that furthered certain goals. Parsons then endeavoured to show
how goals were chosen. As Cohen notes with respect to Parsons, sometimes
meaning is emphasized so much that little is said about acts themselves.
Why?: Cohen (2000) identifies a major conflict among action theorists in
this respect. Some emphasize utilitarian calculations, while others take a
more “romantic” approach and emphasize social meaning, culture, and
personal passions. Utilitarian calculations are, of course, the most common
type of consequentialist analysis. Personal passions in turn suggest a role for
intuition. Social meaning and culture imply an appeal to tradition, but also
raise the possibility of appeals to virtues or rules. Certainly, Weber

29
The selection was not entirely random. Chapters were chosen that seemed focussed on one
type of theory (or at least a closely related family of theories). Chapters on anthropology,
feminism, and postmodernism were ignored, as these covered too wide a range of
theorizing. The chapter on structuralism overlapped significantly with that on
functionalism. I note that Plummer (2000) stresses the similarity between symbolic
interactionism and much postmodern theory. The theories surveyed represent two of the
three types of social theory identified by Jones (1985): theories of consensus (such as
functionalism) and theories of interaction; the main example of theories of conflict
provided by Jones is Marxian theory, which is treated only in passing in Turner.
3. Classifying Theory 85

emphasizes the importance of value-oriented decision-making where one

strives to “achieve a value”. Action theory, then, neither dictates nor
excludes any single decision-making process, though individual examples of
action theory generally do. Indeed, Cohen (74) applauds the recognition that
there is no one right answer here.
Note that while action theories focus on individuals, those who
emphasize cultural influences on individual behavior can in the extreme
portray individuals as merely reacting to overwhelming cultural influences.
Such theories can then start to resemble those that treat social aggregates as
the agents. Cohen thus details conflicts within action theory concerning both
the scientific and ethical implications (with respect to free will) of
individual-level versus societal-level agency.
Theories of praxis recognize that individuals perform many acts
habitually. These theories discuss how individuals both receive and transmit
signals that allow them to interact habitually. In this respect, these theories
resemble symbolic interactionism (below). By recognizing the importance
of habit, praxis theorists open the door to passive reaction. But theories of
praxis realize (and in general applaud the exercise in free will inherent in the
fact) that people do not always act out of habit. While the decision-making
process may be subconscious, and the outcome imperfectly predictable, some
scope for decision-making remains. Dewey, in particular, worried that
agents might be excessively habit driven, and celebrated surprises that would
force agents to think for themselves. Garfinkel notes, nevertheless, that
individuals tend to get frustrated when others act unexpectedly, thus forcing
deviations from habit.
Where?: Very little mention is made of the literal spatial dimension in
Cohen (2000). Giddens is hailed for recognizing that spatial context matters:
it is hard to talk on a crowded bus, for example. But less attention is paid to
what sorts of actions or decision-making might occur in some places but not
others. This in turn reflects the tendency of theorists in this tradition to speak
as if one answer applied to all situations, rather than seek the limits of
particular theories. Cohen (2000) reasonably suggests that different types of
action theory likely apply to different situations.30 What has been said of
literal spatiality can be applied also to figurative spatiality: theorists have
scarcely engaged the possibility that particular sorts of action or decision-
making or process may be more important over some links than others.
Cohen (2000) makes a passing reference to one paper that seeks to identify
different types of social relationships, but this is the only hint of this sort of
research being undertaken.

30
Yet later Cohen expresses doubt that theories of action and praxis can be linked. Why not
view these as complements with different ranges of applicability?
86 Chapter 3

When?: Talcott Parsons posited that social order results from individuals
pursuing integrated ends. That is, there is some process through which
individuals choose largely complementary goals (he, like Weber, emphasized
the pursuit of values); they then act to achieve these. This is clearly an
equilibrium process. Giddens too implies some sort of equilibrium process
in maintaining that practices must be more than locally situated because they
support structural patterns. But Giddens like Dewey celebrates those times
at which individuals break free from habit and recognizes that in doing so
they may encourage cultural change. Giddens and Dewey thus envisage a
more dynamic process than Parsons: pursuit of habit may maintain social
stability, but individuals making conscious decisions will cause societal
change. If these acts are rare, societies may move between equilibria;
otherwise they may be constantly evolving.
Comments: While ‘who?’ and ‘what?’ are fairly well specified, action
theories embrace a very wide range of answers to the other three questions.
It is perhaps not surprising that many action theorists believe that it is the
emphasis on praxis or meaning that is central to their theory, but many –
especially those who emphasize relationships – instead argue that the ‘what?’
is quite secondary to the ‘who?’ (see Cohen, 2000; those who emphasize
relationships thus borrow pragmatically from theories of action and praxis).
This very flexibility has made it difficult to define action theory or provide
typical examples: the typology suggests that the broad title captures several
theory types. Sadly, the flexibility with respect to ‘where?’ has not led
theorists to identify complementary ranges of applicability for different
theories within this family. Nor have action theorists signalled a general
appreciation for theories outside the family.31 The juxtaposition of Table 2.1
and the typology developed here could usefully guide theorists both to
identify where particular versions of action theory are particularly strong,
and also where no version of action theory provides powerful insight.
Judging from Cohen (2000), action theorists have spoken at length of
what in Table 2.1 are called schemas, motivations, and emotions (though
Cohen bemoans the lack of empirical investigation of emotions). Much less
attention appears to have been paid to personality and abilities: the table
might usefully guide action theorists to flesh out their views in these
respects. Likewise it seems that only a subset of the possible influences of
the wider society have been engaged. Cohen (2000) criticizes most action
theorists for abstracting away from power relations, for example.
The 5W typology proved very useful in comparing the detailed
discussions of diverse action theorists in Cohen (2000). And it suggests
ways in which action theory could be improved. The typology illustrates

31
Cohen (2000) notes that network theorists tend to ignore the obvious fact that individuals
spend much time alone not worrying about relationships.
3. Classifying Theory 87

Cohen’s observation that there is no single consolidating principle in action

theory; it also suggests that major subgroups of action theory can be defined
in terms of a core answer to 5W questions.

5.2 Systems Theory and Functionalism

What?: Lechner (2000) describes the core element of systems theory as a

recognition that patterns in social life are not just accidents. When people
act, they are enmeshed in relationships, institutions, and/or societies that
maintain a distinct unity.32 While systems theories must thus deal with
actions, these are generally seen as guided by attitudes, and thus systems
theorists must also engage these.
Why?: The idea of systems implies that people in making their decisions
somehow act in a way that complements the decisions made by others.
Functionalists tend to explain all components of a system in terms of the role
these play in the system as a whole (Note that biologists also use function in
a similar manner, to argue that X was selected to serve purpose Y; Hull and
Ruse, 1998, part IV). It is now widely appreciated that such simple
functionalist explanations must be supplemented with an understanding of
how phenomena within the system encourage other phenomena to behave in
stability-enhancing ways (which often involves evolutionary analysis of how
certain realizations of phenomena were selected for by the environment of
the wider system). The most common approach here is to speak of cultural
attitudes that encourage stability-enhancing decisions.
Talcott Parsons argued that American capitalism depended upon cultural
values which supported economic activity but also a sense of community.
Individuals were urged toward a rational individualism in the economic
realm, but also toward shared norms so that they could work toward
common goals. These norms would guide individuals in their evaluation of,
and reaction to, the actions of others. These shared norms could not be
imposed but were necessarily internalized by individuals. Not surprisingly,
Parsons was criticized for downplaying power relationships and various
negative characteristics of American society. For present purposes, it is
more important to note that many were sceptical of his emphasis on actors
and actions “as if individuals can simply carry structure with them”
(Lechner, 2000). Sceptics doubted that internalized norms alone could
maintain a system, and emphasized instead such phenomena as institutional

32
According to Boyne (2000), the essence of structuralism is similar: a belief in “structures”
like the periodic table in which every element is necessarily related to others in a similar
way. Saussure, for example, had stressed how words only exist in relationship to each
other within the structure of a particular language.
88 Chapter 3

constraints. The emphasis in systems theory, then, is on constraints that

encourage individuals to act in complementary ways. There is, though,
considerable room for disagreement over the form these constraints take.
Jurgen Habermas may be viewed as an exception: he notes that systems
naturally constrain individual freedom and the ability of an individual to be
themselves. He thus both applauds and emphasizes the non-systemic efforts
of individuals to be non-coerced by systems (that is, to be free). Note though
that his starting point is still the existence of constraints.
Where?: Talcott Parsons thought of society as composed of subsystems
which served different purposes in the maintenance of overall societal
stability.33 Within this view, all phenomena were part of one and only one
subsystem [Given the ubiquitous links among phenomena in human science,
this result seems highly unlikely; see chapter 2, Szostak, 2003]. Other
systems theorists have been less inclusive, suggesting that certain
phenomena or links may serve only to effect change in systems of which
they are not part. And still others have emphasized particular systems that
may embrace only a handful of phenomena. Thus, while systems theory is
often all-encompassing it need not be so.
Systems theorists often use a language that suggests that the systems they
discuss are universal. Parsons’ treatment of American capitalism, however,
stressed the uniqueness of that system. The common emphasis on culture in
systems theory suggests that a system observed in one society cannot
function in a society with quite different values.
When?: This is the most clearly spelled out element of systems theory.
The very idea of a system implies that multiple causal links interact in such a
way as to ensure system stability. Analogies are often drawn with organisms
in this respect. This does not necessarily mean that the system as a whole is
unchanging. Some theorists, at least, explicitly recognize that external forces
may act on the system in such a way as to force change (with fewer
recognizing that dynamics internal to the system might also cause change).
The best characterization of systems theory, then, would be as involving
movement between equilibria: the system is viewed as essentially stable, but
shocks may cause it to move. In the hands of some theorists, these shocks
are downplayed such that it seems that a stable equilibrium is being
examined. In the hands of others, either external shocks or internal
dynamics may be such that the system never achieves equilibrium, and thus
some sort of dynamic change seems implied. From the time of Herbert

33
He identified four key functions: sustaining the system with respect to the wider
environment; achieving collective goals; fostering internal integration; and pattern-
maintenance and tension-management. Others have doubted that these are either necessary
or sufficient.
3. Classifying Theory 89

Spencer in the nineteenth century, it has been common to speak of systemic

evolution through a process of differentiation and integration (this is also a
key component of Boyne’s, 2000, discussion of structuralism). But all
systems theories agree that there are systematic interactions within the
system that drive the system to(ward) stability. As Lechner (2000) notes,
systems theorists can recognize conflict and change, but they will argue that
change can only be appreciated by first comprehending the system.
Who?: This question has been left to the end, for it is hardly specified at
all by systems theory. To be sure, Parsons emphasized individuals (though
also relationships), but he was in turn criticized for downplaying institutions.
Since systems can potentially embrace any phenomenon, systems theorists
could potentially deal with every type of agency (systems theory is really a
theory cluster). Indeed, arguably a common flaw in systems theories is that
one or two types of agent are emphasized, even when the system clearly
embraces others. It may be, of course, that these others are simply reactive,
but this needs to be clearly stated and justified within the theory.
Comments: As with action theory, systems theory is well specified in
terms of a couple of the 5W questions, and poorly specified with respect to
others. Interestingly, many scholars have questioned whether systems theory
is really a theory or just a frame of reference. Such critics can be applauded
for urging greater attention to the ‘Who?’ and ‘What?’ questions. Notably,
similar denigration has not occurred with respect to theories that specify
‘Who?’ and ‘What?’ but not ‘When?’ and ‘Where?’ There is no obvious
reason to privilege one over the other, except for the fact that systems theory
must be a theory cluster. Critics worry that the theory is difficult to confront
with evidence.34 This will be so for any theory that fails to specify answers
to each 5W question. But surely the questions of whether stable systems
exist, and if so why, are susceptible in principle to empirical investigation?
It should be emphasized that scientists cannot just simply study an
alleged system, but must be prepared to study all links within it. Otherwise it
is all too easy to assume stability. Only if scientists can identify for every
relevant link ways in which it is encouraged to operate in a way that
generates stability can they be confident that a system exists. Links between
the system and other phenomena should also be investigated, to ensure that
these are as weak as the theory purports. By examining individual links—
both internal to the system and external – another criticism of Parsons and
others can also be answered: are there systems other than that observed
which might support, say, American capitalism? Given the myriad changes

34
There is also a question of whether systems are real or just a scholarly way to make sense
of the world. Similar questions could be asked of most/all theories.
90 Chapter 3

that have occurred in American society since Parsons wrote, and the
continued viability of capitalism, it seems likely that the answer is yes.

5.3 Psychoanalytical Theory

Who?: The focus is on the intentional individual. Indeed psychoanalytic

theory looks within the individual for desires, fantasies, and repressed
feelings from the individual’s dependent childhood. Still, some theorists
emphasize how relationships or even institutions affect the subconscious.
Habermas and Lacan and many feminists stress relationships, while Marcuse
emphasizes the deleterious effect of capitalist institutions (Elliott, 2000).
What?: By focussing on the internal workings of the human mind,
psychoanalytic theorists are naturally concerned primarily with how
individuals develop attitudes.
Why?: Given that individuals cannot know their subconscious mind
perfectly, and yet this influences their decisions, at least some decisions
cannot be strictly rational. Intuitive decision-making will dominate in at
least some circumstances.
When?: Althusser defines ideology as a discourse which leads individuals
to see themselves and others in a way that supports the existing social
structure. He stresses in particular that individuals mis-perceive themselves
to be autonomous, but is unclear on why this outcome is inevitable. His
version of psychoanalytic theory suggests an equilibrium orientation. Others
such as Deleuze and Lyotard have objected to the implication that human
beings are creatures of social forces to such an extent; they have wondered,
albeit pessimistically, if society can be transformed by changing people
(Elliott, 2000).
Where?: While there is little explicit discussion of this question, there
seems a presupposition that the subconscious mind is widely influential.
Moreover, psychoanalytic theorists have generally eschewed empirical
investigation for broad theorizing.
Comments: Psychoanalytic theory gives very firm answers to the ‘Who?’,
‘What?’, and ‘Why?’ questions. (Implicit) Consensus on ‘Where?’ may
unfortunately reflect a preference for abject theorizing over empirical
analysis. There is scope for considerable disagreement on ‘When?’ The
discussion in Elliott (2000) raises concerns that (generally negative) attitudes
toward modern society, rather than empirical evidence, may shape answers
here as well.
While psychoanalytic theory is grounded in psychology, scholars from
other disciplines have suggested various ways in which the subconscious
both affects and is influenced by other phenomena. There is thus scope for
interdisciplinary research, and especially the use of multiple methods.
3. Classifying Theory 91

5.4 Symbolic Interactionism

Who?: As the phrase “interactionism” suggests, the focus tends to be on

relationships. Plummer (2000) identifies four key themes of symbolic
interactionist theory. The first celebrates the human ability to create
symbols, and through these both define oneself and one’s situation; these
meanings evolve through relationships. The third theme is interaction itself:
“The focus of all interactionist work is neither with the individual nor the
society per se; rather its concern is with the joint acts through which lives are
organized and societies assembled” (Plummer, 2000, 195). While theorists
may ask questions about individuals or groups they will strive to understand
these in terms of relationships. Detractors often criticize symbolic
interactionism for having little place for the essence of the self:
subconscious, emotions, and so on. Some symbolic interactionists have
attempted to incorporate a more detailed vision of the self.
What?: The first of Plummer’s themes stresses the negotiation of
meaning; symbolic interactionism can thus be seen to emphasize attitudes.
Especially in the hands of Blumer, symbolic interactionism could involve
analysis of how attitudes influence actions.
When?: Plummer’s second theme stresses change, flux, and the
emergence of new properties from interaction. This indicates a stochastic
perspective on dynamics.
Where?: Plummer’s fourth theme expresses a concern with empirics.
Symbolic interactionists may explore any aspect of the social world, but are
expected to do so with close attention to detail. A great deal of modern
symbolic interactionism was presaged by the work of Simmel, who sought
richly textured vignettes rather than abstract generalizations (Plummer, 2000,
199). While in some sense symbolic interactionists believe that their theory
has universal application (at least among intentional agents), their efforts
tend to be solidly idiographic. There is, though, some discernible effort, at
least by some scholars, to draw generalizations from across studies. Plummer
applauds grounded theory35 in particular here, but speaks also of several
mini-concepts from idiographic research which seem generalizable.36
While symbolic interactionists are to be applauded for an empirical
orientation, it is nevertheless worrisome that they have relied predominantly
on a small subset of methods, particularly participant observation and other

35
Strauss and Corbin (1998) discuss grounded theory. They see it as an approach to using
qualitative data to generate theories. Within a research project there should be an iterative
process of theorizing inductively and testing hypotheses against new information. They
urge a dense textual description of patterns of relationships.
36
Some symbolic interactionists are thus open to structure, while others are opposed.
Likewise some identify themselves as scientists and others as humanists.
92 Chapter 3

case study approaches. It will be suggested in the next chapter that particular
methods lead to a bias toward particular theoretical results.
Why?: There is little explicit discussion of how meaning is negotiated.
The stress seems to be on the intuitive: humans have inherent abilities to
create and interpret symbols.
Comments: Symbolic interactionism (or perhaps Plummer) does a better
job of identifying answers to each 5W question than was the case with
previous theories. Still, there is much room for disagreement among
theorists with respect to both ‘Where’ and ‘Why’.
Plummer (2000, 196) suggests in passing that since symbolic
interactionism holds that meaning is constantly being negotiated, scholars
cannot hope to pin down precisely what symbolic interactionism itself
means. He is, of course, correct in noting that any theory will evolve in
subtle ways through time. But Plummer nevertheless manages to identify
fairly precise answers to the 5W questions. A general lesson can be drawn
here: scholarly communities can and should essay to identify the
characteristics of any theory at any point in time.
Plummer also remarks that symbolic interactionism, though a minority
view in modern human science, nevertheless has wide but generally
unrecognized influences on writers as diverse as Bourdieu, Giddens, and
Habermas. An effort to identify theories typologically should make it easier
to recognize when elements of particular theory types are being applied
elsewhere.
Plummer applauds efforts to codify and draw generalizations from
detailed research: “This is certainly not to argue for an exclusive insistence
upon cumulative classifying work ... But some move in this direction could
help to prevent each new generation of interactionists from having to repeat
history” (2000, 210-1). The project Plummer envisages would be greatly
facilitated by the development of a typology of scientific possibilities (as in
chapters 2 through 5), and efforts to classify scientific documents in more
detail (see chapter 7). It would also benefit from researchers being more
explicit about their theoretical arguments. Interdisciplinary researchers in
particular would benefit from an enhanced ability to recognize when similar
theories are applied in different fields, and then seek commonalities and
contrasts among these.

5.5 Rational Choice Theory

Who?: Individualism is the key assumption of rational choice theory

(Abell, 2000, 231). Abell, though, sees game theory as an offshoot of
rational choice theory, which investigates strategic interaction among
3. Classifying Theory 93

rational actors; game theorists thus emphasize relationships, though in a way

that these are understood in terms of the individuals involved.
Why?: As the name suggests, emphasis is upon rational decision-making.
Abell (2000) identifies two other key assumptions – self-regard and
optimality of decision-making – but notes that all are relaxed on occasion.
Discussion of cultural norms or altruism is usually in terms of explaining
these as outcomes of rational decision-making. Abell suggests that if cultural
norms effectively limit individuals to one choice, rational choice theory
collapses into functionalism. He also recognizes that in some strategic
situations the best strategy for an individual to pursue is to mimic others; this
bears some similarity to decision-making according to tradition.
What?: The general focus is upon actions. Individuals are assumed to
proceed rationally from preferences and schemas to decisions about actions.
Abell notes that rational choice theory is generally unclear on where one’s
preferences come from; he suggests that insights from other theories
regarding learning and emotions could be useful here. Recently rational
choice theorists have begun to investigate the tricky question of how
individuals decide whether it is worthwhile to obtain additional information,
when by definition individuals cannot know how useful new information will
be.
When?: Economists have found it easier to generate equilibrium
outcomes in models with rational agents than have sociologists such as
Parsons who investigate multiple influences upon decision-making (Abell,
2000). Nevertheless, equilibrium outcomes are not inevitable. Game theory
in particular has long sought to identify equilibria, though not always
successfully.
Where?: Rational choice theorists have not only striven to explain all
individual-level behaviors, but also to reduce social behaviors to individual
actions.
Comments: Like symbolic interactionism, rational choice theory provides
some sort of answer to all 5W questions. In the case of ‘What’ and ‘Why’
there is some danger that this precision is achieved only by ignoring the
genesis of preferences and schemas. This is no great problem if, as Abell
recommends, rational choice theory is viewed as a complement rather than
substitute for other social theories (see Szostak, 1999). If rational choice
theorists engage a wider understanding of ‘Who’ and ‘Why’ they would have
to revise their answer to ‘Where’ as well, to appreciate that rational choice is
more important along some links (buying a car) than others (falling in love).
Rational choice theorists are also well advised to remember that rational
choices do not necessarily generate equilibrium outcomes.
94 Chapter 3

Table 3-1. Typology of Selected Theories

Theory Type Who? What? Why? When? Where?
Agency Action Decision- Time Path Generalize?
making

Most Natural Nonintentional Passive Inherent Various Various

Science various types

Evolutionary Nonintentional Active Inherent Not the same Nomothetic

Biology individuals equilibrium

Evolutionary Intentional Active Various Not the same Nomothetic

Human individual (any)
Science (group) equilibrium

Complexity Various Active Not strictly Varies by Generally

Theory and rational version nomothetic
passive

Action Intentional Action Various; Various Generally

Theory individual (attitude) often idiographic
(relationship) rational

Systems Various Action Various; New Generally

theory; and emphasize equilibrium nomothetic
Functionalist attitude constraints

Psychoanal- Intentional Attitudes Intuition; Various Implicit

ytic Theory individual others nomothetic
possible

Symbolic Intentional Attitudes Various Stochastic Idiographic;

Interactionism relationships some
nomothetic

Rational Individual Action Rational Usually Nomothetic

Choice equilibrium

Phenomenol- Relationships Attitudes Various Various Various

ogy (individuals) (actions)

5.6 Phenomenology

An important distinction between scientific and philosophical theory was

drawn at the start of this chapter. This section has eschewed here those
3. Classifying Theory 95

chapters in B. Turner (2000) that were focussed predominantly upon

philosophical rather than scientific theory. Phenomenology is an
intermediate case. According to Vaitkus (2000), phenomenologists celebrate
the fact that there has never been a satisfactory answer to the question of
what phenomonology is. This must render classification difficult.
Moreover, in urging scientists to directly explore individual phenomena and
to free themselves from presuppositions unjustified by prior examination,
phenomenology takes a (laudable) position on philosophical issues, rather
than a position on scientific questions. There is an emphasis on the problem
of intersubjectivity, and thus the ‘Who?’ question can at least tentatively be
answered with relationships and ‘What?’ with attitudes, though there seems
to be much concern with actions (and at least one modern phenomenologist,
Natanson, has explored the reflective ego contemplation of the world).
Answers to the other three questions are even more tenuous.

6. ADVANTAGES OF A TYPOLOGY OF THEORY

Several advantages flow from the application of the 5W typology of

theory. By providing an exhaustive set of what is possible, the typology can
alert researchers to possibilities they may have overlooked.37 Researchers
should be guided to ask what theory types have not been applied to a
particular question. Longino (2000, 63) reviews how scientific understanding
of behavior has benefited from genetic, biological, environmental, and
developmental theories; other areas of science might likewise benefit from
recognition that theories can be complementary.38 Note that even if the link
being analyzed involves only one type of agency, it may still be useful to
employ other theory types. These can show how the phenomena under study
both react to and affect other phenomena. In asking why one institution has
a particular effect on another, considerable insight may be gained from
exploring how individuals or groups affect and are affected by these.
Likewise, S. Turner (2002) argues that an understanding of how the brain
works should inform scientific understanding of how cultural elements exist,
and thus the effects these can have.

37
This approach is similar to Freidheim’s advice that scientists take existing works, analyze
the theories therein along appropriate dimensions, see what is missing, and work toward a
comprehensive theory (more properly, theory cluster). Note that this process involves both
deduction and induction. As Freidheim (1982, 80) notes, “Mine is not a neat system.” But
the world can hardly be comprehended without a complex research program.
38
While many biologists hold to the view that adaptation is the only possible biological
process, others have urged greater flexibility, noting for example that some observed traits
may be side-effects of selection processes (Hull and Ruse, 1998, part I).
96 Chapter 3

At present, researchers are likely to develop theories that reflect dominant

paradigms in their field, or that provide a counterpoint to existing theories;
they might also use induction from data or experience; the typology will
guide researchers to widen their search. Pahre (1996, 218) applauds
Tinbergen for applying Newton’s inverse square law of gravitation to
international trade: trade flows would be positively correlated with income
and negatively with distance. In drawing this connection Tinbergen must
have recognized that both gravity and trade involve to some extent non-
intentional individual agency, passive action, generalizability, and
equilibrium outcomes. A typology of theory could encourage much more
borrowing across seemingly disparate cases.
Most centrally the typology guides scholars to appreciate that scholarly
communities should be striving for a mix of different types of theories.
Necessarily each of these will be deficient in some ways. Rather than
abandoning a research problem because of this realization, scientists are
guided to work toward an ever-greater synthesis (Dogan and Pahre, 1990,
148). Tinbergen might never have touched gravity if seeking the one perfect
explanation, for intentionality can obviously not be ignored in the study of
trade. From this perspective, the goal is to identify the range of applicability
of each theory: understanding of particular causal links will in turn depend
on understanding how important each relevant theory is along that link. And
thus finding limitations to a particular theory should be seen as progress .39
Rule (1997, 18) forcefully makes this latter point:
“ ... no one genre of theory ... can reasonably claim to yield the full range
of insights required for our analytical needs. The best case for progress in
the understanding of social life lies in what I see as the expanding fund of
insights and understandings derived from a wide variety of theoretical
imperatives. Moreover, I shall argue, every theoretical program is prey to
the tendency to fruitless obsession with issues of relevance only from
within its own frame of reference.”
In particular, the typology both allows and encourages scholars and
scholarly communities to move past false dichotomies: individualism versus
emphasis on societal aggregates, interpretivism versus positivism, rational
versus non-rational decision-making, nomothetic versus idiographic

39
Rule (1997, 32) warns scholars against glorying in the simple application of a theory to a
new domain, and calling this progress. It must instead be established that this application
has some empirical validity. Dupré (2001, 16) suggests that theories become less effective
the farther from their home they travel; this is a useful warning, but scientists should be
open to theories that inform their understanding even a little. Dogan and Pahre (1990, 148)
warn scholars to be wary of theories only used in one discipline. While thus dismissive of
idiographic theory, they do a valuable service in pointing to the value of applying a theory
widely.
3. Classifying Theory 97

theorizing, and equilibrium versus non-equilibrium outcomes. The 5W

typology shows that these are all legitimate and necessary types of theory, if
scientists aspire to a comprehensive understanding of our world.
Questions of generalizability are often ignored by theorists. For many
theories, some of the other 5W questions are also ignored. The typology
guides theorists – and empiricists – to specify answers to each question, and
thus encourages one important sort of theoretical progress.
At present it is difficult to identify the essence of particular theories, or to
appreciate the differences and similarities across theories. The 5W typology
summarizes most/all of what the authors in B. Turner (2000) found it
important to say about the theories they discussed. It thus readily identifies
the essence of different theories, and the key sources of theoretical
disagreement both within and across theories. Communication across
theoretical camps, and even disciplines, should be facilitated. Notably, a
systems theorist answering the ‘Who’ and ‘What’ questions as an action
theorist would can be indistinguishable from an action theorist who answers
the ‘When’ and ‘Where’ questions as in systems theory.
Of course, the argument of the preceding paragraph could be turned on its
head to suggest that the essence of theories can readily be surveyed without
the use of the 5W questions. To be sure, encyclopaedia entries can likely be
found that provide such a service for most theories. Each of these, though,
will take a different approach and emphasize different points. The 5W
approach will aid students or scholars in the mastery of key strengths and
weaknesses of theories by organizing these around the 5W questions.
Moreover the typology will greatly facilitate comparison of strengths and
weaknesses across theory types, and also within theory types. In the absence
of clear comparisons, the case for theoretical flexibility is much harder to
make, both in general and with respect to particular research questions.
The fact that the typology provides an exhaustive set of types of theory
deserves emphasis. While infinite permutations of detail may be possible
within theory types, there is a clear finite limit to the number of types. This
typology, in concert with the finite lists of methods and phenomena from
chapters 2 and 4, means that science is essentially bounded. Rule (1997, 8)
worries that, “There is simply no limit to the theoretical agendas that could
conceivably serve to animate our work ...” At least at the important level of
theory types, this is not true. Thus Rule’s concern that social science can be
a non-progressive taste-driven endeavor which moves endlessly from one
theory to another is exaggerated – at least if scholars do not forget past
theories. A mistaken impression that the possibilities are boundless is one
source of postmodern despair, and a recognition of boundaries should
encourage a belief that enhanced understanding is possible.
98 Chapter 3

Scholars often merely “reinvent the wheel,” reproducing research on a

problem abandoned long before. The 5W typology, combined with the other
classifications in this book, provides a potential system of categorization
such that a researcher could immediately know if a particular type of theory
(and method) had ever been applied to a particular set of phenomena (see
chapter 7).
The typology allows a superior definition of the word “theory”. B. Turner
asks what a theory is.40 Is a theory a broad framework for organizing
research? Or a collection of general concepts that direct research? Or a
specific orientation that guides researchers to ask particular questions?
Theory can be all of these things, but cannot be just any one of them.41
Theory clusters, and perhaps some nomothetic theories, may serve as broad
frameworks. But theory must do more than paint the big picture; it must also
fill in the details. Nomothetic theories can be thought of as general concepts;
research that was only guided by these would be severely limited. As for
specific orientation, theory can and should guide the formation of research
questions. But theories must in turn be informed by research, or science can
never progress. And since theory must be broader than any of these
definitions, progress is best evaluated in terms of whether overall scientific
understanding of the strengths and limitations, and especially the range of
applicability, of the theories with which scientists strive to comprehend
reality, has increased.
In sum, the typology serves two main goals: it defines what theory is (and
is not), and shows that there are a finite number of complementary theory
types, each of which has an important role to play in the scientific enterprise.
Perhaps most importantly it shows that each theory type has strengths and
limitations. In doing so, it serves a third purpose: identifying and guiding
scientists toward theoretical progress. All existing theories can be situated
within the typology; in doing so it often becomes apparent that certain
questions have been unanswered. Scientists are thus guided both to improve
individual theories and to seek to identify the range of applicability of each.

40
B. Turner lists four problems with social theory: there is no obvious progress (and much
re-inventing the wheel), it is not clear if the goal is explanation or interpretation, there is no
consensus on what a theory is, and theorists still disagree about various (false)
dichotomies. The typology provides partial or complete solutions to all of these. Turner
later (2000, 16) raises a fifth concern, that the relationship between social theory and ethics
needs to be clarified. This issue will be addressed in chapter 5.
41
The same judgment holds for other definitions the author has encountered. Theories
contain a set of assumptions, but must build upon these in a particular way. A theory may
be a proposition yet to be verified; not only must the nature of the proposition be specified,
but scientists must embrace theories for which there is supporting evidence. Mahner and
Bunge (1997, 87) deal harshly with definitions of theories as the opposite of facts,
equivalent to hypotheses, or generalizations of observed facts.
Chapter 4
CLASSIFYING METHOD

“Of all the courses we teach, none is more important than research
methods” (Singleton and Strait, 1999, xv). Scientific theories were identified
in the last chapter with scientific explanation. It was also noted that scholars
can only have confidence in theories to the extent that these are logical,
coherent, clear, and have been supported by evidence gathered using
scientific methods.1 To be sure, philosophers of science, if not practicing
scientists, now accept that scientific methods can neither prove nor disprove
any theory (nor even any narrow hypothesis). Nevertheless, the application
of scientific methods provides scientists with invaluable if imperfect
evidence with which they can judge whether a theory seems to accord with
reality. The key to science is the imperfect testing of theories against
external reality (Polkinghorne, 1996, 18-19). Scientific theories guide
scientists to ask certain research questions; 2scientific investigations in turn
suggest changes to theories or perhaps even the development of new
theories.

1
Faced with difficulties in defining “scientific explanation” (that is, defining “theory”),
many philosophers of science focused instead on “scientific confirmation.” Hempel
posited that confirmation occurred any time a suggested relationship was observed. The
Bayesian approach speaks of any evidence that raises the probability that the hypothesized
relationship is correct (this unfortunately excludes evidence obtained before the theory was
formulated). The bootstrap hypothesis speaks of confirmation in terms of a wide variety of
empirical tests that approach variables differently (it also applauds the revision of theory in
light of new evidence) (Lambert and Brittan, 1992).
2
Socrates, Aristotle, and Gadamer are among those who emphasized that asking questions
was the key to knowledge. Bacon and Kant were part of a minority of philosophers who
described scientific methods as asking questions of nature. Scientific advance depends on
asking new questions, but also on getting at least partial answers to these to guide further
questioning (Hintikka, 2000, 402-4; he notes that the alternative view, that science
advances through the accumulation of incontrovertible truths, is untenable)
100 Chapter 4

In chapter 3, a distinction was drawn between scientific theory and

various forms of philosophical theory. It is likewise useful here to
distinguish scientific methods from certain elements of scientific practice.
Scientific methods are means by which to obtain evidence of how some
aspect of the natural or human world functions. A more precise definition
will be provided below. As noted above such methods are essential to
establishing the range of applicability of theories. Scientific methods can be
distinguished from a set of prior research decisions – regarding the theories
to test, phenomena to study, and methods to use – that every scientist has to
make. Though these decisions are often misleadingly termed “the scientific
method,” they are better seen as examples of scientific practice, the focus of
chapter 5.
Among the decisions that scientists and scientific communities must
make is how supportive of a particular theory the evidence is. Polanyi has
argued that all scientists learn “tacit skills” for performing this task that
cannot be readily articulated, but will reflect their community’s preference
for the use of particular methods (Polkinghorne, 1996, 17-18). These “tacit
skills” are quite distinct from methods themselves, and indeed will often act
to discourage the use of multiple methods (again, see chapter 5).
It can be noted here that there is now philosophical, if not scientific,
consensus that no one method animates science – “the methods of science
themselves are not unitary” (Stump, 1996, 449) – except in the limited sense
that some way(s) of making research decisions may serve as an exemplar(s)
(Fuller, 1997, 2). That is, the various methods outlined below cannot be
subsumed within one broad scientific method, though certain Wittgensteinian
commonalities will be explored.
Scientific methods can also be distinguished from techniques or tools,
such as experimental design or instrumentation, or particular statistical
packages. These will receive little treatment here: properly a result of the
unpacking of methods, such techniques or tools will be discussed only
insofar as necessary to understand the strengths and weaknesses of methods
themselves. Note, though, that disciplinary scholars are often trained in a
subset of the techniques or tools that might be employed within a certain
method; interdisciplinary scholars especially, but also more specialized
scholars, can usefully be acquainted with a broader range of techniques, as
well as the broader range of methods that will be suggested below.
Particular techniques or tools may have applicability within more than one
method. Scientists may, for example, employ counterfactuals: if a theory
says that X causes Y, evidence may be sought that in the absence of X, Y
does not follow, at least necessarily. Note that the counterfactual could be
tested using a variety of methods, just as the theory proper could be.
4. Classifying Method 101

Scarcely any scientific work is published that does not include appeals to
authority in the form of citations of the work of other scientists. At the level
of individual pieces of research, then, appeals to authority loom large as a
method for justifying one’s results. This reflects the simple fact that science
is a complex and cumulative endeavor, in which no piece of research stands
alone. Scientists, though, should only cite those who they believe to have
valid scientific justifications for the conclusions that they draw. And thus in
citing other scientists they are implicitly citing the methods they (and those
they in turn cite) have employed to achieve their results. And thus authority
does not deserve mention in its own right among a list of methods.

1. CLASSIFYING METHODS

Disciplinary programs, whether undergraduate or graduate, tend to devote

considerable effort to a discussion of the method or methods favored by the
relevant discipline. However the author is unaware of any program, or
indeed any book, which discusses all of the methods employed by scientists.
Notably, works on method almost never juxtapose the methods of the
humanities with those of other sciences. Most disciplinary programs (with
sociology providing the major exception) emphasize but one or two methods.
Programs that take interdisciplinarity itself as a focus tend toward a broad,
though not exhaustive, coverage of methods. Programs that cross only one
or two disciplinary boundaries, such as chemical physics or political
economy, tend to treat only the methods used in the parent disciplines.
By reading works on research methods and methodology across a range
of disciplines, a list of scientific methods can be developed inductively.
Inevitably, it is possible that some methods will have been neglected –
though only if such a method is employed by very few scholars – or that
methods that deserve separate treatment have been grouped together.
Nevertheless, the primary result of such efforts – that the number of distinct
methods used by scholars is manageably small – is robust, and would be
upheld by further research along these lines. While these inductive efforts
have focussed so far upon modern Western research, it is likely that methods
employed successfully in other times and places would have been borrowed,
or at least referenced, in the works of modern Western science.
There are, broadly speaking, some twelve distinct methods employed by
scholars (often in combination):
x experiments (including natural or quasi-experiments)
x surveys
x interviews
102 Chapter 4

x mathematical models (and simulations), including game theoretic models

x statistical analysis (often, but far from always, associated with models)
-including secondary [that is, collected by others] data analysis
x ethnographic/ observational analysis (though “Participant Observation,”
in which the investigator interacts with those under observation, is most
common, discreet observation is also possible).
x experience/ intuition [some would treat this as an important subset of
observational analysis, since we are in effect ‘observing’ ourselves here]
x textual (content, discourse) analysis
x classification (including evolutionary analysis)
x mapmaking (both representational and conceptual)
x hermeneutics/ semiotics (the study of symbols and their meaning)
x physical traces (as in archaeology)
x some would treat “evaluation” of programs as distinct, though it can be
seen as a combination of some of the above methods. Similar arguments
can be made with respect to demography, case study, feminism,3 and
perhaps also hermeneutics.

This chapter, unlike the previous chapter on theory, thus starts with an
exhaustive and manageable set of scientific methods.4 It is quite possible,
then, that both scholars and students could be acquainted with the basic pros
and cons of every method, should this prove desirable. It will prove possible,
indeed, to evaluate the strengths and weaknesses of all methods in terms of
the 5W questions.
Some scholars would attempt to aggregate some of these methods into
broader categories. “Theoretical” methods, designed to look at the
foundations of a theory, are sometimes distinguished from “Empirical”
methods. Yet most methods can serve to establish both the general adequacy
of a theory and the quantitative strength of posited relationships.
Mathematical modelling is often viewed as theoretical, and statistical testing
as empirical; yet the two are often used in combination. Some scholars
would include the logical analysis of theory structure as a theoretical
method; it is best to distinguish various means of internal evaluation of

3
Hall and Hall (1996) note that the feminist critique entails no unique method but rather
concerns about why particular questions are asked and problems addressed, and respect for
the subjects (people) being studied. Feminists have tended to favor qualitative research
(see below).
4
Singleton and Strait (1999) suggest that it is methods that distinguish science from
journalism (but also distinguish natural from social science and each from the humanities).
These views will be critiqued below. Notably, Singleton and Strait suggest that familiarity
with methods can help citizens understand and evaluate scientific results, and also sharpen
individual powers of critical thinking. They thus suggest that everyone can do research.
4. Classifying Method 103

theories (see chapter 3) from scientific methods, which must at least support
the exposure of theories to real world evidence. “Quantitative” methods are
sometimes distinguished from “qualitative” methods. In terms of the
methods that were listed above, quantitative researchers favour experiments,
surveys, statistical analysis, content analysis, and structured observations
(where particular acts are counted), while qualitative researchers favor
participant observation, in-depth interviewing, case studies in general,
biography, and focus groups (Oakley, 2000, 26). Yet exceptions abound in
which advocates of one approach have nevertheless used the methods of the
other. This overdrawn dichotomy will be revisited below.
Works that survey several methods not surprisingly take the point of view
that diverse methods have a place in science (Ellis, 1994, Hartman and
Hedblom, 1988, Singleton and Strait, 1999). For some, the argument is that
certain methods are best suited to the study of certain phenomena. More
commonly, it is suggested that a particular research problem is best
approached with multiple methods. The 5W typology will allow this
conjecture to be evaluated in some detail.

2. TYPOLOGY OF STRENGTHS AND

WEAKNESSES OF METHODS

It was argued in chapter 3 that science should embrace theories that deal
with each of the distinctions identified in the typology of theory.
Necessarily, then, scientists should want methods that would cope with each
of the cells in the typology of theory. Obviously, there will not be distinct
methods to place in each cell. Instead, every method should be investigated
to ascertain how well it copes with the different types of agency, action,
decision-making process, and time path enumerated in the typology of
theory, as well as how generalizable it is. In developing a typology of
strengths and weaknesses of methods, then, the 5W questions should be
asked of method as well. It might be found that one method is clearly
superior for a particular cell in the typology, all methods have some
applicability, or a subset of methods is useful to differing degrees.
Ironically, while a much smaller list of methods has been identified than
was the case with theory types, the typology of strengths and weaknesses of
methods will comprise all of the distinctions relevant to theory, plus several
additional dimensions. That is, in asking the 5W questions of method, a
handful of additional criteria by which to judge methods will be identified.
These additional dimensions refer to potential limitations of methods, rather
than their range of applicability.
104 Chapter 4

Note that if different methods are best suited to investigating different

theories, then in comparing one type of theory to another, scientists must
have recourse to different methods. If only one method is used, it will bias
the results. Theories about particles encourage experiments with particle
accelerators, the results of which are interpreted in terms of particles
(Krieger, 1997, 102). This simple insight is one of the more profound
implications of the analysis in this book; since most disciplines rely on a
single method, bias toward particular types of theory is endemic within the
scientific enterprise.

2.1 Who?

As in the case of theory, the concern here is with “Who is being studied?”
Again intentional agents can be distinguished from non-intentional agents,
and individuals from groups and relationships. These distinctions can have
important implications for method. Freidheim (1982), for example,
hypothesizes that it is easier to precisely measure individual-level
characteristics. If so, there is a danger that scientists will pay more heed to
individuals than they deserve relative to groups and relationships. In any
case, a key question to ask of any method is how well it copes with each of
the six types of agency.
A further type of distinction also looms large here. Does the method
examine one of these agents, several of these, lots of these, or perhaps all
(relevant cases) of these? McNeill (1990, 121-2) argues that in general
surveys deal with lots of people, interviews with fewer, and observation
(especially participant observation) with yet fewer. And Ragin (2000)
devotes his whole book to discussing the advantages and disadvantages of
studying one versus many. His analysis provides a further argument for the
use of multiple methods.
Note that if any subset (from one to many) of the relevant population is
examined, scientists will face questions of sampling. Is the sample biased, or
does it represent the average of the larger population, or perhaps the most
common attributes of the larger population (which can be quite different
from the average)?

2.2 What?

“What is to be studied?” Again, passive action, intentional action, and

attitudes can be distinguished. Different methods will be shown to have
differing strengths with regard to these causal processes. In the previous
chapter the ongoing debate between interprevist and positivist theorists was
discussed, and a compromise urged/celebrated in which the value of both
4. Classifying Method 105

was appreciated. It should hardly be surprising that this debate has had a
methodological side to it, and that this too should be – and to some extent is
being – guided toward compromise. A strict interprevist would argue most
obviously that scholars should study only attitudes and not actions. If it is
accepted that scholars should study both, and passive reaction as well,
methods that can cope with each of these are critical. The interpretivist point
of view has also been associated with a strong preference for idiographic as
opposed to nomothetic analysis; this in turn flows from a belief that there are
huge differences across individuals in how they perceive apparently similar
situations.5 Interpretivists also tend to argue that people can only be studied
in their natural environment. And interpretivists emphasize qualitative versus
quantitative analysis. Strict positivists take the opposing view on all three
issues. The first two of these distinctions will fall out of discussions of
‘why’ and ‘where’ below; this should encourage a suspicion that there is no
necessary correlation among answers to the three questions. The qualitative
versus quantitative distinction does not emerge from any of the 5W
questions: scholars might be especially doubtful that this distinction is
important in distinguishing good from bad research (see below).6
Regardless of what sort of causal process is investigated, there are four
criteria for establishing a causal argument. The first involves showing
correlation. Observations of the causal agent(s) should be associated with
observations of the posited result(s). Second, the direction of effect should
be established. The change agents should appear before the results (as a
general rule: there are some relations in physics where reverse temporal
causation is posited). Third, alternative explanations of the observed
correlation should be ruled out. For example, the possibility that both the
alleged causes and results are in fact the results of some other causal agent
should be investigated. [The first three of these criteria were identified by
John Stuart Mill; Palys, 1997, 274.] Fourth, intermediate (often called

5
Goldenberg (1992, 180-5) notes that scientists might expect the degree of variation to itself
vary across societies, depending on the degree of cultural conformity. He also notes that
there should be a potential for integration: positivists could establish a link between social
class and school attendance, while interpretivists investigated why some poor students stay
in school.
6
The PO method favored by interpretivists was developed by Robert Park and associates in
the 1920s, “though they knew full well that all methods had something to contribute to
sociological understanding” (McNeill, 1990, 4). By the 1960s there was a debate between
those favoring the “objective” approach of positivism, and others emphasizing the need to
understand what social life means to people. Since then it has become acceptable for a
researcher to use both methods in one project, or different methods for different projects
(McNeill, 1990, 6-7).
106 Chapter 4

intervening) variables can be identified. These can aid understanding of why

the causal relationship exists (Singleton and Strait, 1999).
A further question can be asked: is “what is to be studied” dictated
entirely by theory or not? A deductive approach looks only at phenomena
and processes within a theory, attempting to ascertain how well the theory
accords with reality. This can be distinguished from an inductive approach
that embraces phenomena and processes that should be integrated into
theory. Kuhn and Lakatos, and many other philosophers of science, have
stressed that both induction and deduction are necessary for scientific
advance. Interpretivists, notably, tend to stress induction, and positivists
deduction.7 Particular methods may combine elements of induction and
deduction.
One common suggestion is that induction be used to develop a theory,
and then deduction to test it. There is a kernel of truth here, and one that
encourages appreciation of the value of both induction and deduction. If
deduction were relied upon exclusively, theories would have to be plucked
from thin air; a rapid advance in understanding could hardly be expected
unless observations of the world were used as a source of theory. Yet if
scholars depended exclusively on induction they would be rewarded with a
plethora of unrelated observations, and no way of organizing these nor
guessing at their accuracy or representativeness. Nevertheless, the vision of
induction leading to theory, which in turn leads to deduction, oversimplifies
the research process. Indeed, people move recursively between deduction
and induction in their daily lives (Halperin, 1997, 82-3). There is scope for
induction long after a theory has been formulated, in order to suggest
additional elements that should be added to the theory (induction may be
particularly valuable in establishing the range of applicability). Scholarly
ability to judge the validity of a theory will be severely curtailed if scholars
ignore evidence outside the purview of that theory. Without continued
inductive efforts, scholars may come to accept a theory simply because they
have narrowed their gaze to only those phenomena whose study supports it.
They may thus think that they are testing a theory when in fact they only
have recourse to data that does not truly challenge the theory.8
Palys (1997, ch.3) addresses the narrower question of where research
ideas come from. He suggests that there are deductive sources of research
questions, as when scholars wish to test, extend, or limit an existing theory,
or provide a theoretical alternative. There are also inductive sources, as

7
Palys (1997, 43) notes in turn that qualitative analysis tends to be associated with
induction, and quantitative with deduction.
8
John Stuart Mill urged scientists to deplore any method that would “bestow a
disproportionate attention upon [facts] which our theory takes into account, while we
misestimate the rest, and probably underrate their importance” (in Gower, 1997, 254).
4. Classifying Method 107

when a scholar’s own experience or observations [or those of others] guide

them to ask questions. Both inductive and deductive research can suggest
further questions: other scholars may wish to replicate their findings, clarify
or challenge these, investigate anomalies, overcome conflicts, or extend
findings by analogy to other settings.
There is some correlation between this inductive versus deductive
distinction, and the longstanding philosophical dispute between empiricists
and rationalists. The former emphasized the investigation of empirical
reality, and stressed that scientists should not ignore any facts. The latter
emphasized that the world is complex, and thus scientists needed theories
that made order by focussing on only some empirical facts. Philosophers
have long sought a middle ground. They have noted in particular that (in the
words of Goethe) “every fact is already a theory.” That is, theories guide
what scientists see: they tell scientists how to classify both phenomena and
causal processes. Without theory, only a barrage of unrelated stimuli are
observed. While it was argued in chapter 1 that fairly objective classification
is still possible, it is nevertheless clear that scientists can never be entirely
empiricist. As for rationalists, a vision of a blend of nomothetic and
idiographic theory which could eventually aspire to comprehend every causal
mechanism was sketched in the previous chapter. While it can be
advantageous if individual scientists limit their gaze in the interests of
scientific specialization, the scientific enterprise as a whole should exclude
nothing from its gaze. Interdisciplinary analysis has two key roles to play
here: in identifying possible sources of theoretical innovation that lay outside
the purview of specialized researchers, and testing the range of applicability
of theories. Specialized researchers too will often develop new theories
inductively from observation.

2.3 Why?

In the case of theory, the ‘why’ question pointed toward the decision-
making process at work, including the passive decision-making of non-
intentional agents. Individual scientists might wish to investigate the nature
of the decision-making process itself, and/or the results of the decision-
making process. Note that a theory may be largely correct with regard to
either one of these, and yet mistaken with regard to the other. Not only will
particular methods prove to be best suited to answering one type of question,
but methods are differentially applicable to different types of decision-
making process.
“Why do scientists study what they do?” This broad question subsumes
all of the analysis in this chapter, as well as that in chapter 5. The purpose of
this chapter is to type methods in terms of what they do, rather than in terms
108 Chapter 4

of scientific bias. After developing a typology of scientific biases in chapter

5, chapter 6 will explore whether some methods are more immune to these
than others. For the purposes of this chapter it is worth noting again that
scholarly communities are likely less prone to methodological flexibility than
the analysis of this chapter would suggest that they should be.

2.4 Where?

As with theory, the ‘where’ question can be asked both literally and
figuratively. In literal terms, two key distinctions can be identified. First,
phenomena can be analyzed in one place or as they move through space.
Second, either of these types of analysis can be performed in a natural setting
or in an artificial setting (though only limited movement through space is
generally possible in an artificial setting). With respect to the first distinction,
the advantages of each largely hinge on whether the theory refers to
movement (or should do so). With respect to the second, artificial settings
allow the researcher to control extraneous variables and thus isolate a
particular cause-effect relationship; the danger is that agents may behave
differently in an artificial setting. There is also a null possibility, wherein the
research is not spatially dependent at all. Mathematical modelling is
arguably unaffected by the spatial location or movement of the researcher.
Figuratively, the ‘where’ question echoes the nomothetic versus
idiographic distinction drawn in chapter 3. The obvious distinction here is
that the idiographic investigator wishes to establish the relevance of a
theoretical proposition in a constrained set of circumstances, while the
nomothetic researcher is concerned with showing a broad applicability.
Note, though, that the nomothetic researcher could have recourse to
idiographic methods, illustrating the relevance of a certain proposition in
quite different sets of circumstances. The idiographic researcher may wish
to employ nomothetic methods to establish the limited range of applicability
of a theory. Palys (1997) suggests that the two approaches are inherently
complementary: examination of particular cases should bear out (or not)
general rules, while the analysis of generalities should suggest where
idiosyncrasies lie. “Particulars will suggest, circumscribe, and test
generalizations, and in turn the latter will help find and shed light on the
former” (Bunge, 1998, 25).

2.5 When?

Different methods may be best suited to analyzing certain sorts of time

path. Scientists will want to be especially careful here (though it is a
4. Classifying Method 109

concern in all cases) that a particular method is not biased toward a

conclusion that a particular sort of time path is at work.
The literal ‘when’ question was equivalent to ‘where’ with respect to
theory, but is of greater import here. Do scientists analyze a set of
phenomena at one point in time, continuously through time, or at several
discrete points in time? Do they analyze all phenomena at the same time(s)
or at different times? Continuous time has obvious advantages for studying a
process of change: recall the importance of establishing temporal priority and
identifying intermediate variables for any causal argument. Analysis at
particular points in time is generally easier.

2.6 How?

In the previous chapter, it was found that, with respect to theory, answers
to the ‘How?’ question had been covered within the answers to the 5W
questions. The same result, not coincidentally, holds here. To ask “how
does a scientist investigate?,” is to ask the questions above. The potential
biases in scientific practice will, of course, be investigated in chapter 5.

3. QUALITATIVE VERSUS QUANTITATIVE

ANALYSIS REVISITED

Hall and Hall (1996,35) argue that the furore over the quantitative versus
qualitative distinction is dying down: “...nowadays most writers on
methodology avoid such a distinction and stress instead the strategy of
mixing methods.” Tashakkori and Teddlie (1998, 5) ground their book in a
belief that these paradigm wars are effectively over. However, for decades
defenders of one approach to research savagely attacked the alternative, and
there are still many scientists (but fewer philosophers) who staunchly believe
that there is only one right way to do research. Moreover, some have
stressed the inherent incompatability of the two approaches, arguing that they
could or should never be combined. This has not prevented an increased
pragmatic interest in compromise. Tashakkori and Teddlie (1998, 11) echo
the Golden Mean approach outlined in chapter 1, arguing that scientists
should accept both quantitative and qualitative approaches because each has
been long used, accepted, urged, and useful. The relative importance of the
two may nevertheless vary by research question. Interdisciplinary scholars
are likely to be especially open to both, though specialized scholars should
be as well.
The quantitative approach emphasizes evidence that can be quantified:
the number of molecules, people, or widgets. The alternative is to focus on
110 Chapter 4

that which cannot (easily) be quantified: appearance, anecdotes, detailed

recording of a particular set of societal or interpersonal interactions.
Extreme advocates (and the debate encouraged extremism) would
completely exclude the other sorts of measures from their purview. Yet none
of the 5W questions has suggested a quantitative/qualitative distinction
(though intentional agency undoubtedly opens up more scope for qualitative
analysis), and thus the mixing of these methods as reported by Hall and Hall
should hardly be surprising.9
The quantitative/qualitative distinction has often been associated in the
literature (see Palys, 1997, Tashakkori and Teddlie, 1998, Seale, 1999,
Oakley, 2000) with a variety of other distinctions: positivist versus
interpretivist, nomothetic versus idiographic, focus on observables versus
non-observables, deductive versus inductive, non-intentional versus
intentional agency, and concern with ‘why’ rather than ‘what’ or ‘how’. In
all of these cases the correlation is weak. There are qualitative elements to
action, and attitudes can potentially be quantified. Oakley (2000, 28) notes
that interpretivists often quantify: they just use words like “most” rather than
percentages, and speak of correlations without expressing these
mathematically. One can discuss qualitative relationships with wide
applicability, and quantitative relationships with narrow applicability, though
extreme interpretivists would nevertheless suggest that only context-
dependant hypotheses are possible. As Palys (1997) notes, most positivists
have come to accept the need to study some unobservables. Positivists have
certainly long had recourse to induction, while Denzin and Lincoln (1998,
xii), in their overview of qualitative inquiry, note that this can be very
deductive, with set questions at the outset (notably, they use the descriptor
“positivist” for this type of research).10 Seale (1999, 39) recognizes that
qualitative researchers often investigate ‘why’ questions; quantitative
researchers in turn hardly ignore ‘what.’ Many theories in natural science
focussed upon non-intentional agents – such as the germ theory of disease,

9
Abbott (2001, 6-12) applies the concept of fractals, and argues that distinctions such as that
between quantitative and qualitative are repeated at lower levels: quantitative versus
qualitative kinds of quantitative research could be identified, for example. Rare will be the
extremists who make the same choice at each level. The self-definition of scientists will
depend on the approach of those to whom they compare themselves. Attitudes of quite
different scholars will on certain issues appear quite similar.
10
Should scientists ask research subjects what they think, and involve them in research
design? Qualitative researchers often say that it is unethical not to tell subjects what is
happening; Oakley (2000) responds that it is quite ethical to, for example, have control
groups who are not told they are getting the placebo. Consulting research subjects can
often be valuable in suggesting arguments the researcher might otherwise miss. But there
is the danger of seeing things only as subjects do. Oakley (54) also notes that scientists
may not wish to hurt subjects’ feelings with their conclusions.
4. Classifying Method 111

or cell theory – are inherently qualitative (Langley et al., 1987), and

qualitative evidence in support of these could hardly be objected to. While
the correlations are weak, the fact that both sides in all of these other
dichotomies have been found above to be legitimate and important forms of
scientific inquiry suggests that both quantitative and qualitative research
have their place as well (and the same argument can be made in reverse).11
Certainly, attempts to identify two diametrically opposed types of science
along several dimensions have failed signally.
Tashakkori and Teddlie (1998, 10) identify other beliefs associated with
interpretivism and/or a qualitative approach: that reality is constructed and
multiple, that knower and known are inseparable (and thus objectivity
impossible), that inquiry is value-laden, and that scientists cannot distinguish
cause from effect in a world where everything affects everything else. With
regard to the suggestion that science cannot be perfectly objective, it will be
argued in chapter 5 that this and other limitations in scientific practice do not
prevent scientists from achieving scientific advance. Moreover, Tashakkori
and Teddlie (1998, 13) suggest that one key to bridging the gap between
qualitative and quantitative approaches is a common recognition that inquiry
of both types is value-laden and subjective. If scientists ignore the
limitations of their favored approach, it is all too easy to view flawed
alternatives as beyond the pale. To be sure, advocates of quantification
would point to the greater objectivity of their evidence. If scientists wish to
count the number of students in a school at a point in time, different
researchers will likely reach similar conclusions. If researchers wish to
understand what school means to the students, there will likely be greater
differences across researchers. Advocates of qualitative analysis will retort
that scientists should be concerned with the meanings that individuals attach
to situations, and these can only be investigated subjectively: attempts to
quantify meaning must serve to distance scientists from those being studied.
Moreover, even positivists now agree that no bit of scientific evidence is
conclusive, and thus there is always scope for qualitative analysis of any
piece of quantitative evidence (Palys, 1997, 396). After performing

11
Many, but far from all, feminist researchers have embraced the qualitative cause. Oakley
(2000) bemoans, much better than I could, the limitations this places on feminist research,
and the not-so-subtle pressures to conform. She notes, in particular, that one can hardly
convince others of gender discrimination without some quantitative evidence. Oakley
herself had begun her career as an advocate of qualitative research but came to appreciate
the advantages of combining the two. She thus also came to realize that the longstanding
emphasis on quantitative research was not simply a relic of male-dominated science.
112 Chapter 4

statistical analysis, for example, scientists should ask the qualitative question
of how important the results are.12
Seale (1999) speaks of a tendency of interpretivists to be wary [explicitly]
of causality, for fear that it compromises free will. Tashakkori and Teddlie
(1998) attribute a similar attitude to qualitative researchers. In both cases,
the correlation is far from perfect, with many qualitative and/or interpretivist
researchers openly embracing causality. Even those who do not explicitly do
so end up describing how some phenomena influence others. Interpretivists
tend to prefer the word “process” to the word “causal”, feeling that this
better signals the interrelationship among diverse causal links, but they must
still deal with causation. While it is true that (almost) every phenomenon
influences and is influenced by every other, this hardly prevents scientists
from identifying each causal link, nor making causal arguments (as was seen
in chapter 2).
Quantitative analysis especially is undertaken in both natural and
artificial settings. And we shall see that there are advantages and
disadvantages associated with each of these. In psychology especially there
was nevertheless a heated debate along these lines (Tashakkori and Teddlie,
1998).
Is there a real world out there, or is reality constructed by individuals?
Positivists think that fairly stable causal relationships exist out there to be
uncovered by research. Constructivists think that humans collectively create
reality: there is no “cheating” or “single parenthood” out there, but rather the
definition and implications of these depend on social and cultural
circumstances. But note that these are causal arguments, and scientists could
hope to understand how various cultural attitudes affect single parents. And
thus scientists can investigate the degree to which a certain causal
relationship is conditioned by realizations of other phenomena. Hammersley
and Gomm (2000, 157) note that the claim that there is no one reality is itself
a universalistic claim; there is a long-recognized inconsistency involved in
asserting absolutely that all knowledge is relative. Palys (1997) celebrates
critical realist philosophy, which bridges the gap between the two, and urges
the investigation of diverse causal links (while Tashakkori and Teddlie,
conversely, urge a common acceptance of multiple constructed realities).
In terms of the methods that will be discussed below, quantitative
researchers favor experiments, surveys, statistical analysis of secondary data,

12
Various authors in Pole and Burgess (2000) suggest that researchers from differing cultural
backgrounds are more likely to reach different conclusions using qualitative than
quantitative analysis. Moreover researchers studying the words or actions of those from
different backgrounds can easily misunderstand these (and often ask questions that make
sense only to them). There is also a difference in the popularity of qualitative research
across cultures.
4. Classifying Method 113

content analysis, and structured observations (where particular acts are

counted), while qualitative researchers favor participant observation, in-
depth interviewing, case studies in general, biography, and focus groups
(Oakley, 2000, 26). It will be seen that each of these has strengths and
weaknesses. For now it is sufficient to recognize that exceptions abound in
which advocates of one approach have nevertheless used the methods of the
other. Goldenberg (1992, 322) thus asserts that (at least) experiments,
interviews, and historical methods are not inherently quantitative or
qualitative (nor positivist nor interpretivist). And note that the boundaries
are not always clear: when does an interview with a fixed set of questions
cease being qualitative?

4. EVALUATING METHODS

Before proceeding to place individual methods within the typology of

strengths and weaknesses of methods, it is useful to first establish criteria by
which methods might be evaluated. Then, rather than just saying that
method X can be applied to question Y, it will be possible to gain some idea
of how well method X investigates question Y. The utility of a method does
not just depend on the range of questions to which it can be applied, but the
confidence that scientists can have in the results it generates. Fortunately,
scholars of methodology have long since identified three criteria by which
the trustworthiness of any method – or, indeed, any application of any
method – might be judged: reliability, validity, and representativeness.
Unfortunately, there is much room for judgment in applying these criteria,
and thus interpretivists and positivists have decried the reliability and
validity of the others’ methods (see Goldenberg, 1992).
More seriously, some interpretivists have questioned whether any criteria
should be used, and if so whether validity and reliability are reasonable. As
Seale (1999, 43) notes, a belief in multiple realities is at odds with efforts to
discern the likelihood that given research reflects reality. He concludes,
though, that interpretivists in practice do appeal to standards of validity and
reliability (and indeed make interesting observations about how these could
be achieved). Oakley (2000) summarizes the efforts of several qualitative
researchers to develop such criteria. Notably, half of these develop
quantitative criteria and the other half qualitative criteria by which
qualitative research might be judged.13 Tashakkori and Teddlie (1998, ch.4)

13
Together they urge authors to be clear about the questions addressed, the methods used, the
definitions of key concepts, the context of research, the relationship between researchers
and subjects, the extent of interaction with subjects, the relationship of the evidence
compiled to alternative hypotheses, and the generalizability of results (among
114 Chapter 4

perform a similar exercise to that of Oakley and conclude that there are
greater similarities than differences in how quantitative and qualitative
research are evaluated.
Reliability: Does the method yield similar results when applied to
identical situations? With respect to individual pieces of research, questions
can be asked such as “could/has another researcher duplicated the results?,”
or even “can the same researcher reproduce the results?” At the level of a
particular method itself the concern is with the degree to which replication is
possible. Is the research closely tied to the idiosyncrasies of an individual
researcher, or to the idiosyncrasies of a particular research situation, or can it
reasonably be expected/observed that another researcher can do very similar
research and gain very similar results?
Seale (1999, ch.10) notes that qualitative studies are hard to replicate.14
Nevertheless, reliability is still an important criterion, and this can be greatly
enhanced by various practices developed by qualitative researchers.
Reflective methodological accounting requires the researcher to carefully
report what evidence was collected and how, and how this was interpreted
(Seale emphasizes the importance of distinguishing the recording of
information from the drawing of inferences). Peer auditing involves having
other researchers review the research and comment on issues of reliability.
Seale discusses the heated debates over early ethnographic studies – such as
Margaret Mead’s study of adolescence in Samoa – and notes that these early
researchers did not leave detailed notes from which the reliability of their
studies might be judged.
The importance of reliability should be obvious. If a method or particular
piece of research is not reliable, scientists will and should lack confidence
that the results obtained would be obtained again, and can thus hardly be
confident that they reflect reality. Note, though, that several examples of
research which reach similar results, even if the research is not strictly
replicable, may still help scientists to increase their confidence in reliability,
and perhaps also guide them toward appreciating the range of applicability of
a theory.
Reliability has of necessity been discussed at the level of both methods
and particular pieces of research. While some general range of reliability can
be attributed to a particular method, it is important to appreciate that
particular applications of that method will differ in terms of reliability.

other criteria). Silverman (2000, 178-85) stresses in turn the importance of seeing if any
posited relationship applies over the entire data set, and striving to explain deviant cases.
14
Strict constructivists argue that replication is both impossible and undesirable (because
there are multiple realities). Seale retorts that to abandon replicability leaves no special
place for science relative to fiction.
4. Classifying Method 115

Singleton and Strait (1999), for example, suggest that reliability can be
enhanced by pretesting one’s research instrument, adding items to the
research project which seem important, and dropping items which the pretest
suggests have no impact on the outcome.
Validity: While reliability asks whether there is consensus across
applications of a method, validity deals with the more central question of
whether the results of research reflect an external reality. Note, though, that
if research is not reliable it will almost always be invalid as well. Since by
definition scientists can only approach understanding of external reality
through research, there is necessarily more room for subjective judgment in
evaluation of validity (with some scholars doubting the very existence of an
external reality). Some communities of scholars may thus be tempted to
revel in the reliability of their methods, while ignoring questions of validity.
There are some key questions that can and should be asked. First, does
the research actually investigate the links posited by theory – or, in the case
of induction, those that could be posited by theory? If the results of research
do not accord with such links, as might occur, for example, if the data do not
accurately represent the phenomena in question, scientists should be
sceptical that the results reflect an external reality. For qualitative research it
can be asked whether the explanation provided is grounded in the
observations, or perhaps only consistent with these. Second, it can be asked
whether the research deals with alternative hypotheses. If not, scientists
should worry that the results can be explained in diverse other ways, and thus
that external reality does not accord with the particular theory favored by
researchers.
Goldenberg (1992) treats the two previous questions as tests of internal
validity, for they focus upon the relationship between research design and
theory, and the following questions as tests of external validity. Third, it can
be asked if the results are robust. Note that when concerned with reliability
scientists should aspire to getting similar results from similar studies; here
they should aspire to getting similar results from different studies. This can
be an argument both for using one method across many situations, and for
using many methods. Scientists should not, though, look only for situations
where the theory predicts similar results from different situations, but also
look for where it predicts different results, and ensure that these indeed
follow (Singleton and Strait, 1999). Fourth, it can be asked if the results
allow scientists to predict. Fifth, if a scientific community considers other
theories or measures to be valid, they can look for consistency with these.
Different questions may be more appropriate for different types of
investigation. Nevertheless, a community of scientists should be careful of
focussing on only a subset of these: economists long cherished the predictive
power of their theories, even when results were loosely connected to theory,
116 Chapter 4

only to find that predictions became poor due to changes in un-specified but
relevant phenomena in the 1970s.
Representativeness: Virtually all methods require some sampling of a
larger population, and thus raise questions of the representativeness of the
sample. One problem with Mead’s study of adolescence in Samoa was that
her sample was small and potentially biased (Seale, 1999). Techniques of
achieving representative samples for surveys are well developed; problems
loom larger for other methods. It is noteworthy that most books on method
have an introductory chapter on sampling that precedes discussion of
particular methods. Meltzoff (1998, ch. 4) lists several common types of
unrepresentative sample, and notes that for methods like interviews and
experiments that require a significant time commitment from subjects,
representativeness is particularly difficult to achieve.

5. CLASSIFYING PARTICULAR METHODS

WITHIN THE TYPOLOGY

“There is never a single way to conceptualize or address a problem. Nor

is there any great consensus about how successfully various methods deal
with particular issues” (Goldenberg, 1992, xiii). Hopefully the
classifications of phenomena and theory types in previous chapters will aid
in the conceptualisation of research questions. Researchers would then
benefit from a handy guide as to which methods are best suited to the
question at hand. While there is no assumption at the outset of this analysis
that the results must assume a particular form, such as urging methodological
flexibility or orthodoxy, casual empiricism suggests that for at least some
research questions multiple methods will prove appropriate.
Of any method it can be asked whether it reliably and validly deals with:
x non-intentional and intentional agents; individuals, groups, and
relationships (with particular attention to the number investigated)
x passive causation, actions, attitudes
x the four criteria for identifying a causal relationship: correlation, temporal
order, evidence against alternative hypotheses, intermediate variables
x decision-making processes and/or the results of decision-making
processes
x phenomena and processes outside of a particular theory
x nomothetic versus idiographic relationships
x spatiality: whether it allows study of movement through space (with a
focus on whether it studies phenomena through space or in one place)
4. Classifying Method 117

x change through time (with an emphasis on whether it allows continuous

or static coverage)
x various time-paths: equilibrium (same or different), dynamic, stochastic

Five of these dimensions replicate those in the 5W typology of theory:

they indicate what kind of theories a method can investigate. The 5W
questions also guided the identification of new dimensions: these identify
how well certain key tasks can be performed by a method. In answering both
sorts of question, it will often be found that a certain method is simply
incapable of dealing with some cells in this typology. At other times a
method sheds indirect or partial light on a particular question. More
generally methods can perform certain tasks but imperfectly. When these
imperfections (or lack thereof) are particularly remarkable, the reliability,
validity, and/or representativeness of applications of these methods will be
explored. Often imperfections can be alleviated by the use of information
generated by a different method. When problems of reliability or validity are
present the likely direction of bias can at times be identified.
Several “stand-alone” methods are first surveyed in alphabetical order.
Each of the questions identified above is asked of each method. The results
of this analysis are summarized in Table 3.1: note that the table cannot
capture many important caveats about the characteristics of various methods.
A handful of key “combinatorial methods” are then investigated, again in
alphabetical order. Rather than repeating the detailed questioning for the
latter, some general observations are made.

5.1 Classification (especially evolutionary)

Type of agents (and number investigated): All types of agent can

potentially be classified.
Type of causation: Evolutionary classification is a combination of theory
and method; by classifying a species or a gene, scientists are suggesting what
it evolved from, and pointing the way toward understanding how. Scientists
must be wary of biases in both theory and classification. Classifications
distinct from theory cannot speak to causation. They are, though, essential
building blocks in this process, for scientists should not say that X causes Y
unless they have first classified X and Y. To say that poverty encourages
economic crime, for example, is meaningless unless poverty and crime can
be distinguished from other states and behaviors.
The four criteria for identifying a causal relationship: Classification
allows scientists to seek correlations by other methods. These correlations
will only be as reliable as the classifications themselves. If poverty is not
118 Chapter 4

adequately distinguished from wealth, correlations with crime may be

undeservedly weak.
Decision-making processes?: Classification alone can tell scientists
nothing about decision-making processes.
Phenomena and processes outside of a particular theory?: There is an
especial danger that in classifying scientists ignore items that do not fit, or
place them in a category in which they do not belong and forget about them.
Imagine a periodic table in which helium was slipped in with hydrogen. But
scientists may be guided to investigate items that do not fit.15 Recognition
that helium had unique properties could stimulate further research.
Nomothetic versus idiographic relationships: Classification sets the stage
for both.
Spatiality (studies through space?): Classification systems can have a
spatial element. Species, for example, can be classified in terms of
ecological range.
Change through time (continuous or static coverage?): Evolutionary
classification can obviously aid scientific understanding of change over time
(with the danger of oversimplifying this). Even atheoretical classifications –
say, of occupations – allow scientists to identify (subject to possible biases)
changes through time that need explanation.
Various time-paths: Classification alone cannot answer such questions.
General Comments: The reliability and validity of classifications may too
readily be assumed. “Many classification systems are artificial constructs,
laden with jargon, which often do little to advance understanding” (Valiela,
2001, 11). Scientists must ask whether a particular system is artificial or
reflects reality. In particular, scientists must beware that classification serves
social and political purposes, and may thus reflect biases of those who create
and use classifications; in looking for possible biases it is useful to ask what
use is made of particular classifications (Bowker and Star, 1999, 9).

5.2 Experiments

Type of agents (and number investigated): Experiments can deal with

non-intentional or intentional agents, and individuals or relationships.
Analysis of group behavior is generally though not always infeasible.
Experiments usually deal with small numbers of agents, and thus questions
of their representativeness can loom very large (especially in human science,

15
Bowker and Star (1999, 10) note that ideally a classification should exhibit exhaustiveness,
mutual exclusivity, and a consistent and unique organizing principle. None of the
classifications they studied met this ideal.
4. Classifying Method 119

given the propensity to rely on college students as subjects).16 Natural

experiments, in which particular occurrences are analysed as they occur in
the real world, allow greater attention to group processes, as well as non-
intentional agents such as planets or volcanoes, but at the cost of limiting the
researcher’s ability to control realizations of other variables.17
Type of causation: Experiments are very good at identifying both passive
(such as chemical reactions) and active (such as intentional reactions to
stimuli) causation. They are less successful at measuring changes in attitude,
for these must either be inferred from behavior or interviews.
The four criteria for identifying a causal relationship: Experiments are
very good with respect to all four of these, though there is some potential for
bias in how competing hypotheses are tested. Polkinghorne (1996, 28) notes
that electrotweak theory was ignored for years, as researchers interpreted
experimental evidence in a way that supported previous theories; once
certain mathematical advances in theory had occurred, experimental
evidence was largely viewed as supportive of the theory. Experiments are
potentially highly reliable: they can be easily repeated with all sorts of subtle
changes to research design. This is important, for experimental results –
especially those regarding intentional agency – might be confounded by a
variety of un-recognized variables: idiosyncrasies of researchers or subjects
or raters or apparatus, chance occurrences (such as a plane flying overhead),
and procedural and situational variables (such as how instructions were read)
(Metzloff, 1998, ch. 5).
Decision-making processes?: Experiments are best at establishing links
between cause and effect. “While the other research designs provide useful
information, the experimental design provides the most rigorous test of a
hypothesis which specifies that X causes Y” (Bouma and Atkinson, 1995,
126). They can illuminate some aspects of decision-making, such as the
degree to which people are swayed by the views of others, but notably only
to the degree these aspects of decision-making can be operationalized in the
experimental setting. That is, the experimental design must include some
way for participants to act on their decisions.
Phenomena and processes outside of a particular theory?: Experiments
are primarily a deductive tool: the subject is manipulated in a particular way

16
“...in education and the social sciences generally, large groups are needed if the many
variations and ambiguities involved in human behavior are to be controlled” (Bell, 1999,
15).
17
Singleton and Strait (1999) stress that quasi-experiments violate the experimental rule of
changing one variable at a time. Oakley (2000, 309-12) gives examples in medicine, social
work, and income support where uncontrolled quasi-experiments yielded the results
desired by researchers (that the programs worked), while later controlled experiments
yielded the surprising result that the programs in question were much less successful. She
concludes that in general studies without control groups show much higher success rates.
120 Chapter 4

and results measured. Krieger (1997, ch. 3) worries that rare events may be
missed by exclusive reliance on experiments, and thus advocates that
experiments be supplemented by real world analysis of natural processes.
With respect to phenomena, the experiment can only disclose the importance
of phenomena unexpected by the researcher if some sort of mistake is made
in experimental design, either through the introduction of some unplanned
element, or through failure to appreciate how the experimental environment
would influence results. Such serendipity is surprisingly common in the
history of natural science. Sir Alexander Fleming, for example, discovered
penicillin by observing a bacteria-free circle surrounding a mold growth that
had contaminated a culture of staphylococci. As experimental controls are
enhanced by technology, induction may rely increasingly on other methods.
Nomothetic versus idiographic relationships: Both are quite possible, for
experimental design can be manipulated to see how a particular cause-effect
relationship is affected by any number of circumstances. Experiments in
human science are usually conducted by those confident of general patterns
of cause and effect (McNeill, 1990, ch.3).18 Scientists must be careful of
generalizing from a small number of subjects studied in an artificial setting.
Spatiality (studies through space?): Experiments are usually constrained
to operate in an artificial and small laboratory setting. There is thus a huge
validity problem: do agents behave the same in their natural setting? This is
usually not a great concern with non-intentional agents, though some
biologists have doubted that experimental results concerning the effect of
UV radiation on amphibian reproduction support valid explanations of
decreased reproduction in the wild (Longino, 2000, 267-8), and Rothbart
(2000, 83) argues that chemists often seek observational verification of
experimental results. The problem looms much larger for intentional agents.
Other methods, such as participant observation, can then provide a useful
check on the validity of experimental results. Moreover, scientists cannot
follow spatial behavior requiring more space than the laboratory. Note,
though, that some experiments occur in fairly natural settings. A drug trial is
an experiment in which some people are given a drug and some a placebo:
beyond visits to the clinic they live their normal lives. An experiment on
spatial movement could be imagined in which researchers give people
misleading directions and see how they cope. But even in natural settings,
the mere knowledge that one is being observed can cause a change in
behavior: the “Hawthorne effect” refers to studies at GE’s Hawthorne plant

18
Advocates of qualitative research, including feminists, have often been hostile to
experiments (partially due to ethical concerns; because experiments involve control and
manipulation they seem an abuse of power, but Oakley argues that ethical experiments are
possible). Oakley (2000) notes that we all experiment in our daily lives – when cooking or
gardening for example – and should thus be willing to experiment in science as well.
4. Classifying Method 121

in the 1930s that found that workers increased effort whenever they were
being observed.
Change through time (continuous or static coverage?): Experiments
occur at a point in time. They are thus limited in their ability to analyze
changes that occur over lengthy periods. The experimental efforts of
geneticists have nevertheless informed the study of evolutionary processes,
by identifying what sorts of change are possible. Natural experiments are a
bit different, allowing scientists some scope for analyzing revolutions or
planetary formation.
Various time-paths: Experiments are best at identifying simple cause-
effect relationships. They can identify equilibria in very simple systems,
such as setting up a pricing game and seeing how a market-clearing price is
established, but in general are poor at identifying complex interactions.
Reliability: With intentional agents researchers must worry about
signalling the desired result to subjects. As well, subjects sometimes leave
during an experiment, biasing the results in perhaps unknown directions.
Careful design (for example distancing researcher from subject) and
replication (with the same and different research protocols) can help. Even
with non-intentional agents scientists should appreciate that there are a wide
variety of experimental set-ups possible, and thus experimentation cannot be
reduced to a routine process (Gower, 1997, 11).
General Comments: As Gower (1997, 10) notes, with citations from texts
in chemistry and physics, experiments are often treated as “the” method in
natural science. Obviously, experiments have many strengths, especially for
the analysis of non-intentional agents. Researchers may be oblivious to the
role played by various elements of experimental design in generating a
particular result. Thus the results of experiments are “rarely decisive,” and
“often ambiguous and sometimes opaque” because experimental results must
be transformed into broader theoretical arguments, and can generally be
interpreted in multiple ways (Gower, 11).19 It is generally useful, then, to
seek corroborating evidence from other methods. While biologists have had
much success in this regard, psychologists often find that experiments
produce different results from observation (Fuller, 1997, 13-14).

19
Gower goes on to maintain that philosophers of science have come to appreciate “the
practical difficulties in obtaining ‘good’ data from experiments...” (1997, 18). Yet
elsewhere Gower himself seems to treat experiment as “the” scientific method. Collins and
Pinch (1993, 3) are much more critical: “The problem with experiments is that they tell
you nothing unless they are competently done, but in controversial science no-one can
agree on a criterion of competence. Thus, in controversies it is invariably the case that
scientists disagree not only about results, but also about the quality of each other’s work.”
Valiela (2001, 13) counters that “The issue of experimental artifacts is real but something
of a red herring”; scientific advance suggests that problems can be overcome with
appropriate experimental design.
122 Chapter 4

5.3 Interviews

Type of agents (and number investigated): Interviews cost much more

than surveys, and thus tend to involve fewer individuals, though more than
Participant Observation (below). They can speak only indirectly to
relationships (unless there are joint interviews) and group processes.
Type of causation: Interviews attempt to elicit attitudes. They might in
this regard identify attitudes that encourage(d) particular actions. But people
often do not know why they act as they do (Goldenberg, 1992, 352). Some
questions can elicit insights into constraints/incentives imposed by
impersonal agents.
The four criteria for identifying a causal relationship: Interviews might
establish a correlation between attitude pairs, or an attitude and action, if
enough subjects are asked questions about both that meaningful statistics can
be generated. They are limited in their ability to identify temporal priority,
in large part as interviewees may mislead on this and other points, either
purposely or through faulty memory. The ability to distinguish among
hypotheses and/or identify intermediate variables is limited and dependent on
the researcher asking appropriate questions. Note that there are limits to how
long an interview can last.
Decision-making processes?: People generally find it easier to answer
‘how’ than ‘why’ questions; more in-depth interviewing is often required to
comprehend decision-making processes. When interviewers ask why people
did something, they are usually given a chronological account lacking detail
on how decisions were reached. There is always a problem that interviewees
may mislead about why they acted/thought as they did/do. Even scientists,
queried about scientific practice, will give the answer that they think is right,
rather than describe what they actually did (Fuller, 1993, 194-5)
Phenomena and processes outside of a particular theory?: Some
interviews are more open in allowing respondents to respond at length in
ways not imagined by the researcher, or by pre-existing theory. There is,
though, a loss in objectivity relative to closed interviews.20 Feminist
researchers in particular stress the advantages of mutual exchange between
interviewer and subject (Palys, 1997, 187). Others suggest that interviews
are inherently a relationship, and thus tell us about the relationship rather
than the subject (May, 1997, ch. 6). Surveys, with a less personal
relationship between researcher and subject, may provide an antidote.

20
Oakley (2000) argues that when women interview women, the principle of respecting
one’s subject may introduce an even greater degree of bias toward identifying with the
subject’s point of view. Nevertheless, interviewers may forget the power relationship
inherent in interviews.
4. Classifying Method 123

Nomothetic versus idiographic relationships: Interviews necessarily

involve small numbers of people, and thus generalizations require integration
of results across many studies. Interviews can, though, usefully identify
idiosyncrasies (that is, realizations of other phenomena) that prevent a
general rule from operating.
Spatiality (studies through space?): Interviews occur at one or a few
points in space. Researchers must rely on the interviewee’s memory and
honesty (note that they may mislead themselves as well as researchers) in
constructing spatial processes.
Change through time (continuous or static coverage?):Interviews occur
at one or a few points in time. Researchers must rely on the interviewee’s
memory and honesty (note here and elsewhere that subjects may mislead
themselves as well) in constructing continuous processes.
Various time-paths: Interviews on their own would face great difficulty in
identifying the nature of causal processes.
General considerations regarding reliability and validity: Interviews face
greater problems in terms of subject bias than Participant Observation, but
lesser problems in terms of researcher bias. Still, interviewers can all too
easily encourage subjects to respond in desired ways. May (1997, ch.7)
suggests that interviews may elicit different responses depending on the
gender of the interviewer. Drawing connections across many different
studies reduces both problems.
Note: One extended form of the open-ended interview is the life history.
This has the advantage over the autobiography of emphasizing questions of
interest to the researcher. There are, though, huge problems of subject bias,
for individuals tend to understand their lives in self-serving ways. The life
history can allow scientists to appreciate the multifaceted nature of any
individual, and may guide scientists to look for similar personal
characteristics (see chapter 2) and their effects in others. Scientists may be
able to draw general conclusions from many cases, but must worry about
representativeness: life histories of interesting people are more likely to be
recorded (see McNeill, 1990, ch.4).
To overcome problems of researchers biasing results through their choice
of question, some favor narrative analysis where people are asked to tell
their own stories, and the researcher structures these around themes. This
may attenuate but not eliminate researcher bias, while limiting the
researcher’s ability to generalize across cases. Bell (1999, 16-7) suggests a
further advantage: that narrative is a different way of knowing, and that
humans can naturally appreciate others’ stories, even if their lives are quite
different from ours. While narrative may have rhetorical value it should not
be viewed as a substitute for explicit causal analysis.
124 Chapter 4

5.4 Intuition/Experience

Type of agents (and number investigated): Human intuition provides each

individual with some unique insight into themselves as individuals, and thus,
hopefully, into the inner lives of all humans. It may also provide unique
insight into relationships and groups. Individuals feel intuitively drawn to
some and repelled by others, and learn from experience how useful or
misleading such intuitions are. Individuals also “know” from experience
how various impersonal forces affect them. More reliable understanding of
non-intentional agents can generally be obtained from other methods.
Individuals must, in any case, worry that their subconscious mind misleads
them about many things.
Type of causation: Human intuition gives individuals fairly accurate
indications of their feelings at a point in time, but may be quite misleading in
terms of the causes of either attitudes or actions. Notoriously, humans tend
to attribute bad outcomes to circumstances but good outcomes to their own
actions. More generally, humans have difficulty seeing themselves from
outside. Inevitably intuition exaggerates some causal links and neglects
others.
The four criteria for identifying a causal relationship: Individuals may be
able to correlate feelings and actions (felt rage; struck someone), and perhaps
establish temporality, but will be poor at establishing intermediate variables
and especially at dismissing alternative explanations.
Decision-making processes?: Individuals have unique access to their own
decision-making process. They can generally be more confident of this than
of what another tells them. But individuals may still mislead themselves.
Phenomena and processes outside of a particular theory?: Intuition is an
invaluable inductive tool, and often suggests hypotheses to researchers that
would not have resulted from deductive logic. In particular, intuition may at
times guide scientists to extraneous phenomena which experience has taught
them to be important. However, intuition can be misleading: psychological
disorders often stem from mistaken connections drawn as children, and one
may repress causal connections that one does not wish to recall. Scientists
should not ignore intuition, but should be aware of possible biases.
Nomothetic versus idiographic relationships: Intuition is inherently
idiographic, but as noted above individuals may draw broad lessons from
limited experience.
Spatiality (studies through space?): Intuition/experience is grounded in
human movement through space. Individuals may be biased in their
treatment of places that were the site of particularly good or bad memories.
4. Classifying Method 125

Change through time (continuous or static coverage?): Intuition reflects

how humans move through time. Note that different individuals may
perceive the passage of time quite differently.
Various time-paths: Intuition is generally in the form of simple
cause/effect relations, though it may well guide individuals to view the world
as unchanging or chaotic or progressive.
Reliability and Validity: Obviously, intuition faces some of the greatest
problems with respect to both reliability and validity. Nor can validity be
assured by comparison of multiple intuitions, for humans may all mislead
themselves in similar ways. Intuition can serve a vital inductive role, and
likely has played some such role in most scientific theorizing. Scientists are
guided to validate intuition with other methods. Far more than with any
other method, scientists should be wary of exclusive reliance on intuition.
Scientists take this position to extremes, and almost never mention the role
that intuition may have played in their research: this obscures the true nature
of the scientific process. Beyond this general role for intuition, some aspects
of reality may not be accessible in any other way.

5.5 Mathematical Modeling

Mathematical models may be used for statistical analysis. The problems

inherent in statistical analysis will be discussed below. But mathematical
models (and models more generally) are often used even when statistical
testing is not foreseen.
Type of agents (and number investigated): Scientists can model anything,
though there is a strong bias toward modelling quantifiable relationships..
Type of causation: Scientists can, again, model anything. The huge
danger here is that, if scientists are not going to expose their model to
external reality they have very limited means of judging its validity.
Modelling can, though, aid scientists in identifying logical inconsistencies.
The four criteria for identifying a causal relationship: Models will
identify what is put into them. They will ignore intermediate variables
unrecognized by the researcher, and provide no means on their own for
dismissing alternatives that are not logically incoherent.
Decision-making processes?: Decision-making processes may be
modelled. It has generally, though, proven easier (and thus more enticing) to
model rational decision-making (or passive reaction), for actions can then be
treated as a logical reflection of the circumstances.
Phenomena and processes outside of a particular theory?: A model by
its nature includes only those phenomena judged important by the
researcher. In working out the mathematical implications of a model,
126 Chapter 4

however, the researcher will occasionally be surprised by a result. Attempts

to achieve a desired result may encourage the search for additional variables.
Nomothetic versus idiographic relationships: Modelling sets the stage for
both. Valiela (2001, 25) suggests that models can usually not be accurate,
realistic, and general at the same time. That is, in seeking generality some
important details are necessarily neglected.
Spatiality (studies through space?): Models can have a spatial element,
but, by adding extra dimensions to the analysis, this often makes the
mathematics intractable. There is thus a bias toward ignoring spatial
considerations.
Change through time (continuous or static coverage?): Models can
embrace change through time, but it has generally proven easier to model
smooth change through time. In recent decades, models that suggest more
unstable, and even stochastic, time paths, have been developed: chaos
models are an important example.
Various time-paths: It is easier to model equilibrium systems, and thus
scholars must be wary of a bias toward these.
General comments: Models are often hailed for making an author’s
theoretical arguments clear. In that sense, models may be almost perfectly
replicable, for any researcher should achieve the same solution to a system of
equations. Hedstrom and Swedberg (1998) urge all social scientists to
develop formal models, since research always at least implicitly involves
models. However, it is important to distinguish an insistence upon an author
clearly stating their assumptions and arguments (see chapter 5) from an
insistence that this be done in a particular format.
Note: Game theory modelling is a common form of mathematical
modelling. When individuals or groups must interact strategically – as is
often the case – game theory allows scientists to model how they might
develop strategies, and how these will be adjusted as they interact with
others. Unfortunately, game theoretic models often have either none or
multiple solutions, and thus have limited ability to explain many observed
interactions.
Models sometimes lend themselves to simulations. A range of possible
values for the model’s parameters are entered, and the model is run by
computer to generate results. Simulations are particularly valuable for
models that do not generate precise mathematical solutions. Simulations
may identify under what conditions different sorts of time path are generated.
The value of simulations depends on the validity both of the model itself, and
of decisions about the possible range of values for particular parameters.
4. Classifying Method 127

5.6 Participant Observation (including ethnographic

field work)

This is by far the most common sort of observational analysis, though

discreet observation, in which researchers observe agents without the latter’s
awareness of their presence, is also used, particularly in the study of animals.
Participant Observation (PO) researchers generally let participants know that
what they are doing, and then often include interviews with their
observations. Covert observation raises ethical questions, but may provide
the only means of observing, say, criminal groups.
Type of agents (and number investigated): PO is focussed on intentional
agents, but could comment on the constraints/incentives imposed on these by
non-intentional agents. A drawback is that only a small number of
individuals or relationships, and generally only one group, can be studied in
one research project.
Type of causation: PO researchers emphasize attitudes, but could also
study actions, as well as constraints/incentives. With respect to attitudes,
there may be no better way to ascertain how delinquents, for example, make
sense of their lives. Researchers can thus come to appreciate how delinquents
can come to view their behavior as appropriate. (Or not; researchers may
always be outsiders, no matter how hard they try.) The very presence of an
observer may cause participants to behave differently, and pretend to
different attitudes (see Goldenberg, 1992); if observers are there for a period
of months, though, participants start to forget about their presence, and
would in any case find it difficult to pretend over a long period of time.21
The four criteria for identifying a causal relationship: PO is potentially
capable of identifying all four. Note, though, that PO researchers are often
less concerned with identifying cause-effect relationships than to “tell it like
it is” (McNeill, 1990, 64). Research reports often take a descriptive rather
than analytic form. Many PO researchers feel that the method is better for
exploration/induction than hypothesis testing, though others differ
(Goldenberg, 1992, 322).
Decision-making processes?: Another strength of PO is that it follows
subjects as they make decisions. PO researchers will often ask subjects to
explain why they acted as they did.
Phenomena and processes outside of a particular theory?: PO is
inherently inductive, though individual researchers may be guided to ignore

21
Goldenberg (1992) notes that humans are in general not reliable witnesses (though trained
observers may be much better). He suggests that there will be biases both in what is
observed and in how this is reported.
128 Chapter 4

evidence which conflicts with their desired conclusions.22 Indeed,

ethnography in particular is often criticized for eschewing theoretical
preparation. Note that a PO researcher can usually not take notes while
observing, and thus there is a danger of selective memory. PO researchers
may also come to identify with the point of view of their subjects, and ignore
conflicting information. An opposite danger is that the researcher may be so
overwhelmed by a wealth of data that causal connections cannot be
established; in practice this rarely seems to occur. Note that PO is a passive
and patient activity; the researcher should not encourage actions or attitudes
or events that they wish to observe.
Nomothetic versus idiographic relationships: PO is idiographic in focus,
but studies across a variety of situations might identify general relationships.
Palys (1997) notes that it is possible to take a quantitative approach to PO;
quantification of observations would facilitate cross-study comparisons.
Individual PO studies can usefully identify idiosyncrasies (that is,
realizations of other phenomena) that prevent a general rule from operating.
And PO may be the best, or indeed only, way to study certain sudden or
unique events.
Spatiality (studies through space?): The PO researcher can potentially
follow participants as they move through space. In practice, though, this
may be constrained: one cannot follow firefighters or police or criminals
everywhere. Research is conducted in a natural setting. PO researchers both
seek and are granted access to some settings more than others.
Change through time (continuous or static coverage?): PO necessarily
follows subjects over a period of time. The coverage is fairly continuous: the
observer may not spend every hour with subjects, but will strive to spend
various parts of different days. PO studies generally last only a few months,
and thus cannot fully comprehend processes that evolve over longer time
frames.
Various time-paths: The PO researcher can identify movement toward
either existing or new equilibria only if these occur within the timespan of
the research. More generally, the researcher will identify some causal links
that tend to maintain stability and others which imply change. It will be
difficult within the confines of one study to identify the kind of change, or
even to be confident that stability is maintained.

22
Margaret Mead famously argued that Samoan adolescents were immune to feelings of
sexual jealousy. This result is now viewed with great scepticism. Still, Goldenberg asserts
that PO is very good at “theory-building”.
4. Classifying Method 129

General concerns re reliability and validity: The active involvement of

the researcher raises unavoidable questions of reliability and validity.23 “A
record or text of a participant observer influenced by a functionalist approach
and a participant in a conversation in a hermeneutic tradition yield vastly
different data” (Herda, 1999, 4). PO studies are inherently non-replicable.
(Goldenberg, 1992, thus urges the use of multiple observers wherever
possible.) Note that both problems can be reduced by attempts to detect
similarities and differences across PO studies (and this will be aided by PO
researchers being explicit about the nature of their research). Care must be
taken, though, that studies, at least in aggregate, are representative; there may
be a tendency to study unusual (especially powerless) groups, in part because
it is easier to gain access.
Palys (1997, ch.9) notes that PO is almost always integrated with the use
of interviews, surveys, and/or textual analysis. This allows researchers to
compare what people say and do. This likely serves to decrease the problem
of researcher bias. Still, Silverman (2000, 10-11) argues that all forms of
qualitative research raise the danger of anecdotalism, where unrepresentative
snippets of conversation are used as evidence.
Note: One form of Participant Observation is protocol analysis, whereby
participants are asked to perform a task and describe verbally their thoughts
while doing so. Protocol analysis effectively combines elements of PO and
interviewing. It is more deductive than other types of PO. It is also more
artificial, and thus raises questions of whether participants will both do and
think as they would in a less artificial environment.
Non-participant observation (such as through a two-way mirror into a
classroom or workplace) is also possible. It is generally difficult to observe
people for any length of time without being observed. New technologies,
such as video surveillance or the monitoring of Internet conversations,
facilitate this sort of research. Doing so, however, raises ethical questions.
Where feasible, non-participant observation alleviates some of the concern
about subjects altering their behavior; the cost is that one is distanced from
subjects’ thought processes.

5.7 Physical Traces (as in archaeology or geology)

Type of agents (and number investigated): Non-intentional agents (some,

such as volcanoes) and intentional agents leave traces of themselves

23
Atkinson and Hammersley (1998) recognize that ethnographic reports necessarily absract
from reality; they nevertheless disdain the “linguistic turn” which stresses style over
substance in such reports, and celebrates their fictional quality. They suggest that
ethnographic studies should be judged in terms of validity and relevance to theory and
policy.
130 Chapter 4

(skeletons). Agents may also leave traces of what they touched: pottery, or
the plains scoured by glaciers. The traces left by relationships and groups
might be considered indirect traces of the individuals that comprised them.
Type of causation: Physical traces speak to passive causation (as when a
mountain pass facilitates movement) and intentional action (like making
pottery). Scientists can often infer attitudes (with a bit less confidence in
validity) from burial mounds and decorations.
The four criteria for identifying a causal relationship: Physical traces can
at times establish correlations – as when fossils are found in a particular
geological stratum – and (less often) temporality. They may point scientists
to intermediate variables, and may serve to discredit alternative hypotheses.
Decision-making processes?: Physical traces by their nature tell scientists
about the results of decisions, though they may thus provide indirect
evidence of decision-making processes.
Phenomena and processes outside of a particular theory?: This is quite
possible, and often occurs. Biologists have wondered about punctuated
equilibria because of gaps in the fossil record, and archaeologists have
posited links between groups on the basis of similarities in pottery. Still the
favored theory may guide scientists to look in places that are likely to contain
physical traces in accord with that theory.
Nomothetic versus idiographic relationships: Individual physical traces
are situation-specific, but generalizations can be drawn from many.
Spatiality (studies through space?): Physical traces can be collected from
different locales, and thus inferences drawn about spatiality.
Change through time (continuous or static coverage?): Physical traces
can be collected from different points in time, to identify what seems to have
changed, and perhaps evidence on why.
Various time-paths: Physical traces are best at highlighting changes in
particular phenomena over time, but can allow much insight into interactions
among these. Equilibria might be posited if the physical record exhibits little
change over some period of time.

5.8 Statistical Analysis

This generally involves analysis of statistics collected by others, but may

be applied to any statistics, including those generated by surveys or
experiments. Secondary data analysis is the most popular single method in
social science. Singleton and Strait (1999) worry that this reflects the simple
fact that it costs less – in time and money – than other methods. In some
disciplines, notably economics, it is treated as “the” scientific method. The
interdisciplinary scholar will often struggle to make specialists heed evidence
from other methods.
4. Classifying Method 131

Type of agents (and number investigated): Data are usually collected at

the individual level, but sometimes in terms of organizations. Data are often
available for huge numbers of individuals (often approaching all). Data can
often be aggregated to represent group tendencies, and also differences
within groups. Relationships are usually measured indirectly at best. Data are
often available for both non-intentional agents and, especially, intentional
agents.
Type of causation: Data are predominantly recorded in terms of actions,
though numerous surveys of attitudes exist. One can hope to identify
correlations between attitudes and diverse other phenomena, and interpret
these as causes or consequences of attitudes. But secondary data cannot
provide detailed insight into how any intentional agent forms attitudes.
Where data on non-intentional agents are available, secondary data may be
able to illustrate very well the constraints and incentives these impose on
intentional agents.
The four criteria for identifying a causal relationship: Statistical analysis
can establish correlations extremely well. If data are available for many
periods, it may be possible to establish temporality. But this may be difficult
if the lag between cause and effect is unknown or indeterminate, and if both
causes and effects occur frequently through time. Even when a correlation is
established; scientists must use judgment and rely on theory in inferring
causation (Silverman, 2000, 6-9). McKim (1997) concludes that statistical
analysis should only be taken as evidence of a causal relationship if
reinforced by plausible theory and direct evidence from other methods of the
causal process in operation (he also stresses the importance of
operationalization of phenomena, and exploration of all relevant causal
links).24 Correlation might imply causation in either or both directions, or
might reflect the operation of unknown intermediate or contextual variables
(McKim, 1997, 10). Scientists must also worry about the strength of a
relationship; researchers often celebrate the “statistical significance” of a
result without taking the necessary – and inherently qualitative – step of
asking if the relationship is important. They may thus too easily embrace a
theory with limited explanatory power, or reject a theory with great
explanatory power because their sample size was small and thus statistical
significance not established. McCloskey and Ziliak (1996) analysed a large

24
Economists have in recent decades embraced a critique of macroeconomic models which
argues that changes in government policies will change attitudes and thus the strength of
causal relationships: economists can thus not use these models to predict how people will
respond to changes in policy. It has not been appreciated that the same critique
undermines all types of statistical analysis. Unless scientists understand why observed
correlations exist, they can have little confidence that a correlation observed in one time
and place will hold in another.
132 Chapter 4

sample of papers from a leading economic journal, and found that the vast
majority had failed to evaluate the real world importance of correlations.
Alternative hypotheses can potentially be tested, though in many important
cases these cannot be distinguished statistically. Intermediate variables can
be identified if they are recorded, and if the researcher includes them in the
analysis. Ragin (2000) devotes much of his book to arguing that standard
statistical procedures do not properly engage questions of sufficiency or
necessity: they instead establish potentially misleading correlations that may
ignore important ways in which causal factors interact. He is particularly
critical of casual assumptions that causal phenomena are independent, and
notes that the statistical techniques for dealing with interdependent variables
yield ambiguous results unless the number of variables is small relative to
the size of the data set; there is thus a tendency to only investigate
interactions preordained by theory (2000, 311-5). Statistical techniques can
also not distinguish between a variable which makes an outcome possible
and one that actively causes the outcome (Meltzoff, 1998). While statistical
procedures exist to distinguish causal influences in opposing directions, in
practice these can easily be conflated.
In the mid-nineteenth century, during a cholera outbreak in London,
William Farr found a strong negative correlation between incidence of
cholera and elevation; this supported the reigning theory of the time, that
cholera was transmitted through the air. John Snow argued forcefully that
cholera was transmitted through water. Unlike statistical analysis, case
studies and natural experiments did provide support for this alternative
theory. Only then was it recognized that the cholera flux was heavier than
water, and the pumping technology of the time caused less of it to be
transmitted to neighborhoods at higher elevation. S. Turner (1997) takes this
cautionary tale as evidence of the danger of relying on statistical analysis
alone, especially when unknown causal processes may be at work.
Decision-making processes?: By their nature, secondary data speak to
results rather than decision-making processes. As with respect to
mathematical modelling, there may be a bias toward assuming rational
decision-making, for this is easier to model.
Phenomena and processes outside of a particular theory?: There are
statistical techniques that allow the identification of correlations across
disparate bits of data: there is thus some inductive potential (though Valiela,
2001, 16-7, doubts that the search for correlations unguided by theory is
generally worth the effort). And even when a theorist has an idea of which
variables to include, statistical analysis can give surprising insights into their
relative importance. In general, though, secondary data are used
deductively. And note that researchers are guided to look at only those
phenomena that are quantifiable and for which data exists. Data rarely exists
4. Classifying Method 133

for every phenomenon a researcher might wish to investigate (McNeill,

1990, ch.5). Results of statistical analysis will be misleading if important
causal variables are not included.25
Nomothetic versus idiographic relationships: Both are quite possible.
Regression analysis tends to look for one equation that best fits the whole
data set: it may miss out on identifying a host of idiographic relationships
hidden within the data set (see Ragin, 2000).
Spatiality (studies through space?): Data on spatial location (or, say, time
spent driving to work) may allow some analysis of the causes and
consequences of location and movement decisions. Note, though, that it can
be quite difficult, due to the multidimensionality of space, to code space into
statistical analysis, and thus data analysis alone cannot capture the full
richness of spatial movement. However, data coded by geographical region
can allow detailed analysis of regional differences. Due to the difficulties of
coding for location, spatial information is entirely absent from many data
sets where it would provide useful information.
Change through time (continuous or static coverage?): Data are by
definition collected at a point in time. In some cases, such as price data,
daily or even second-by-second movements (as on a stock exchange) are
collected. More often, data are intermittent in nature: analysis of temporal
change is still quite possible, as long as scientists are confident that the same
data collection techniques (including relevant definitions and sampling
protocols) were used across time. Even data collected at one point in time
can be used to draw temporal conclusions: if the young measure differently
than the old it can be hypothesized that today’s young will behave like
today’s old in the future. But such a hypothesis cannot deal directly with the
possibility that particular historical circumstances shaped the observed
behavior of the old.26
Various time-paths: Statistical techniques are well-developed for
evaluating whether a system of phenomena tends to move through time to

25
“Once we have a data set, it is a good idea to really try to understand the data before
plunging them into statistical tests now temptingly easy to do using statistical packages”
(Valiela, 2001, 44). Valiela guides researchers to examine outliers, see whether there are
obvious differences across samples, and see whether the data obeys common statistical
assumptions. Sloppy statistical analysis will lessen the potential for induction, and also
reduce the reliability and validity of the analysis.
26
Dale and Davies (1994) discuss various problems in analyzing change through time with
data from one point in time: “age” and “cohort” effects are conflated, respondent reports of
previous experiences are especially likely to error with regard to previous attitudes, and
duration effects may thus be underappreciated (the longer one is married the more likely
one will stay married). They thus encourage various forms of longitudinal research, where
data on the same or very similar subjects is collected at several points in time.
134 Chapter 4

new or old equilibria. There is a great danger that scientists will find
equilibria more often than these exist, simply because techniques for
identifying chaotic or disequilibrium systems are much less well developed.
Techniques of partial analysis, which measure how a change in one
phenomenon tends to cause (how much) change in others, are also available,
and often used. Analysis at the level of particular causal links can serve as
some antidote to a tendency to exaggerate system-level equilibriating
tendencies. Even here there has historically been a problem in that it has
been much easier to estimate relationships that are linear than to estimate
exponential relationships; the latest statistical packages are alleviating this
difficulty. Newell (2001) has hypothesized that interdisciplinarians will
often/always be interested in non-linear relationships, for causal links that
can be comprehended linearly will have been been embraced by disciplinary
researchers.
General considerations re reliability and validity: There is “a common
tendency to treat the validity of numerical data as given despite their
constructed character and the sources of potential error built into them”
(Hammersley and Gomm, 2000, 155). Scientists must look at how data are
recorded, and ask whether either those reporting or those recording had
likely biases. Government bureaucrats may define variables for convenience
in ways that differ from scientific definitions (and recall the discussion above
that definitions of crime or mental illness may reflect a desire for social
control). And they may exaggerate the reliability of statistics (such as crime
figures); researchers may be unable to check this.27 Moreover, neither those
collecting nor using statistics are as careful with respect to changing
definitions as they should be. Note that all of these problems tend to increase
as the researcher moves backward in time. Researchers may also be
selective in which data were included in the analysis. While statistical
analyses should be replicable, there are few professional rewards, at least in
human science, for attempting to replicate results. There is evidence that
mistakes in data entry or use of statistical techniques are common, though
there are only a few key studies here, again reflecting the limited rewards for
that type of research. In any case, the possibility of using different data sets
and different statistical techniques means in practice that quite different
results are derived from quite similar data. Perhaps most importantly,
scientists – and particularly interdisciplinary scientists – must worry that
phenomena that cannot or have not been measured quantitatively will simply
be ignored.

27
“Any data series is partly descriptive and partly interpretive; the thoughtful researcher pays
heed to both aspects”; those collecting data must necessarily have recourse to much
“qualitative” decision-making (Palys, 1997, 239).
4. Classifying Method 135

5.9 Surveys

Type of agents (and number investigated): Surveys are naturally focussed

at the individual level. Results are often broken down on a group basis.
These necessarily speak to average tendencies of group members rather than
group processes themselves. Surveys can point to important differences
among group members, which might not be exposed by other methods.
Note, though, that surveys may contain too few causal variables. Dogan and
Pahre (1990, 132) bemoan the fact that studies of voting behavior long
concentrated on variables such as race and gender commonly measured by
surveys; only as interviews came to probe important psychological variables
were these incorporated into theory. Surveys usually speak only indirectly to
relationships; network analysis, in which people are surveyed about whom
they interact with, can, however, identify relationships.
Surveys must be carefully worded. Still it is likely that different
respondents will interpret questions differently. Researchers in turn must
thus be careful in interpreting results. One stratagem is to add an
interviewing component that may alert researchers to the possible range of
interpretation.
Type of causation: Surveys can provide quantitative data on attitudes at a
point in time, and some limited idea of how those attitudes emerged (subject
to surveyee bias, and the ability of the surveyor to ask questions that
accurately identify possible causes). Only with (expensive) longitudinal
analysis, where the same subjects are repeatedly surveyed, can changes in
attitude over time be observed. Like other methods focussed on attitudes,
surveys can give some insight into actions predicated on those attitudes,
though again this depends critically on asking the right questions and
obtaining unbiased answers. Surveys can ask questions about actions of the
“how often do you X?” variety; these are especially subject to surveyee bias.
Note that while surveyor biases can be ameliorated through multiple studies
with different questions, this strategy may do little about surveyee bias.
There is some scope for asking indirect questions that can elicit information
that subjects do not wish to admit. On the other hand, surveyors can (often
unintentionally) introduce bias and error into results by improper wording of
questions (this problem can be alleviated by varying the wording, and pre-
testing the survey). Other methods, such as interviews and PO, may alert
researchers to likely surveyee biases.
The four criteria for identifying a causal relationship: Surveys can
identify attitudes that are correlated, and can associate attitudes with actions
to the extent that these can be measured. One problem here is that there is
often an inconsistency between what people say and do. Surveys are less
likely than interviews to establish temporal priority, both because the
136 Chapter 4

questions are less flexible and the potential for removing bias less. In some
cases, though, people may be more honest in an anonymous survey (unless
they are misleading themselves) than in a face-to-face interview. And if
individual surveyee biases are likely to extend in opposing directions, the
larger sample available with surveys will be particularly advantageous. Still,
researchers must be careful in inferring causation from correlation; evidence
from other methods can be invaluable.
Decision-making processes?: Surveys will in general be less successful
than interviews at identifying decision-making processes.
Phenomena and processes outside of a particular theory?: Surveys,
unless they allow for open-ended responses (in which case they blend into
interviews; they also become much harder to administer, and create both
practical difficulties and questions of bias in evaluation), are very poor at
identifying these. Those surveyed may find that they cannot respond in the
way that they would wish to. Note that in drawing causal conclusions from
survey data alone those items not included in the survey will inevitably be
excluded. Critics of survey analysis worry that researchers impose their own
analytical concepts on the questions asked.
Nomothetic versus idiographic relationships: Surveys, with their
potential to ask a range of questions of a very large number of individuals,
may be able to identify general relationships or very specific ones.
Spatiality (studies through space?): Surveys can usually only ask people
about movements through space, and are thus subject here to the various
biases noted previously. Surveys can investigate how people in differing
spatial situations respond to survey questions. Longitudinal studies can
provide some measures of migration.
Change through time (continuous or static coverage?): Surveys take
place at a point in time. Longitudinal studies catch the same people at
different points in time, and can thus measure more accurately than a single
survey how people’s attitudes and behaviors change over time: scientists no
longer have to worry about errors of memory. Large-scale surveys, like the
census, can provide similar insights, to the extent that researchers can assume
that the “sample” is the same at different points in time.
Various time-paths: Surveys on their own will have great difficulty in
identifying these.

5.10 Textual Analysis

As with other methods, there are a variety of techniques for the

performance of textual analysis. These are often grounded in different
theoretical understandings of the meaning of a text. In particular, theorists
disagree whether, and if so to what degree, the core message of any text can
4. Classifying Method 137

be identified. The analysis here pursues a Golden Mean approach whereby it

is argued that this can be done, albeit imperfectly.
Type of agents (and number investigated): Texts speak most directly to
the intentions of their author, but can also provide valuable insights into the
author’s perceptions of groups, relationships, and non-intentional agents.
Love letters and other correspondence can provide special insight into
relationships. Group-sponsored texts can speak to group attitudes. There are
limitations to the insights that can be drawn from any text, and also the
number of texts a research project can embrace. It is increasingly
appreciated, however, that no text can be understood fully in isolation from
other texts, since language is symbolic. That is, the meaning of a word
reflects its use in multiple texts.
Type of causation: Texts speak most directly to intentions, but provide
indirect evidence on passive and active causation. In the case of historical
events, texts (sometimes in conjunction with secondary data) may be the
only information available on any type of causation. Researchers must be
wary that the author has biased their treatment of events. Content analysis,
by which quantitative analysis is performed of how often particular ideas or
phrases appear in a text, is one technique which can potentially identify
intentions of which the author was not even aware.28 This does not obviate
concerns with authorial bias. While authors in general cannot react to the
particular researcher or research project, they may nevertheless wish to leave
a biased interpretation for posterity. Authors are nevertheless often
inconsistent, and provide multiple perspectives in a single text. The
availability of multiple texts (and multiple techniques for analysis) can
provide some solace here, though note that most texts have historically been
written by well-educated, higher income white males. Those texts that
survive are likely to have pleased the rich or powerful. Researchers must
worry about whether the author had first-hand knowledge of events, or has
had to rely upon perhaps biased reports from others.
The four criteria for identifying a causal relationship: Researchers may
attempt to discern all four elements within a text. While they may reach
conclusions quite different from the author, they can only build upon the
information that the author (consciously or not) provided.
Decision-making processes?: Authors often provide some discussion of
why they or others behaved as they did, though this may be ignorant or
purposely biased.
Phenomena and processes outside of a particular theory?: Texts are
usually quite rich and can thus provide a diversity of interpretations. As

28
Goldenberg (1992) worries, though, that quantitative techniques of textual analysis will
miss that which is implicit in the text.
138 Chapter 4

with PO, there is a danger of the researcher being so overwhelmed by detail

that they cannot discern causal processes. As with PO this danger is
generally overcome in practice.
Nomothetic versus idiographic relationships: While individual texts tend
to deal with the particular, scholars can hope to discern general rules by
drawing on many texts. Note, though, that there are many situations,
historical and otherwise, for which no texts exist.

Table 4-1. Typology of Strengths and Limitations of Methods

Criteria Classifica- Experiment Interview Intuition/ Mathemat-
tion Experience ical
Modelling
Type of All All; but Intentional Intentional All
Agent group only individuals; individuals;
in natural relationships others
experiment indirect indirect

Number All Few Few One All

Investigated

Type of Action Passive, Attitude; Attitude All

Causation (evolution- Action acts
ary) indirectly

Criteria for Aids each, Potentially Might pro- Some insight All; limited
identifying a but limited all four vide insight on with respect
causal on each correlation, to
relationship temporality intermediate,
alternatives

Decisionmak Indirect No Some Yes; may Some insight

-ing Process insight insight; mislead
biased

Induction? Little Some If open Yes; bias Little

Generaliz- Both Both Idiographic Idiographic Both

ability

Spatiality Some Constrained From From Difficult to

memory memory model

Temporality Some Constrained From From Simplifies

memory memory

Time Path No insight Little insight Little insight Little insight Emphasize
equilibrium
4. Classifying Method 139

Criteria Participant Physical Statistical Survey Textual

Observation Traces Analysis Analysis
Type of Intentional All; All; Intentional Intentional
Agent individual; groups and groups and individuals; individuals;
relationships relationships relationships groups others
groups? indirect indirect indirect indirect

Number Few; One Few Many/all Many One/few

group

Type of Action Passive, Action, Attitude; Attitude,

Causation (attitude) Action Attitude acts Action
indirectly

Criteria for All, but Some insight Correlation Some insight Some insight
identifying a rarely done to all four and on on all
causal temporality correlation
relationship well; others
maybe

Decisionmak All No No Little Some

-ing Process insight;
Biased

Induction? Much Much Some Very little Much

Generaliz- Idiographic; Idiographic; Both Both Idiographic;

ability nomothetic nomothetic nomothetic.
from many from many from many
studies studies studies

Spatiality Very good; Possibly Limited Rarely Possible

Some limits infer

Temporality Very good Possibly Static, often Longitudinal Possible

up to infer frequent somewhat
months

Time Path Some insight Some insight Emphasize Little insight Some insight
equilibrium

Spatiality (studies through space?): Authors may discuss movements of

themselves and others through space, subject to the usual biases.
Change through time (continuous or static coverage?): Texts often
provide continuous coverage over lengthy periods, subject to the usual
biases.
Various time-paths: Evidence of all sorts of time-paths can be gained
from texts, subject to the usual biases.
140 Chapter 4

Reliability and Validity: As with many other methods, there are obvious
advantages to replication. Yet there is virtually no explicit replication, either
in historical studies or in the evaluation of fictional texts. When the same
text is revisited, it is generally to show how a different interpretation can be
brought out by a different theoretical perspective. As noted before, one
interpretation is that it is possible to find anything in a text. In general,
though, one can identify the main messages.29
The literary theorist Roland Barthes argued that interdisciplinarity must
involve a focus on a subject such as “the text” that belongs to no discipline
(Moran, 2002, 83). The analysis of this book would suggest that theories of
textual interpretation and methods of textual analysis should be
distinguished, and that neither should be the monopoly of any one
community of scholars. Interdisciplinarians should be open to the use of
multiple theories and methods. While an expertise in literary theory or
method may be invaluable, interdisciplinary scholarship should not be
limited to these.

5.11 The Case Study

This is really a grouping of methods, rather than a method in its own

right. As such it is likely the most common method in human science
(Ragin, 2000, 67). Bouma and Atkinson (1995, 114) call case studies “the
basic building block of research design” in human science, and praise their
inductive and deductive potential. Case studies are also much used in natural
science, especially biology and geology. The evidence for the case study
might be drawn from one or more of physical traces, textual analysis,
secondary data analysis, interviews, PO, or perhaps other methods. Since
case study research benefits from a familiarity with multiple methods,
interdisciplinary researchers may have an important role to play.
Case studies can usefully be contrasted with statistical analysis: the
former look at one particularity in rich detail, while the latter tends to seek
patterns across numerous cases (this point is emphasized by Ragin, 2000, 25;
he also notes that there is little research between the two extremes of one
case and many cases). Case studies can, however, be quantitative and/or
qualitative. In case studies, “There is no claim to representativeness, and the
essence of the technique is that each subject studied, whether it be an
individual, a group, an event, or an institution, is treated as a unit of its own”

29
“Most philosophers and critics would be quite willing to admit that slippage of meaning
occurs (poetry works on just such a principle), but would draw the line at saying that
nothing but slippage occurs: it is hard to see how, if that were the case, we could even
communicate such a state of affairs.” (Sin 1992, 109).
4. Classifying Method 141

(McNeill, 1990, 88) Case studies are very useful for idiographic purposes.30
Ragin (2000, 68) suggests that case studies usually turn up evidence that
does not fit any theory perfectly.
Note, though, that a causal sequence cannot be distinguished from an
accidental sequence by relying on any one case study (Little, 1998, 213).
However, general rules may be built up or supported by comparing a number
of case studies (a point made with respect to each of the case study methods
above).31 Some of the detail of individual cases is inevitably lost in
generalizations (Stake, 1998). Scientists could usefully strive both to derive
generalizations and to explain deviations from these. One disadvantage of
statistical analysis is that outliers – data points that fit the estimated
relationship poorly – are rarely subjected to special analysis; a case study
approach could complement statistical analysis here.
McNeill (1990) emphasizes the value of case studies for induction.
Researchers should be careful not just to report those observations that seem
to lend themselves to generalizations; other information may encourage the
development of alternative theories or the recognition of limits to existing
theories. Note that cases are often chosen for their intrinsic interest (but also
to provide insight into a particular issue or theory; Stake, 1998), and thus
may not be representative. This potential bias could be combated through
random sampling, or through choosing quite different cases. While there
seems a danger that a researcher could draw any lesson from the multiplicity
of data contained in any one case study, in practice this danger is limited, and
researchers are often surprised by their results (Ragin, 2000, 68).
Note: Community studies represent one type of case study. A particular
community is analysed with the use of a variety of methods, including PO,
surveys, interviews, and textual analysis (McNeill, 1990, 93).

5.12 Evaluation Research

In evaluation research, the focus of study is a particular program

intervention. In other words, scholars aim to evaluate the effects of a
(intentionally created) non-intentional agent in terms of its effects on
individuals or groups. As noted at the start of the chapter, evaluation
research can utilize many distinct methods. “Evaluation research is not,

30
One major criticism of area studies programs has been their case study approach. Geertz
in 1973 responded in favor of the “thick description” inherent in case studies, “without
which the complex interrelationships among social processes and events cannot be
meaningfully unravelled.” As proponents of case studies have suggested criteria by which
these can be evaluated, opposition to case studies has fallen away (all in Tessler, 1999).
31
McNeill (1990) speaks of a ubiquitous comparative method. With case studies,
comparison is made across studies. With experiments, there is internal comparison with
the control group.
142 Chapter 4

then, a specific body of research techniques but an approach to research

which differs from basic or academic research in its purpose” (Hall and Hall,
1996, 46). Singleton and Strait (1999, 449), for example, note the advantages
of secondary data analysis if a program affects a whole population, and
surveys of both the affected group and a control group, or experimental
design, if the program is new; others would emphasize PO or interviews.
There is an increased danger of researcher bias here, for researchers
often want to believe that programs work (though in addition to evaluation,
researchers are often also interested in information on how to improve a
program). The subjects of research may also be biased in this way: those
who (think that they) gain from a program will exaggerate its benefits and
downplay its costs.
Evaluation research might be seen as a subset of action research: while
there are different definitions of this, they seem to share a focus on concrete
problems.32 In developing specific knowledge for a specific problem in a
specific setting (a very popular approach in education research), a variety of
methods are used (Bell, 1999, 8-9). Reason (1998) advocates an activist
approach, in which the researcher does one or more of three things: actively
develops strategies in concert with a group; develops knowledge useful for
group action; or identifies theories in the literature useful to group action.
Obviously there is great scope for biases, both in that which the researcher
communicates to the group and that which the researcher reports about group
activity. Such action researchers, not surprisingly, are very sceptical of the
possibility of objective science (Hall and Hall, 1996). But we shall see in the
next chapter that participation in policy analysis can provide valuable insight
into human science theory.

5.13 Hermeneutics

As the study of symbolism and meaning, hermeneutics can be based upon

information gained through a variety of the above methods: textual analysis,
participant observation, and interviews being prominent. Yet unlike case
studies or evaluation research it is more than just an amalgamation of
simpler methods. Hermeneutics emphasizes the transmission of multiple
meanings through symbols. Symbols must be analysed both on their own
and within larger systems of understanding. In textual analysis, then, a
circular process is advocated of analysing particular words and sentences,
analysing the text as a whole in terms of these more specific analyses, and
then using this holistic understanding to investigate the particulars again.

32
Note, though, that some interpretivist researchers use the phrase “action research” in order
to distinguish their research from the causal emphasis of positivists.
4. Classifying Method 143

Since by definition symbols exert unobservable effects, hermeneutics faces

inherent difficulties in establishing reliability and validity.

5.14 Mapmaking

This is a combinatorial method, for it involves mapping information

gathered from other sources: generally secondary data or surveys, though
texts are another possibility. It must thus reflect the strengths and weaknesses
of its sources of data. As with hermeneutics, mapmaking involves a set of
techniques beyond those involved in the methods it draws upon. Its obvious
strength lies in terms of spatiality. There are, quite simply, some spatial
movements that cannot be properly visualized nor comprehended without
recourse to maps (see Pounds, 2001). But while mapping is generally
considered in terms of spatiality, mapping is possible along a number of
dimensions. The system of phenomena and causal links outlined in chapter 2
has been spoken of, for example, as a map of (some aspects of) human
science. And Goldenberg (1992) discusses network analysis (generally
grounded in survey analysis; see above), which maps how individuals
interact (using various kinds of map).33 In general, maps are useful when
scholars are talking about relationships among numerous places or things. In
particular, maps can be useful in representing movements through time as
well as space, and can thus be crucial to comprehending change through
time.34 Note that in placing information on a map, hidden relationships may
become evident; mapmaking can thus be very important inductively as well
as deductively.
Scholars must recognize, though, that no map is a flawless rendition of
reality; all maps emphasize some aspects over others. “Chemists create and
refine models [maps of molecular structure, for example] in an effort to
deepen their understanding of phenomena and to afford predictions that
guide future experiments. But models [maps] are not physical reality; they
constrain thought and even mislead at times, and this is especially likely
where our philosophical understanding of their status is incorrect or
incomplete” (Bhushan and Rosenfeld, 2000, 10). Maps may represent the
concerns and interests of the powerful, as when black population centres

33
“It is difficult to describe network analysis methodologically, since it seems to be more of
a perspective or even a theory than a method” (Goldenberg, 262). I suspect that the answer
to this quandary (which causes most writers on method to ignore network analysis) is that
mapping is inherently a method used in conjunction with other methods.
34
Pounds (2001) criticizes historians for not more often thinking cartographically. They use
verbal cartography when maps would be both more precise and objective (without maps,
researchers choose which spatial movements to report). He recognizes the costs involved
in drawing maps, but suspects that the problem is rooted in a lack of cartographic training.
144 Chapter 4

were ignored on maps of South Africa. Or they may exaggerate aspects of a

situation that are easier to map: maps of economic activity stress production
over consumption and industry over services because it is easier to map
activities that are geographically concentrated. The boundaries of maps must
also be questioned: Black (1997) notes that even weather maps tend to be
done on a national basis, though the weather is indifferent to such boundaries
(Fortunately an exhaustive map, such as that in chapter 2, can overcome this
problem.) Maps may mask certain differences: relative city sizes are
obscured on most road maps because the vertical dimension is ignored (As
noted in Szostak, 2003, some of the causal links discussed were much more
important than others; these differences could potentially be represented
diagramatically). Maps may make points look closer or farther apart than
they actually are. Even choices of color can affect one’s perception of a
map. In sum, while maps can be very useful, scholars must realize that
“subjectivity is central to the production and understanding of maps” (Black,
1997, 168).

6. ADVANTAGES OF A TYPOLOGY OF METHODS

It has hopefully been shown that the typology provides a means for
evaluating the strengths and weaknesses, and especially the range of
applicability, of each method. And it yields a most interesting result: each
method has strengths and limitations. The legions of scientists and
philosophers who have argued otherwise can be thanked for making this
Golden Mean conclusion so important.
It was noted in the previous chapter that the 5W typology of theory
provided an answer to the vexed question of “what is a theory?” Gower
(1997, 249) bemoans the fact that the failure of philosophers to provide an
“account of scientific method” to which no exception can be found, has, as
with respect to theory, opened the door to those who would deny the very
existence of “scientific method,” or any standard by which this might be
preferred to other forms of investigation.35 As with theory, the provision of a
list of scientific methods provides a more compelling (Wittgensteinian)
definition of scientific method than has existed heretofore. Moreover, the
5W questions show that all of these methods can be evaluated in a similar
fashion, and the important result derived that each has strengths and
weaknesses. Given that each proves to have different attributes, it is hardly

35
Sceptics point to the fact that the dominant method in science has changed over time. But
this merely shows that there is more than one scientific method, and none is perfect.
Advances in technology (such as improved telescopes) have influenced the popularity of
particular methods at particular points in time.
4. Classifying Method 145

surprising that philosophers have found it difficult to define scientific

method precisely in one or two sentences.
There are, of course, other possible schemes by which one might compare
and evaluate methods. But this typology has a claim to objectivity: note in
particular that other researchers could readily replicate the approach of
asking the 5W questions. Such efforts might result in the addition of new
dimensions; they are unlikely to result in the removal of any of the
distinctions drawn above. The approach taken above, if not itself exhaustive,
at least points to an exhaustive system of evaluation. Singleton and Strait
(1999, 410) evaluate their four favorite methods in terms of seven criteria.
Inevitably, they thus capture some but not all of the distinctions made here.
They worry, for example, about the possibility of replication, the effects of
researcher character and behavior, and whether subjects react to being
studied. Some of their criteria are quite general. For example, “limits to
what can be studied” is suggestive, but nowhere near as concrete as the
discussion above of how methods fare in terms of five distinct questions.36 In
other words, the 5W questions identify precisely where limitations lie.
Bouma and Atkinson (1995) evaluate methods in terms of five very broad
questions: “What is happening?”, “Has there been change over time?”, “Are
A and B different?”, “Are A and B different over time?”, and “Does A cause
B?”; they conclude that one broad type of method is best suited to each type
of research question identified. They thus severely under-appreciate the rich
variety of strengths and weaknesses in method.
Existing texts in research methods tend to pursue a “how to” orientation
(Ellis, 1994, Hartman and Hedblom, 1988). This necessarily leaves readers
with a limited understanding of the advantages and disadvantages of
particular methods. Other texts are aimed at making the reader a better
consumer of research (and some try to do both; Singleton and Strait, 1999,
xvi). They thus discuss strengths and limitations. But this is rarely done in
terms of a common set of questions, and thus readers have difficulty
comparing the strengths and weaknesses of methods for particular research
questions. Even when there are common questions these (as was seen above)
are of a general nature such that comparisons are still difficult.
As noted in chapter 3, while the essence of theories might be summarized
without recourse to the 5W questions, the 5W approach had two clear
advantages. First, by providing an organizing structure, it would greatly aid

36
They also suggest that all social research involves five questions: what do we want to
know and why?, what is to be observed? (time and place, units of analysis, relevant
characteristics, process in which units embedded), which and how many objects are to be
examined?, how are they to be observed, and what degree of certainty can be applied to the
results? The 5W typology has dealt with all of these questions and more.
146 Chapter 4

mastery of the material by both students and scholars. Second, and crucially,
it would facilitate comparison by providing a hopefully exhaustive set of
questions. The same can be said with respect to methods. The number of
dimensions has doubled, but is still quite manageable. As with respect to
theoretical flexibility, the typology provides a sound base from which to urge
methodological flexibility, both in general and with respect to particular
research questions.
Specialized researchers can use the typology to ask whether methods
eschewed by their community might have something important to say.
Interdisciplinary scholars with expertise in a particular method could use the
typology to gauge where that method might usefully be applied
“[S]ystematic explorations of new domains have played a large role in the
history of many sciences,” as when scientists explored how magnets operate
on different metals. “Existing knowledge – a taxonomy of substances and
their properties and components – is the major prerequisite for the generation
of such programs” (Langley et al., 1987, 306). If this strategy works for
magnets, why not for (theories or) methods?
One further advantage of the typology is that it suggests that the scientific
community does not take full advantage of the potential of particular
methods. It was often noted that a particular method is not used within its
full range of applicability. The inductive potential of secondary data analysis
is not fully appreciated. PO researchers could try harder to identify causal
relationships.
The advantages of replicating various types of research were also
highlighted. In the case of PO and interviews, comparing results across
different studies can greatly enhance both reliability and validity. In the case
of statistical analysis, replicability is often assumed, but scholars should
make greater efforts to illustrate this. Validity is also problematic; multiple
studies using different data and techniques are highly advisable. These
points are perhaps well understood, but are rarely reflected in especially
human science research.
More generally, the analysis above points to common flaws in the
application of methods. Better scientific practice can and should be
encouraged with respect to all methods. Palys (1997, 186) notes that while
researchers often criticize the practices of other researchers, they are
generally critical not of particular methods per se but of how these are
applied. They also tend to gloss over problems with the application of their
favored method. A greater appreciation of the problems, and best strategies
for overcoming these, within each method, should thus act to encourage a
greater appreciation of methodological diversity.
The 5W approach provides a means by which both scholars and students
could be introduced to the possibilities inherent in all methods. Mastering
4. Classifying Method 147

any of these methods is a time-consuming process, and it could hardly be

expected that most scholars or students will be adept at more than a couple of
these.37 Since people will be exposed in their lives to examples of each,
however, it would be very useful to be acquainted with their pros and cons.
In particular, individuals should appreciate the sorts of biases that can creep
into results of inquiries utilizing each method, and the guidelines that might
be used in order to evaluate the quality of a particular piece of research. If
students were exposed to the sort of analysis above, they would inevitably
also be exposed to the important principle that scholars and citizens should
always seek evidence from more than one type of method. This sound
scientific principle may already be communicated within many university
programs, but can easily be lost if students are then exposed predominantly
or exclusively to evidence derived from only one method.

6.1 All Methods Are Good

The single greatest advantage of the 5W approach is that it shows that

there is a place for all of these methods in the scientific enterprise as a whole.
Each has unique strengths, but also unique limitations. The very fact that
methods can be differentiated along several distinct dimensions cries out the
fact that no one method should dominate science. What is true at the level of
the scientific enterprise as a whole is also true at the level of individual
scientific questions: these too tend to lend themselves to the use of multiple
methods. It is not, however, necessarily true at the level of individual
researchers: there are obvious benefits to specialized expertise. Even here,
though, there are benefits to being able to appreciate and evaluate research
using other methods. At least some members of any scholarly community
should have an expertise in multiple methods, so that results from different
methods can be integrated.
The implicit assumption of disciplinarity is that certain issues should be
analysed with the use of only one or two methods (though there may well be
scientists on the fringes using other methods). Various authors in Creath and
Maienschein (2000), including Schaffner (2000), suggest that if biology

37
Given the difficulty of mastering different methods, there are advantages to academic
specialization by method. The problem with disciplines is that this methodological
mastery is thus artificially conflated with a set of theories and phenomena, when all
methods should be applied to all relevant theories and phenomena. “[Research methods]
are not the private domain of any particular discipline, but rather are an abstract set of rules
or contingencies that may be applied to both physical science and social science problems”
(Hartman and Hedblom, 1988, 5). It is thus of some interest that some economists,
geographers, and cultural anthropologists would wish to define their disciplines in terms of
method alone.
148 Chapter 4

rather than physics had been the focus of philosophical attention, the value of
methodological plurality would be more widely appreciated; yet even in
biology many urge the exclusive reliance on experiments (Magnum, 2000).
A superior assumption can be proposed: given that no method is perfect,
scientists should always seek justification for any theoretical proposition in
terms of different methods. There may well be exceptions to this guideline,
but these should be established, not assumed. In the words of the Pragmatic
philosopher Charles Sanders Peirce, truth [or at least what is thought to be
likely true] is the confluence of independent streams of good evidence. One
of the costs of disciplinarity is that this common sense notion is not more
widely appreciated.38 Given the imperfections in every scientific method this
yields two types of unfortunate result. In the first case, a hypothesis receives
strong support from analysis using the favored method, but would be
questioned by analysis using other methods. Fuller (1993, 194-5), referring
to the study of science itself, notes that interviews, ethnography, and textual
analysis each have different biases that support different interpretations of
how science operates. In the second case, a hypothesis receives ambiguous
support from the favored method, but would be resoundingly supported by
research using other methods.39 The discussion of cholera and statistical
analysis above is an example.
The 5W questions allow this argument to be fleshed out in detail. Are
there particular types of agent, causal process, decision-making process, and
so on, for which one method far outperforms the others? The clearest
positive answer to this question lies in the applicability of the experimental
method to the study of the interactions among non-intentional agents. Here,
the main limitations of the experimental method, such as the danger that
agents act differently in artificial settings, difficulties in penetrating
decision-making processes, and spatial or temporal boundaries, often do not
come into play (but not always; some natural processes occur on a large
physical or temporal scale such that experimentation is infeasible). The
many strengths of the experimental method thus operate unimpeded. It

38
There are some exceptions. Fraser and Gaskell (1990, 13) celebrate the range of methods
used by social psychologists – interviews, observations, group discussions, content
analysis, and statistical analysis (though Himmelweit in the same volume worries that
some methods are still used less than they should be). And Pervin (1990) concludes his
handbook on personality theory thus: “At least for the present, however (and, we may
hope, for the future) the time of dismissal of some phenomena as not legitimate for
investigation and of hegemony of some research methods over others has passed.”
39
Oakley (2000) notes that as early as 1957 Trow had urged scientists and philosophers to
move away from asserting the general superiority of one method, discuss their relative
usefulness for different questions, and use the widest array of methods suited to the
problem at hand. Campbell and Fiske argued in 1959 that psychological traits in particular
had to be identified with multiple methods (Tashakkori and Teddlie, 1998, 41).
4. Classifying Method 149

should hardly be a surprise, then, that the experimental method dominates the
study of chemicals and atomic particles, for these are well suited to the
laboratory (and vulcanologists try to duplicate this method with natural
experiments). Note, though, that experiments are still fallible: researchers
may be oblivious to the role played by various elements of experimental
design in generating a particular result.40 Even here, then, scholars should
not entirely dismiss other methods. While those methods which involve
asking questions of subjects are obviously unsuited to non-intentional agents
(though chapter 5 will discuss the benefits of querying those scientists who
study these), classification, intuition (chemical reactions are regularly
observed in our daily lives), mathematical modelling (to identify possibilities
and exclude impossibilities), physical traces (of interactions in a natural
setting), and even mapmaking (the Periodic Table can be seen as a map) can
usefully be applied.
Other cases in which a particular method outshines others can be
identified. Mapmaking can give unique insight into spatial and temporal
aspects of causal systems (among other things). Intuition likely gives unique
insight into how and why, for example, art moves humans. Surveys are
advantageous for gauging attitudes of large numbers of individuals. PO may
be the only way to comprehend how others make meaning of their lives. But
again all methods are problematic and scientists would wish to supplement
these wherever possible.
Human science is often viewed as a weak relation to natural science,
precisely because it is not dominated to the same extent by one method. The
5W approach shows that the difference is one of degree rather than kind
(albeit an important difference of degree). Not all natural scientists can avail
themselves of the experimental method. Even those that can use experiments
should nevertheless seek insights from other methods. Human scientists
generally, but not always, focus upon intentional causation. Since all
methods are applicable to intentional agents, but not all are applicable to
non-intentional agents, human scientists must then have recourse to a wider
range of methods than those utilized in natural science. They should neither
idealize nor condemn the methods of natural science, but draw from the full
range. Interpretivists who claim that only attitudes matter neglect the twin
facts that actions result from a mixture of external (to the individual) and
internal causes, and that internal causes can be evaluated in terms of (non-
intentional) personality characteristics, schemas, and so on; they are thus
mistaken in suggesting that human science can have no recourse to the

40
This realization is of critical importance in modern philosophy of science. It was the
mistaken idea that (at least) experiments could be infallible that had supported positivist
arguments earlier in this century that scientific statements could be proved or at least
disproved.
150 Chapter 4

methods of natural science, but correct in asserting that human science must
do more than ape natural science (Singleton and Strait, 1999).
The 5W analysis can guide the research process in many ways. First a
theorist should be guided (by the analysis in chapter 3) to specify answers to
the 5W questions. They can then ask what sorts of method are best suited to
each of the elements of a particular theory. While it may be impossible for
an individual researcher to use all implicated methods, they might strive to
use those underemployed by previous researchers. Note that for any non-
trivial scientific question in human science it will likely be possible to
identify some application of all methods. Most questions in natural science
will likewise lend themselves to a handful of methods. Scholars writing
survey articles should thus strive to report research using all methods, or at
least urge scientists to broaden their approach so that this can be done.
“Unfortunately, social researchers rely altogether too frequently on a
single method or measure when a number of approaches could be brought to
bear on the research question” (Singleton and Strait, 1999, 394).41 Where a
particular research enterprise has been dominated by one method, the
researcher can use the typology to identify the most serious limitations of
that method, and their relevance to the particular research questions being
addressed. If, for example, statistical analysis of secondary data has
dominated, one would be concerned about the possible exclusion of
relationships from the analysis, the role of attitudes and decision-making
processes, the possibility of exclusion of non-equilibrium processes,
trivialization of the spatial dimension, omission of variables not implicated
by theory (or for which data could not be found), interactions among causal
factors more complex than could be handled by standard statistical
techniques (at least before the latest generation of statistical packages),
and/or failure to properly identify causal relationships. One could then turn
to methods that are better in each of these respects. Not surprisingly, various
case study methods loom large here. But again all methods will likely have
something to contribute.
Since particular methods are best suited to the investigation of certain
theory types, the complete reliance of any scientific community on one
method – except in those cases where one method is ideally suited to the
phenomena in question – will almost certainly bias results in favor of the
preferred theory (for surely the community will have chosen a theory and
method that deal with the same type of agent, action, and so on). In other

41
Pressures to conform are felt early in one’s career. “Every student whether writing an
essay or completing a doctoral thesis has felt this pressure. The methods used may satisfy
examiners but they are not suitable for capturing the topic. The brave may abandon the
study and their careers. Most adjust to conform to the standards required for qualification”
(Shipman, 1988, 166).
4. Classifying Method 151

words, scholars are supposed to test theories against alternatives, and if this
is done only in the context of a method best suited to one of the competing
theories, the community will be more confident than it should be in the
validity of that theory. Economists, for example, should be aware that
statistical testing is best suited to rational individual actions and equilibrium
outcomes, just as is the rational choice theory that dominates the discipline.
It is not enough that a community uses multiple methods: they must apply
these correctly, There is a danger that methods will be mis-applied such that
they generate desired responses (Fuller, 1998, 130). Familiarity with the
typology above would be invaluable in spotting such errors; this might be
coupled with a sense of the disciplinary perspective or worldview of the
scholarly community in question (see chapter 5, Szostak, 2003a).

6.2 Triangulation

Advocates of the use of multiple methods speak of triangulation (e.g. Hall

and Hall, 1996, 44, Oakley, 2000, 54): this refers to the method used by
surveyors to identify, say, the source of a radio signal, by establishing its
direction from two or more vantage points. Different methods provide
different viewpoints, and can thus hopefully guide researchers to the correct
answer. The analysis in this chapter broadly supports the idea of
triangulation. While the analogy is powerful, like all analogies it must be
treated with care. The 5W approach suggests two caveats or clarifications.
First, researchers should appreciate that a research question can usefully be
broken into parts, and different methods may cast light on different parts of
the broader question. This point is made by Tashakkori and Teddlie (1998,
42-3); they suggest that different methods can examine different facets of a
question, one method can provide inputs useful to the application of other
methods, and the use of multiple methods can add breadth to a research
endeavor. Second, it is entirely possible that different methods will point in
different directions. Singleton and Strait (1999) hope that if different biases
are at work in different methods (which is likely the case), the correct
answer will be approached. The fact that different methods have different
strengths and weaknesses is supportive of this conjecture: “Social science
methods should not be treated as mutually exclusive alternatives among
which we must choose...Our individual methods may be flawed, but
fortunately the flaws are not identical...A diversity of imperfections allows
us to combine methods...to compensate for these particular faults and
imperfections” (Brewer and Hunter, 1989, 16-7). Nevertheless, scholars
should be careful of assuming that the biases inherent in every method will
152 Chapter 4

cancel out in a particular investigation.42 If different methods resolutely point

in opposing directions, triangulation will be impossible, and the scholarly
community is left with the unpalatable task of choosing amongst competing
forms of evidence. The typology provides some basis from which to
evaluate the relative reliability and validity of these. Oakley (2000)
discusses how experiments with control groups often find medical therapies
to be less successful than do interviews or surveys; awareness of possible
biases in the latter methods in such cases would usually guide scientific
communities to rely more heavily on experimental evidence.
While triangulation across methods is most commonly spoken of,
triangulation can be and is performed in terms of phenomena/data, theory,
and different researchers as well. While it is spoken of here, discussions in
other chapters of phenomena, data, theory, and practice also highlight
potential biases, and thus suggest that triangulation of all types will be a
valuable – albeit imperfect – scientific procedure. Of course, triangulation
with respect to theory is desirable only if relevant methods suggest that more
than one theory has validity: one would thus not triangulate between
heliocentric and Ptolemaic theories.
Triangulation can occur within one study. Tashakkori and Teddlie (1998,
18) develop a typology of mixed method studies: different methods can be
used sequentially or simultaneously; they can be given equal status or one
may dominate; and different methods may be applied at different levels of
aggregation.43 Singleton and Strait (1999, 412) give examples where surveys
and PO, surveys and interviews, and experiments and interviews, were
usefully combined. A host of other examples could be provided. Indeed
since scientists rarely ignore their intuition (but rarely admit this),
triangulation within a research project may be endemic. Even where only
one method is (explicitly) pursued, reference is often made to previous
studies that utilized other methods. Even staunch advocates of one method
do so: Ragin (2000) celebrates the fact that case studies often cite statistical
analyses and vice versa.

42
This can be seen as a special case of the problem of induction: just because all existing
cases point in a particular direction does not mean this will hold in future (Seale, 1999, 57-
61).
43
Tashakkori and Teddlie also urge researchers to explicitly draw from both of what they see
as broader quantitative and qualitative paradigms. The analysis above suggests that it is a
mistake to reify qualitative and quantitative paradigms. Seale (1999) argues that research
should not be guided by a paradigm; he urges researchers to eclectically do what works
best. Hammersley (1996, 168) suggests that viewing qualitative and quantitative
approaches as distinct paradigms implies that researchers first choose a philosophic
orientation; in practice researchers develop their philosophical and methodological
orientations pragmatically through interaction with others.
4. Classifying Method 153

Just as some interpretivists object to judging research in terms of validity

or reliability, they object also to triangulation, because it assumes that there
is one reality to triangulate upon. Seale (1999, 57-61) suggests, though, that
triangulation would still be a useful process. First, it can identify the degree
to which convergence toward a common view occurs. Second, if there is
little or no convergence, scientists should be guided toward the creation of
new theories. Third, even if scientists cannot move beyond multiple
perspectives, these should still be identified.44
While specialized researchers may and do triangulate over some set of
methods or theories or phenomena, there is an obvious role for
interdisciplinarians to identify the full range of relevant methods, theories,
and phenomena, and attempt triangulation across as diverse a selection as
possible. They will, in doing so, want to be conscious of the various biases,
including disciplinary perspectives, which may have affected results of
previous studies. Triangulation should not be a license for scholars to
assume that every method should be applied to every question – indeed with
limited resources even marginally useful methods might reasonably be
eschewed – but should be an antidote to the all-too-common assumption that
only one method or theory or set of phenomena is sufficient. It can only
serve as such an antidote if applied carefully.

6.3 Meta-analysis

Meta-analysis involves drawing together the results of previous studies.

This, in a broad sense, can perhaps make the best claim to being “the”
scientific method, for science does not advance except through judgments of
multiple pieces of research. Meta-analysis can identify results that seem
robust across studies, and exceptions to such generalities. Properly done,
meta-analysis will identify weaknesses in different pieces of research, and
look for countervailing strengths elsewhere. In the words of Hunt (1997) it
is “how science takes stock.” The usual approach is to statistically compare
the results generated by different studies, using some quantitative coding of
the approaches taken in those studies. Meta-analyses thus indicate which
elements of particular studies generated particular results.
Meta-analyses have become very common in recent decades with respect
to clinical trials of new medical procedures. Given the expense of such
trials, and the common occurrence of conflicting results, meta-analyses have

44
Stake (1998) suggests a different interpretation of triangulation. It “has been generally
considered a process of using multiple perceptions to clarify meaning, verifying the
repeatability of an observation or interpretation. But acknowledging that no observations
or interpretations are perfectly repeatable, triangulation serves also to clarify meaning by
identifying different ways the phenomenon is being seen.”
154 Chapter 4

proven invaluable in drawing strong conclusions. Meta-analysis has been

employed across natural and human science in such areas as the correlation
between television violence and aggressive behavior, the risks associated
with second hand smoke, and the effectiveness of spending more money on
schools (Stanley, 2001). One important characteristic of meta-analyses is
that they can find statistically significant correlations among phenomena that
do not show up in the smaller sample sizes of individual studies.
The fundamental guiding principle of meta-analyses is that all relevant
studies should be included. Rather than excluding certain studies on the
belief that they were flawed – a process open to a variety of biases – the
meta-analyst should code the research procedure used, and hopefully the
meta-analysis will detail how these influenced results. Note that, as with
triangulation, the results of meta-analysis can be misleading if all of the
studies included contain the same errors or biases. This is particularly likely
if these studies only embrace one method, as is commonly the case (usually
only experiments or statistical analyses), but might occur simply because of
biases in the worldview of the community (or its journal editors) responsible
for the studies covered. The analysis in this chapter suggests that meta-
analyses should be performed across methods wherever possible. Since
coding of research protocols is an essential element of meta-analyses, it is
conceivable that qualitative analyses might be embraced by this method.
Efforts to perform meta-analysis are made difficult by the tendency of the
authors of individual studies to be unclear about various elements of their
research procedure, and for there to be no clear dimensions along which
different studies can be compared and contrasted. This problem is especially
difficult when studies using different methods are compared. The 5W
questions point toward solutions to both problems, by urging individual
authors to be explicit in answering each of the questions used above, and
urging meta-analysis to proceed along these lines. Hopefully, by applauding
methodological flexibility and replication, this book will encourage the
scientific community to place even greater value in meta-analysis than is
presently the case, while not thus ignoring qualitative studies that may be at
present difficult to include.
Chapter 5
CLASSIFYING PRACTICE

Chapters 2, 3, and 4 classified elements of science that have long been

recognized: phenomena, data, theory, and method. This chapter moves into
somewhat more contested terrain. Some philosophers and many scientists
would wish that we would stop where we are. “Philosophers of science have
always steered clear of ‘psychologism’ [the study of the individual as
scientist] for fear that it would rob science of its much-prized objectivity”
(Ziman, 2000, 6). But it is not just the logic of the 5W typology of chapter 1
that propels this project forward. The questions to be explored in this
chapter have in recent decades received a great deal of attention. McCloskey
(1990) thus disdains previous attempts to “take the scientist out of science”
and pretend that the facts speak directly to scientists; instead it must be
recognized that scientists shape the evidence that they compile in many
ways.1 Some of these potential biases have long been recognized and have
indeed long motivated sceptical thought: suspicions of the human ability to
understand have bubbled to the surface from time to time within every
philosophical tradition (Collins, 1998). In particular, as troubling questions
are raised about the increasing importance of private financing of scientific
research, it is noteworthy that private financing of research, and concerns
about this, have existed as long as science has.
Two extreme views of science were reviewed in chapter 1: that it is a
perfectly rational and objective pursuit of knowledge, and that it is an
entirely subjective game in which scientists argue for positions that they

1
Kuhn showed that “’the’ paradigm or ‘the’ methods characteristic of a particular science
are a gross simplification; their real content comes in their application ... This means that
we must look at the actual practice of science rather than just its textbook formulations”
(Kincaid, 1996, 36-7). Kincaid refers specifically to problem-solving strategies, skills, and
implicit knowledge.
156 Chapter 5

like.2 The Golden Mean was invoked at the time. Freeman (1994) guides
scholars to expose the influence of power and envy on the scientific
enterprise, “and still recognize that the laws of nature cannot be bent and
cannot be broken by power and money;” he concludes that “the history of
science is most illuminating when the frailties of human actors are put into
juxtaposition with the transcendence of natural laws.” Collins and Pinch
(1993, 1) suggest that, rather than seeing science as all good or all bad, it be
viewed as a “bumbling giant” that possesses some imperfect ability to move
in the desired direction. Later (142) they suggest that scientists should be
more humble; pretending that science is perfect encourages those who are
sceptical of science. In the context of the evolution/creation debate, Kitcher
(1982) argues that false claims that science rests on infallible proofs makes it
too easy for those who disagree with certain scientific conclusions to
discredit these: neither evolutionary theory, nor the evidence in support of it,
is flawless. An understanding that science is inherently imperfect would,
ironically, have the effect of enhancing the authority of science.3 The
reverse argument can also be made: critiques of science that deny any
connection between scientific theories and reality are too readily dismissed
by the scientific community. Scientists can thus be blissfully unaware of the
biases and errors that might afflict their work.4 Moreover, the public,
exposed primarily to extreme views, is left unsure of how to evaluate
particular scientific arguments: “To raise the level of scientific literacy [in
the public], we need to better explain what science does ... [in particular] that
science works in imperfect ways” (Valiela, 2001, 268). A public that
understood the variety of biases that might afflict scientific research, but also

2
Ward (1996, 95) identifies a third extreme: postmodernists who argue that understanding
of reality cannot be separated from the symbols used to represent it. While there is
inevitably some ambiguity even in scientific texts, central messages can still be identified,
at least if the text is clearly written (Szostak , 2003, ch.2). Ward favors a focus on the
process by which scientific discoveries are first made and then spread through the
community of scientists.
3
A further irony comes from the fact that it was from the church that seventeenth century
science borrowed the idea of proof: Copernicus was told that one needed proof to disagree
with Scripture (Motterlini, 1999, 35). Other anti-science arguments reflect jealousy of the
status of science, frustration at inability to understand it, suspicion of any powerful group,
or celebration of the mysterious (Holton, 1993, esp. 154-6, 173-5). The Golden Mean
approach allows arguments to be discredited that can clearly be traced to such motives.
4
This is regrettable, for “A mind-set of critical skepticism is essential for scientists”
(Valiela, 2001, 268). While the social study of science has established that science is not a
uniquely rational activity, even the broader community of sociologists behaves as if it were
(Cole, 1992, ix). Sociologists of science may exaggerate their case out of fear that if they
recognize any role for reality scientists will just assume that this dominates (Cole, 1992, x).
Scientists do, of course, search for errors when judging the works of others; the danger
here is that they look for only a subset of possible errors, perhaps only some commonly
recognized difficulties with the preferred method and data of their field.
5. Classifying Practice 157

the fact that these do not render science impossible, would be much better
prepared to evaluate the barrage of scientific information in the daily news.5
Where does the Golden Mean lie? This is inherently an empirical
question: potential influences of both reality and society [including
individual psychology] in generating scientific results can be identified
theoretically, but there are no simple theoretical grounds on which to
determine the relative importance of these (Cole, 1992, x).6 While some
vague idea of the relative importance of these two types of influence on
average can be gained by looking across several case studies of scientific
practice, common sense suggests that the balance will likely differ across
types of science. “I think that most people working in science studies would
agree that one encounters scientists that one admires more than others,
findings that seem more compelling than others, and uses to which the
sciences are put that seem more enlightened than others” (Cussins, 2000,
331).7 In other words, the types of biases and errors that might afflict science
must first be identified. Then these can be investigated across the types of
theory, method, and phenomena/data identified in previous chapters, to
establish under what conditions particular errors and biases are most likely to
exert a key influence on scientific outcomes.8 This chapter will develop a
typology of critiques of scientific practice. The next chapter discusses how
these are likely correlated with previous classifications.
“For all their labors, the philosophers have failed to come up with any
simple, generally agreed principle on which belief in science might be safely

5
Carl Sagan warned that a public that does not understand the essence of science can
embrace magic as equally valid (Valiela, 2001, 256). Much public angst arises from
exaggerated reports of weak results (Valiela, 259). The public should be informed that
science is imperfect, and scientists encouraged not to exaggerate (Shipman, 1988). Katzer,
Cook, and Crouch (1998, vi) appreciate that “Most consumers of research will need to
evaluate various types of research.” Nevertheless, they seek a common set of criteria to be
applied in each case. They recommend an error approach, where common errors are
looked for; if scientists do not find these they can have confidence in the result.
6
Cole (1992, 234) worries that social constructivists have failed to identify empirical results
that they would accept as evidence that reality also matters. Many scientists in turn reject
evidence that social influences play some role.
7
As Cussins details, the field of science studies is characterized by different schools of
thought. Some schools evince extreme skepticism while others celebrate science; the
majority fall somewhere in between. Not surprisingly these schools of thought tend to
favor theories and methods that support these starting premises (see also Fuller, 1993).
8
Already in 1973, Stoner (xvii), reprising the pioneering work in sociology of science of
Robert Merton, appreciated that the nature and growth of science could not be understood
“without dealing specifically with the contents of science: its concepts, data, theories,
paradigms, and methods.” Yet scholars must also “distinguish the behavior of scientists as
scientists from the details of their ‘output’,” and thus need to identify the social influences
on scientific practice.
158 Chapter 5

grounded. But sociological critiques are vacuous without reference to the

specific contexts in which beliefs are held or made” (Ziman, 2000, x). By
classifying phenomena, data, theory, and method above, the stage has been
set for a superior appreciation of the strengths and limitations of different
types of science. Indeed it has already been seen that even at its best no
theory or method is perfect. And it was noted that theories and methods are
often misapplied. The next step is to recognize various errors and biases that
will affect the sorts of questions scientists ask, the degree to which they
adhere to best practice guidelines of scientific inquiry, the conclusions that
they draw from necessarily imperfect scientific evidence, and the rate of
scientific discovery or advance. Sadly there has been no previous attempt to
classify scientific critiques: “We know much more about science studies
nowadays than can be put together into a comprehensive, coherent package”
(Ziman, 3). Different scholars and schools of thought within the field of
science studies emphasize different biases or errors (see Cussins, 2000); a
classification of these can potentially bring coherence to this diverse field.

1. ERRORS AND BIASES IN SCIENCE

Critics of scientific practice often use the words “error” and “bias”
without defining their meaning. Hammersley and Gomm (2000, 151-3)
suggest three main definitions of the word “bias”. While they recognize that
this list may not be exhaustive, it suits this chapter’s purposes nicely; all of
the biases to be identified here fit within one of these meanings of the word.
Notably, at least within their definitions, it only makes sense to speak of
biases or errors if it is thought that there is an objective truth to deviate from;
nevertheless constructivists often accuse others of bias or error (156-7).
A first definition of bias involves the pursuit of a particular point of
view. Scientists will then inevitably emphasize certain aspects of a question.
The most powerful antidote for this sort of bias lies in synthesizing different
viewpoints, as recommended in previous chapters; scientists thus reduce the
risk of ignoring important aspects of the question at hand. Many relativists,
though, recognizing that some bias of this sort is inevitable (at least in human
science), urge scientists to choose a bias that is socially responsible. If
scientists thus miss important insights, though, they may be unable to give
accurate guidance to those they wish to help. 9 Hammersley (2000) notes

9
Some partisans imply that any social change is good; history suggests that misguided
efforts to help the disadvantaged could leave them worse off. An alternative justification is
that different social groups bring different perspectives and values to research. But if all
knowledge is relative scientists have no way of determining which perspectives are best.
To leap from recognition that bias cannot be eliminated to a conclusion that scientists
5. Classifying Practice 159

that such relativist arguments are often assumed rather than justified; he
identifies a handful of possible justifications and finds each wanting.
A second definition of bias would involve any systematic deviation from
reality. Hammersley and Gomm speak of sampling bias or measurement
bias as important examples. If a sample is not random, or a method of
measurement consistently under- (or over-) estimates the true value, results
will likely be erroneous. This is the sort of bias that was dealt with in
chapter 4. No method is perfect, and particular applications of particular
methods introduce scope for further imperfections. At times, such
imperfections are random and should cancel each other out over large
enough samples. Hammersley and Gomm are concerned with imperfections
that are biased in a particular direction. Note, though, that even random
imperfections are problematic, for they introduce noise into results.
The third definition is the most specific: it speaks of a tendency of
researchers to collect, interpret, and present data in a way that supports
preconceptions. This might lead them – and those who accept their results –
to a false conclusion, or away from the correct conclusion. Such a bias
might reflect religious, political, or other attitudes of the researcher. Or it
might represent the desire of all researchers to provide further support for
positions they [or perhaps those they hope to impress] have previously
advanced. Note that such a predilection might be conscious or subconscious.
Hammersley and Gomm emphasize that this bias [like the others] can affect
all methods: they speak in particular of the design of experiments, phrasing
of survey or interview questions, and interpretation of ethnographic data.
Hammersley and Gomm treat biases as a subset of errors. Other errors
include any non-systematic imperfection in the research process. That is, for
them a bias must mean errors in one direction. Note that biases then
necessarily detract from the validity of results. These random errors may
reflect researcher negligence or the inevitable imperfections in methods. In
the latter case Hammersley and Gomm speak of non-culpability.
Hammersley and Gomm emphasize the fact that in any research enterprise
scientists must start from a set of unprovable assumptions (see chapter 6); if
these are wrong the results will likely reflect this. Both negligence and non-
culpability can lead to systematic error as well.
Such definitions of bias and error can provide only a starting point from
which to identify the potential sources of bias and error. As noted above, a
fairly exhaustive list of these should be possible. As with respect to
scientific theory in chapter 3, though, there appear to be a diversity of

should abandon the pursuit of objective understanding is “to assume that an ideal that is
not, or never can be, fully realised is of no value. This is true of some ideals but not of all”
(Hammersley, 2000, 33). This is, of course, a classic Golden Mean argument.
160 Chapter 5

critiques, and there is some sense that new questions arise as science evolves.
This chapter will develop a typology of critiques of scientific practice, and
show where major biases and errors fit, but cannot pretend to engage every
critique that has or could be made. It is thus useful to begin with a fairly but
not completely exhaustive list of the types of questions that might be asked
about scientific practice.10 The following list reflects a reading of diverse
studies of scientific practice; it is much less concrete than the list of methods
in chapter 4, but much more useful than the list of grand theories at the start
of chapter 3:
x What is science, and (how) can it be distinguished from non-science?
x Who possesses scientific understanding?; What are its warrants; How is it
acquired? What does it do? (see Cussins, 2000, 330).
x Is science objective in practice? How objective can it be? What can be
done to make it more objective? Is there an external reality to be
understood?
x In particular, to what degree can evidence support (or not) a theory?; how
much scope for human judgment is there in science?
x To what extent do the perceptual and psychological limitations of all
humans limit the scope for scientific understanding?
x Are scientists biased? Can the directions of bias be identified? (How) can
bias be combatted?
x Is the institutional structure of science conducive to its practice (that is,
does it serve to combat subjectivity, bias, etc,)?
x In particular, how are scientists trained, and is this training appropriate?
x Is science too specialized? What is the role of interdisciplinarity?
x Are the values held by the scientific community appropriate? (How) are
they inculcated, and are they widely held?
x In particular, are scientists likely to exaggerate the evidence in support of
their conjectures, or even obscure evidence that contradicts these?
x Are scientists biased toward producing results that support the economic,
social, and political status quo?

The answers to these questions will have an obvious effect on people’s

willingness to believe particular scientific statements. They will also suggest
whether a superior institutional structure and/or culture for science are
advisable (here and in chapter 8).

10
Madden (1989, vii) provides a list of persistently recurring questions in the philosophy of
science. These involve theory construction, hypotheses, causality (especially the
universality thereof), lawfulness [that is generalizability], probability, experimental
methods, the problem of induction, and the problem of distinguishing evidence from
assertion. Most of these have been discussed in previous chapters. The last issue deals
most directly with the questions of bias to be engaged in this chapter.
5. Classifying Practice 161

2. TYPOLOGY OF (CRITIQUES OF) SCIENTIFIC

PRACTICE

The 5W questions are a useful guide yet again, but will prove somewhat
less exclusive here than in earlier chapters: for example, the incentives faced
by scholars (Why?) reflect the institutional structure in which they operate
(Where?) Recall from chapter 1 that some fuzziness in boundaries is
acceptable if undesirable for a classification. It will still be possible to
identify distinct subsidiary questions under each heading. The classifications
developed in previous chapters will prove useful in pursuing deduction:
when investigating ‘who?’, it will prove useful to examine individual, group,
and relationship agency among scientists. The classification of phenomena is
particularly noteworthy: each element in the typology of critiques can be
understood in terms of causal links toward the phenomenon ‘science’ (or a
definition of science, in the case of ‘what?’). The diversity of analysis in the
field of science studies suggests that a fairly exhaustive classification is
possible by combining as in previous chapters a deductive approach with an
inductive reading of the science studies literature.

2.1 Who?

Are scientists objective? No. Scientists are not computers but complex
organisms with competing genetic drives and quite imperfect perceptual
capabilities. They will often see what they want to see, or what they think it
is important to see, and can be expected to miss much that either conflicts
with or seems entirely extraneous to their view of the world. Are scientists
completely subjective, then? No. While some methods are open to a greater
exercise of subjectivity than others, all require some confrontation with
reality in a manner that limits the researcher’s ability to ignore evidence that
is uncongenial. This fact will form a key element of the very definition of
science to be proffered below.
“Philosophical theories of science are generally theories of scientific
rationality. The scientist of philosophical theory is an ideal type, the ideally
rational scientist” (Giere, 1988, 2). Social constructivist views go to the
other extreme, and deny any influence of reality upon scientific thought.
Giere (1988) argues for an “in-between view” – that is, a Golden Mean – that
speaks of scientists using their imperfect perceptual apparatus to interact
with reality. Scientists strive for explanations that fit both reality and “their
own emotional longings among the alternative world views” (Feuer, 1995,
ix). Writers such as Bunge (1998) and Ziman (2000) bemoan the amount of
ink that has been spilled, notably but not exclusively in recent decades, in
defending one or the other of these untenable extremes while glorying in
162 Chapter 5

the impossibility of the alternative. This book can proceed to the less
sensational but more useful task of identifying the degree and sources of
bias. To ask “who is the scientist?” is to query the degree to which any
human(s) can shake off both limitations in ability and selfish motives in
order to pursue scientific understanding.
One insight of sociology of science is that scientists are not superhuman
but “merely” humans with a particular expertise (Fuller, 1997, 8-9). It is
useful to start, then, with a sense of basic human capabilities. Humans have
likely been selected genetically for certain sorts of scientific inquiry
(Kantorovich, 1993, 201-2). Humanity’s hunter-gatherer ancestors needed to
make judgments in their pursuit of food, clothing, shelter, and other basic
needs, and were likely selected for an ability to reach accurate perceptions of
at least some elements of their environment. Since they dealt with changing
environments they would have been selected not for particular attitudes
appropriate to one environment but an ability to evaluate various sorts of
evidence. Since hunter-gatherers cooperated, they should have been selected
for an understanding of some elements of human as well as natural science.
Along with a selection for certain sorts of inquiry humans have also been
selected for general cognitive abilities, and thus likely have some ability to
tackle any question. Nevertheless, human ability to tackle some questions –
due both to limitations in ability and biases – may be very limited.
Fuller (1997, 28) urges the identification of human cognitive
imperfections, and then of the degree to which these affect scientific practice.
He notes that certain human failings – such as a tendency to deviate from
rules – can at times yield good scientific outcomes, though this is likely not
the general case. Halperin (1997, 21) lists 21 common “fallacies” in how
humans make and evaluate any argument: these include logical problems
such as circular reasoning or assuming that correlation implies causation,
emotional problems such as succumbing to appeals to pity, pride, ignorance,
or tradition, and errors in judgment such as accepting appeals to authorities
with no particular expertise, or accepting evidence that is inherently
unknowable (such as reports that unreported crime is increasing). Later
(1997, 215-6) Halperin lists several biases in decision-making: humans
exaggerate easily recalled information, are excessively optimistic, and allow
emotional states and preferences to overwhelm reason. Scientists are
hopefully less subject to these various biases – scientific training urges
scepticism of authority, for example – but are not entirely immune (see
Kruglanski, 1989). Valiela (2001, 268), notably, worries, as does Halperin,
that scientists all too often criticize fellow scientists rather than the
arguments that these put forth. This human tendency to personalize
arguments is not conducive to scientific advance.
5. Classifying Practice 163

Giere (1988, ch.6) argues that scientists are satisficers: they seek theories
that fit reality, but are not overly concerned that there is likely another theory
that will fit reality even better [especially if this may fit less well with their
worldview]. Valiela (2001, 269) also suggests that scientists are too prone to
inertia in this regard. This is not entirely a bad thing: scientists plagued by
extreme doubt would be unable to proceed. But if Giere is right then
scientists may show too little openness to novel theories and also to evidence
that the existing theory does not flawlessly reflect reality.
Table 2.1 suggests that it is generally useful to look not only at general
human characteristics but also individual differences. Particular scientists
may have qualities that encourage certain sorts of discoveries or certain sorts
of bias. Ralph Waldo Emerson in 1837 observed that “character is higher
than intellect” for a scholar (in Glassick et al., 1997, 62). What personality
characteristics seem broadly conducive to scientific advance?11 Emerson
stressed patience, courage, and self-trust, to which Glassick et al. add
integrity and perseverance. Beveridge (1957) lists curiosity, honesty,
openmindedness, imagination, tenacity, and humility.12 Einstein suggested
that there were three types of scientists, those who worked to exercise their
talents, those who viewed science as a career and could just as easily have
pursued other careers, and those who viewed science as a vocation; he
argued that the last group had an impact much greater than their numbers
(Beveridge 1957, 143).
Beveridge (ch. 5) expands on the role of imagination. Solutions to
problems tend to come from the subconscious when the scientist has
temporarily diverted their attention from the problem: when they are in the
park or the bath. Such moments of inspiration, however, occur only after
lengthy periods of conscious focus on the problem. Beveridge recommends
that scientists read widely and talk with a diversity of others about the
problem, in order to give their subconscious the widest range of information
to synthesize. After a moment of inspiration, they must be able to recognize
that this was imperfect. Beveridge suggests, indeed, that the best scientists
are those best able to recognize errors. He quotes Planck, Einstein, and
Helmholtz; each speaks of many errors on their road to discovery, and how
they were guided by a vague vision of the direction to go. It may be that the
present structure of science (see below) limits the exercise of imagination.
Specialization and the pressure to publish combine to discourage reading

11
In investigating this question, scholars should be wary of spurious correlation: the
psychological attributes, discoveries, and biases of a community may reflect its historical
evolution and/or institutional structure (‘where’ and ‘when’ below; see Bromne, 2000).
12
Amabile (1996) suggests that independence, sensitivity, and a preference for complexity
are also conducive to creativity of various types. An ability to concentrate may also be
crucial, but one must also be able to let go.
164 Chapter 5

widely. The formation of narrow elite groups with a particular worldview

discourages the diverse conversations that Beveridge recommends. The
pressure to publish drives leading scientists to spend almost all of their time
at work, decreasing the opportunity for moments of inspiration. Beveridge
notes that excessive worry or focus on a problem will discourage acts of
inspiration. And the celebration of works that provide research agendas for
others acts to discourage efforts to recognize errors before publication (while
errors which will not encourage further research are instead papered over).
Beveridge suggests one key advantage of humility: by recognizing
collective ignorance scientists are encouraged to appreciate the limits of their
own discoveries (he quotes Newton and Pasteur on how small their
discoveries were relative to human ignorance). As well, the humble should
be less likely to exaggerate the importance of their ideas, and paper over
their defects. They would also be less likely to give policy advice that
simply ignores their areas of ignorance (Szostak 1999, 2003). The arrogant
often fail to analyze the sources of their success, and thus can more readily
be seduced into accepting the existing structure of the scientific enterprise
simply because they are treated well within it.13 And humility encourages
openmindedness; in facing one’s own ignorance one becomes willing to seek
enlightenment in a variety of directions. The arrogant may be guided to
focus on trivialities by a false belief that they already have the answer.
Kant responded to a 1783 newspaper query about the nature of
enlightenment with the oft-quoted words: “have the courage to know”. The
ethical scientist needs courage in many ways: to face the unknown without
fear or self-deception, to battle inappropriate scientific institutions, to face
the fact that the best scientific project is often not the best career path, to
recognize insights which threaten one’s existing schema system, and to
expose one’s thoughts to the world honestly and clearly rather than
attempting to hide their defects. Indeed, courage may be the single most
important characteristic required of a modern scientist.
In terms of a variety of other personality dimensions, diversity is likely
best. Beveridge is right to emphasize that the successful scientist must be
motivated in order to collect diverse bits of information, and to work through
the errors inevitably embedded in any inspiration. The highly motivated,

13
Getman (1992) argues that (especially leading) scientists like to see arrogance in their
colleagues, for this serves to reinforce their own inflated view of their importance [and
people like to hire people like themselves anyway]. A humble colleague, by their very
existence as well as their research efforts, questions the pretenses of the elite. There may
thus be selection away from humility – which may foster gender and ethnic discrimination.
This tendency will be exacerbated by the training of the next generation of scientists:
rather than telling them of our collective ignorance, their teachers celebrate even trivial
accomplishments, and thus signal that arrogance is not just acceptable but de rigeur.
5. Classifying Practice 165

though, may not let their brains rest in order to achieve inspiration. Certain
sorts of discovery may thus be more likely from the moderately motivated.
Likewise, since extroversion is conducive to interaction with others and
introversion to independent thought, shy people may provide insights that
extroverts cannot. Both holistic and analytical thinkers should be heard, as
should both the calm and emotional. Mood swings may encourage certain
sorts of discovery: Storr (1996) treats Newton and Einstein as examples.
Conscientiousness has obvious advantages for any complex endeavor, but
untidy thinkers may be able to draw connections between elements the
conscientious would never juxtapose. Organizational theorists suggest that
creative firms must draw together people with both diverse talents and
personalities. Scientists, and especially elite scientists, may tend to hire
those who resemble themselves, at least in terms of some personality
dimensions (Getman, 1992, 260). What is lost if the shy or holistic are thus
selected out? Very little research has been performed along these lines.
In addition to biases inherent in genes and individual personalities, a host
of societal influences can generate further biases. Table 2.1 again proves
useful, as all of the major categories could generate some sort of bias:
scientists threatened by volcanoes or earthquakes may have different
attitudes toward nature, as may those in a disease-ridden environment. The
art of a society may influence its scientists, and its economic well-being will
almost certainly have an influence on the amount if not direction of scientific
research. Scholars of science have, though, tended to emphasize cultural,
political, and social considerations.
Scientists are part of a wider society and, even if they consider themselves
rebels or outsiders, will have internalized some of the values of that wider
society. They will thus wish to uncover scientific understanding that is in
accord with these values. This may seem of little import in natural science,
but recall the religious furore occasioned in the past over heliocentric views
of the solar system or theories of biological evolution, and which may well
be occasioned in the future with respect to genetic engineering. Even less
controversial elements of science may reflect cultural values. Newton saw
himself as illuminating the order of Godly creation: who can guess to what
extent scientists exercise a bias toward order in their investigations for
similar cultural reasons? Fuller (1997, 56-7 suggests that scientists generally
believe in progress so powerfully that it is impossible to convince them
otherwise. With respect to human science, cultural values generally play an
even greater role: if a society believes that art is ineffable or market
outcomes fair or gender distinctions normal, then scientists may be biased
toward finding these same results.
Michel Foucault is justly famous for emphasizing the political element in
all of this. Those with power in a society will encourage cultural attitudes
166 Chapter 5

that serve to maintain their power. The values that scientists embrace will
thus often not reflect the will of the people but of the elite. And thus
scientific biases may serve to support the status quo. Elites, moreover, act to
directly influence scientific outcomes through their financing of research (see
below). Foucault spoke in particular of how scientific analyses of
criminality and insanity served as mechanisms of social control, but intended
his analysis to apply much more broadly.
Scholars have also written at length of late about the social element in
values. Society has different expectations of people depending on their
gender, ethnicity, and class/occupation (and also sexual orientation, age,
marital status, and so on). White men can be expected to be more favourably
disposed toward attitudes that support their social status than are women or
minorities or the poor. Since the scientific enterprise has long been
dominated by white men, it likely reflects such biases.14 Again, this problem
might be expected to loom larger in human science, but the priorities chosen
by natural scientists are also affected (as in choosing to investigate prostate
versus breast cancer).
Sadly, much of the relevant literature starts from the assumption that such
biases are pervasive. The work of Foucault and many others can leave little
doubt that such biases do influence scientific outcomes. But innumerable
counter-examples can also be pointed to of white male scientists arguing
against cultural attitudes that serve the elite and/or discriminate against
women or minorities. These can be attributed to two sources. First, culture
is not as monolithic as some scholars of culture presume (Szostak, 2003),
and thus all humans have some capacity to disagree with their society’s
values. Scientists do this in their everyday lives, and are even more likely to
do it in their lives as scientists. Secondly, scientific methods themselves, as
noted above, force scientists to face up to reality and ask questions and reach
conclusions that they might have preferred to avoid.
The preceding analysis has proceeded largely at the individual level. The
typology of theory in chapter 3 would suggest that analysis of groups and
relationships might also be important. As has already been suggested, these
might shape individual personalities and behaviors. Indeed, Fuller (1993, 5)
notes that his analysis of science bears a strong similarity to that of Giere,
except that Giere emphasizes individuals and Fuller groups; elsewhere Fuller

14
Cussins (2000, 350-1) also reviews various arguments that there are inherent masculine
and feminine ways of doing science – say because notions of sexuality influence by
analogy perceptions of the world, or because women are inherently less likely to see nature
as a separate entity. The idea of a distinct feminine perspective has been attacked for
ignoring both personality and ethnic/class differences among women.
5. Classifying Practice 167

applauds Latour and others for stressing the role of scientific relationships
(see Galison, 1996).
A belief in science cannot depend on the virtues of individual scientists,
but must instead hang on the ability of communities of science to more often
than not choose good ideas over bad (Ziman, 2000, 160). Can group
processes overcome (some) individual-level biases? Recall that individual
scientists are influenced by biases, but are often able to rise above these
biases in order to achieve insights that challenge widely held beliefs and/or
offend the powerful. At the level of communities of scholars, if scientists
can be relied upon to take conflicting results seriously, and attempt to test
these against each other by exposure to real world evidence, it can be hoped
that any tendency toward overcoming biases in research will in the long run
be reflected in scientific understanding. That is, no matter how strong the
tendency toward bias, if there is some conflicting tendency, and if
contradictory results are evaluated seriously by the scientific community,
bias will be overcome in the long run. Note, though, the two big “ifs” in the
preceding sentence. As noted above, scientific evidence is never perfect, and
no scientific method infallible. Real world evidence may thus not always be
powerful enough to overwhelm bias. Social constructivists argue that
empirical evidence is often incapable of solving scientific disputes (Cole,
1992, 5). Collins and Pinch (1993) provide several examples for which they
claim that certain theories were accepted or rejected by a scientific
community despite ambiguous evidence. Moreover, the long run can be
very long, and in human science especially the world may have changed long
before scientists arrive at the correct interpretation of how it works. The
point to emphasize here is that the question of bias is essentially an empirical
question, not solvable by theoretical speculation alone, and one that has been
addressed to a very limited extent.
A naive view of science would hold that scientific debates are decided on
the basis of which side marshals the most persuasive evidence and logical
argument. To be sure, these are important; otherwise one could have little
confidence that science would lead to enhanced understanding. But humans
are not (entirely) rational calculating machines. Students of rhetoric show
that metaphor and story are important in making a compelling argument, and
that this is as true in science as in everyday life. Thus scientists buttress
their arguments with detailed (often hypothetical) examples, and strive to
show how the relationship they posit is similar to another already widely
accepted by the scientific community. McCloskey (1990) stresses the
advantages of this: it is harder for a scientist to tell a “big lie” if their
arguments are judged in terms of four criteria rather than two. To be sure,
scientists in evaluating the evidence and theoretical arguments of another
scientist can beneficially ask themselves if they can imagine compelling
168 Chapter 5

examples of the theory at work, and also whether the theory accords well
with other elements of their scientific view (schema) of the world. However,
a flashy story and facile analogy may too easily compensate for faulty logic
and limited or misleading evidence. The antidote, again, is the conscious
application of multiple methods to different types of data, and the clear
statement of how well the evidence accords with the theory. Being aware of
the rhetorical approaches that humans use and heed would also be helpful.15
Not surprisingly, some rhetoricians suggest that rhetoric trumps validity in
determining scientific outcomes (see Brown, 1992); Montgomery (1996)
explores language use in science, and recognizes its importance, but reaches
the Golden Mean conclusion that while rhetorical practices have an
influence, scientists can still aspire to considerable accuracy in their
description of the world.
In the case of social biases, much can be achieved by the entry of large
numbers of women and ethnic minorities into the world of science (though
some ethnic groups, notably Aboriginals, are still hugely underrepresented,
and so is the working class and especially the underclass). Scientific
communities can also benefit from an expanded understanding of cultural
processes, and especially the political element in these (see Szostak, 2003,
for a discussion of these linkages). Scientists can thus become more aware
of potential biases. More generally, Palys (1997, 399) stresses the
importance of not privileging any particular point of view; if scientists do so,
they will then place alternatives at an unfair disadvantage by judging these in
terms of the standards of their favored view. Scientists should strive for
mutual respect and understanding. But most centrally perhaps, scientists can
strive to ensure that the scientific enterprise is open and empirical: that it
takes conflicting views seriously, and subjects these to analysis through the
use of (generally more than one) scientific methods.
Disciplinary biases may be of even greater import than cultural, political,
or social biases. Moreover, disciplinary biases can reinforce the other types
of bias (see Szostak, 1999). Scientists face much greater pressure to accord
with disciplinary values than with wider societal values: their ability to be
hired, promoted, published, heard, and cited are all influenced by their
willingness to act in accord with disciplinary preferences with respect to
theory, method, data, and phenomena of study. Those who insist on using
unusual theories or methods, or examining unusual topics, may find
themselves ignored and even unemployed. It would be an unusual human
being who would be completely impervious to such pressures. Scientific
curiosity may drive research, but is itself shaped by the community in which

15
Halperin (1997, ch.3) warns scholars to beware of, and defend against, emotional
language, name-calling, vagueness, labelling, ambiguity, arguments from etymology,
reification, bureaucratese, misuse of definitions, euphemisms, and misuse of analogy.
5. Classifying Practice 169

scientists operate (Ziman, 2000, 22-4). Even if consciously dedicated to the

selfless pursuit of understanding, scientists will self-consciously constrain
their quest in order to meet disciplinary standards. This book has and will
illustrate the dangers of such disciplinary preferences, and hopefully will
encourage a greater scientific openness to the examination of all phenomena
by all methods and with recourse to the widest range of theory. It
encourages specialized researchers to greater flexibility, and recognizes a
clear role for interdisciplinary researchers in ensuring that the broadest range
of theory and method are applied.
One critical point to be emphasized here is that biases are generally
subconscious. Some scientists, to be sure, may consciously try to please the
powerful, but sycophancy is not by far the primary danger. Cultural attitudes
are for the most part internalized within the subconscious; they thus affect
individuals without them being consciously aware, and can thus become
embedded in scientific theory without any scientist being consciously aware.
Much of the antipathy of scientists to modern sceptical analysis comes from
this simple misunderstanding. Scientists who consciously strive to perform
their research objectively are deeply insulted by accusations of bias.
Misunderstanding the (often unclear) nature of the attack, they dismiss it.
And by dismissing it they fail to ask what sorts of bias might afflict their type
of research, and thus be on guard against it.

2.2 What?

What do scientists do? By definition, they do science. The 5W questions

thus suggest a novel approach to the question of “what is science?” Rather
than trying to define science in terms of first principles, a task that eludes
philosophers to this day, this section will define science in terms of what
scientists actually do. This approach accords with the advice of Charles S.
Peirce to define science “in terms of the activity of those whose lives are
animated by the desire to seek truth” (Kent, 1987, 18). Likewise, Taylor
(1996, 4), noting the problems in defining science, suggests that, “we might
be tempted to turn for answers to those who ought to know best – scientists
themselves. After all, science might be understood as ‘what scientists do’.” 16
Ragin (2000) and Ziman (2000, 5) both note that sceptics often focus
exclusively on the similarities between science and non-science, and ignore
the differences, while defenders of science do precisely the opposite. The
similarities between science and non-science can be expressed in terms of
the various human imperfections discussed in this chapter: scientists and

16
Ziman (2000, 12) argues that scholars should not aspire to a simple definition of science,
precisely because science is inherently multidimensional and diverse. But that should not
stop us from knowing what it is (Ziman refers to science as a natural kind).
170 Chapter 5

non-scientists alike are afflicted by biases, perceptual limitations, competing

influences, social pressures, and so on. By focussing exclusively on these
one can soon reach the conclusion that there is no such thing as science.
Oakley (2000) is hostile to this conclusion on the practical grounds that it
precludes humans from enhancing our understanding of the world, and thus
solving the problems that humanity faces. To avoid this conclusion,
differences between science and non-science must be identified.
The differences between science and non-science are to be found in
chapters 2, 3, and 4. Science involves, and can thus be defined as, the use of
theories which specify answers to each of the 5W questions outlined in
chapter 3, examination of well-defined phenomena or causal links (properly
operationalized) as discussed in chapter 2, and the use of one or more of the
methods outlined in chapter 4.17 Scientists, by following these procedures,
are able to limit bias to a much greater degree than non-scientists who do not.
“Science may be socially produced, but particular scientific conclusions may
be more or less credible and trustworthy” (Oakley, 2000, 302-3). But this is a
result of the distinction, not its essence: even if it is doubted that this
difference is large, or even if it is speculated that for some questions it
disappears (as some might do with respect to certain aspects of art, for
example), a distinction has been identified on procedural grounds that still
stands.18 Of course, the distinction may only be of interest because of the
possibility that these procedural differences have important and beneficial
effects on the outcomes of scientific research. Still, the discussion has at
least moved beyond the question of whether science exists, and can proceed
to more focused questions about whether this matters.
Mahner and Bunge (1997, 186) provide a definition of “a science” which
also speaks to phenomena, theory, and method. Lacking well-defined
classifications of these, they can only insist upon study of some set of
phenomena with logical theory and checkable and explainable method(s).
They also speak of outlook: a science should be characterized by logical
values such as exactness and consistency, semantic values such as clarity,

17
Postmodernists often note that regular people have “theories” too (Seale, 1999); these are
often poorly specified and almost never tested with a scientific method. Early in this
century positivists strove to define science in terms of falsifiability – philosophy and
astrology could be viewed as non-science if they could not identify empirical tests that
could falsify their arguments. As philosophers came to appreciate that no statement could
be proven false empirically, this definition collapsed. Many philosophers, including
Feyerabend, Foucault, and Derrida, thus concluded that there was no such thing as science.
18
Is science rational? We could ask, first, if its means further its goals (of enhanced
understanding). Or we could ask if it pursues a method(s) such that its theories are tested
in the real world, for it is inherently rational to support beliefs with evidence. Lambert and
Brittan (1992) conclude that science cannot be distinguished in either way from non-
science in terms of rationality. This suggests the value of a procedural distinction.
5. Classifying Practice 171

methodological values such as testability and reflection on method, and

attitudinal values such as openmindedness. All of these values have been
advocated at some point in this book. The logic of earlier classifications
suggest their inclusion here too: science itself was unpacked in terms of
phenomena/data, theory, method, and practice in chapter 1, and thus a
definition of science might be expected to speak to all four. Moreover, it
was found when asking the ‘what?’ question in chapter 3 that both attitudes
and actions should be addressed by science; the definition above just speaks
to actions. However the analysis of this chapter suggests that it may be hard
to clearly distinguish science from pseudo-science on attitudinal grounds, for
most/all scientists are guilty of some attitudinal flaws. While scientist
attitudes are a valid concern, and scientists should certainly be urged to both
recognize and attempt to minimize their biases, practical considerations
suggest that a definition in terms of actions alone will be most useful.
Kincaid (1996) is sceptical that a precise distinction between science and
non-science is possible. Nevertheless he thinks it useful to identify
characteristics of science. He recognizes that previous attempts to define
science in terms of some unique strategy, such as a unique scientific method,
have failed. “We must instead look in detail at the various different claims,
kinds of evidence, and concrete methods that are used ...” in order to
ascertain whether a certain activity deserves to be called a science (1996,
26). Later he identifies symptoms of good science: that it relies on evidence,
seeks explanation, uses theories, and reflects a reliable process. The
definition of science proffered above draws on the characteristics and
symptoms identified by Kincaid. By providing precise and exhaustive
definitions of theory types and methods we are able to go a step further and
identify a precise demarcation between science and non-science.
The process of doing science involves the “discovery” of new theories or
empirical regularities, and the verification of these: historians of science tend
to emphasize the former and philosophers of science the latter (Langley et
al., 1987, 3). Yet these are mutually supportive activities. A definition of
science must thus embrace both theory and method, while permitting
individual scientists to specialize in one or the other. Notably, philosophers
have devoted much more attention to how scientists should test theories than
how they should develop these in the first place. Langley et al. (1987) argue
that discovery can be treated in large part as a particular type of human
problem-solving: scientists thus develop heuristics, or rules of thumb, that
guide their efforts to solve scientific puzzles, but some heuristics are more
productive than others. Science will advance most rapidly if, along with
appropriate procedures for verification, its rules of thumb for discovery are
optimal. The analysis of this book suggests some important guidelines here:
172 Chapter 5

asking the 5W questions of any problem, and not arbitrarily limiting the
range of phenomena, theory, and method to be studied.
Philosophers have uncovered no simple definition of science that serves
to include what is generally thought of as science and exclude that which is
not. The definition above, derived as it is from observations of scientific
practice, serves this goal very well. Note that the inclusion of mathematical
modelling among methods allows some elements of mathematics to qualify
as science: this discipline is often a challenge for those who would define
science. However, while some applied mathematics is thus science, much
mathematical theory is not scientific but philosophical (see chapter 3); as
with philosophy more generally this sort of analysis can provide important
inputs into science, but is not itself science. By including experience and
intuition among methods, the study of art is also included within science.
Philosophers have long attempted to define science in terms of “the
scientific method” rather than twelve scientific methods. The identification
of diverse but equally valid scientific methods suggests that this must always
have been a difficult enterprise. Gower (1997, preface, 7) suggests that
“People do not now believe in scientific method” and thus they doubt that
philosophy has much to say about science. Hopefully the classification of
method solves this particular philosophical muddle. Can some commonality
across all these methods be identified that might serve as a more precise
definition of science? It might be posited that all reflect a desire to confront
scientific theories with evidence from the real world. But does intuition
really engage the real world? While the general principle of confronting
reality is important, a Wittgensteinian conclusion seems warranted: the list of
a dozen methods is more concrete than any brief statement of overarching
principle could aspire to be.
While no phenomenon lies outside the purview of science, good science
requires the careful specification of the phenomena involved in any research
project. Too much research in human science either does not identify which
phenomena or links are being studied, or does so only in terms of vaguely
defined terms (such as culture, which is defined in hundreds of different
ways). Since the list of phenomena of interest to human scientists is both of
manageable length and susceptible to hierarchical organization, there is no
valid reason for such sloppiness. In chapter 2 it was argued that scientific
concepts should either be carefully specified in terms of phenomena, theory,
or method, or jettisoned from the scientific lexicon. Research that involves
well-specified theory and at least one scientific method, but fails to carefully
identify the phenomena or causal links under investigation, must be viewed
as either non-science or bad science.
As noted in chapter 2, scientists can usefully identify patterns in the data
for which they have no theoretical justification. As long as these patterns are
5. Classifying Practice 173

not just imagined, but result from the proper use of scientific methods, and
are inputs into the process of developing theory, they are part of science.
Scientists should be careful in their use of such regularities until they
understand why these occur.
Early in the last century, many scholars tried to establish the “unity of
science” (see Neurath et al., 1970). Positivists often imagined a unity of
science grounded in method (Potter, 2000, 69), though they tended to an
unnecessarily narrow view of acceptable scientific methods. The idea is no
longer popular (Lenoir and Gombrecht, 1996). It is now appreciated that
there are huge differences in theory and method across disciplines (Fuller,
1997, 13-14). Ian Hacking has argued that scholars of science should
emphasize harmony rather than singularity: science involves quite different
types of endeavor, but these are interconnected (Galison, 1996). The
definition of science developed above, and indeed the entire approach of this
book, is consonant with such a view of science. It was argued in chapter 2
that causal links exist between most pairs of phenomena; science is thus a
huge complex enterprise that should strive for a mutually consistent
understanding of every link. In doing so, scientists should be open to the full
range of theory and method, and wary of every possible type of bias.
Scholars of science have long been curious about “boundary work”: when
individuals operate between different types of intellectual enterprise.
Necessarily, such boundaries must be identified before boundary work can
be analysed. Much of this book has served to identify boundaries between
different types of science. Here a way of defining the most central boundary
of all, between science and non-science, has been developed.
Pseudosciences such as astrology might be excluded from this definition of
science on two grounds: astrologers do not in general specify causal agents
or relationships very well, nor do they test their conjectures against reality in
terms of any of the dozen methods. Astrologers may make a case that they
rely on intuition or experience, but it was argued in chapter 4 that relying on
these exclusively is bad scientific practice: the scientist must strive to ensure
that their intuition is not biased. That is, the very method of intuitive
analysis requires that scientists analyse their intuition, not just blindly accept
it. Astrologers that can point to a carefully articulated theory and the careful
use of at least one method other than intuition, could be deemed scientific:
again the purpose here was not to exclude certain groups a priori but to
establish objective principles by which to judge how scientific an enterprise
is. Everyday experience is also excluded from the definition of science
because it is rarely formulated in detailed theoretical terms, and rarely
verified by other methods. As Fuller (1997, 25-30) notes, some definitions
of science have suggested that science could be performed by anyone; the
definition here suggests that people in their daily lives could, but generally
174 Chapter 5

do not, achieve a scientific ideal. Likewise, much of what is called

technology can be distinguished from science by its reliance on non-
theoretical insights; engineers, for example, have developed various rules of
thumb grounded primarily in experience rather than theory.
Other means of identifying pseudosciences have been proposed, but all
are problematic. Kitcher (1982, 48) suggests that pseudosciences pursue an
opportunistic problem-solving strategy, apply tests that rely on convenient
auxiliary assumptions, ignore unresolved problems, and have limited
applicability to diverse areas of research [Kitcher thus downplays the
importance of idiographic analysis]. Note that all of Kitcher’s criteria call
for the exercise of judgment; few scientists, perhaps, could entirely escape
guilt with respect to all of these. Dupré (2001) suggests that bad science can
readily be comprehended in terms of the motivations of scientists. A
deliberate effort to both identify and overcome biases would thus be a sign of
good science; as noted above, scientists could more diligently confront their
biases. Mahner and Bunge (1997) suggest that a science must be
characterized by theories that are complementary to the theories of other
sciences. Such a criterion might exclude revolutionary science. Moreover,
as a criterion to identify pseudo-science this is more than a bit tautological:
pseudo-sciences are often decried precisely because of inconsistent theory.
Science, according to the definition proffered above, thus comprises most
of what happens within universities, and only very little of what happens
elsewhere. The key exception is that much of philosophy and mathematics
can be distinguished from science to the extent that it relies on reason rather
than evidence. But clearly there are cases where philosophy shades into
science. Indeed the enterprise in this book relies on a mixture of
philosophical argumentation and scientific analysis. Scholarly specialization
creates an unfortunate gulf between the careful reasoning of philosophy and
the empirical evidence of science; science must rely on both logic and
evidence.19 Philosophy can also be distinguished from science to the extent
that it engages questions of how the world should be rather than how it is.
Yet such analyses should be grounded in a scientific understanding of reality
(Szostak, 2004).
Imre Lakatos, in developing his ideas of degenerating and progressive
research programs, hoped to shift the demarcation debate away from science
versus non-science toward good versus bad science. Moreover, he did not
want to prohibit any scientist from pursuing bad science, though he thought
journal editors and granting agencies might want to have standards of good
scientific practice firmly in mind when making decisions (Motterlini, 1999).
It is an empirical question worthy of inquiry as to how valuable is research
that falls a little bit below the standards set above. If it has value, and it

19
I thank Barry Blesser for drawing this issue to my attention.
5. Classifying Practice 175

would have been difficult to rectify the shortcoming, the research may have a
place within the scientific enterprise.
This definition of science not only provides a response to those who
doubt the existence of science, but also to those who would define it so
narrowly as to exclude much university research. Philosophers and scientists
through the ages have spent a great deal of time arguing for the priority of
certain phenomena, types of agency, theories, and methods, and often for the
complete exclusion of others from the domain of science (just as
philosophers have generally done with respect to ethical theory; Szostak,
2004). A simple answer can be provided to questions of which phenomena,
theories, and methods qualify as scientific: all of them. Every theory that
answers the 5W questions qualifies. Every one of the methods listed in
chapter 4 qualifies. And every phenomenon deserves scientific attention.
The various classifications developed in this book have shown that all
deserve a place in science.
Much of the criticism of science comes from humanists. It is perhaps no
coincidence that natural scientists often exclude the humanities from their
view of science. This both reflects and encourages humanist suspicions. Yet
the classifications developed in previous chapters suggest that the subject
matter (art, various aspects of culture), methods (primarily textual analysis,
but also intuitive appreciation), and theories of the humanities deserve a
place in science (though some sorts of philosophy, as noted above, likely
deserve special treatment). This analysis supports Ziman’s (2000, 26)
suggestion that the portrayal of humanists and scientists as two distinct
cultures is – or at least should be – hugely overdrawn.
The defining characteristics of the humanities have changed over time. In
the nineteenth century, the humanities were commonly distinguished from
objective science in terms of their emphasis on the meaning and purpose of
phenomena. As was argued in chapter 3 with respect to interpretivism, this
perspective should be incorporated into the scientific enterprise. Early in the
last century, humanists often stressed the idiographic nature of their analysis;
again this hardly renders them unscientific (and note that humanists often
draw broader lessons from their idiographic research). Texts have always
been important: “The humanities might almost be said traditionally to have
been elaborate exercises in various kinds of reading and writing … ”
(Kervin, 1997, 9). Some have nevertheless emphasized the intuitive
perception of images and ideas. Both methods are important far beyond the
humanities. The point to emphasize here is that both the favored method and
subject matter of the humanities are suitably scientific. A minority would
define the humanities in terms of a holistic approach to everyday life; it has
been seen that science requires a synthesis of specialized knowledge and can
thus encourage greater interaction between humanists and other scientists.
176 Chapter 5

Finally, some would define the humanities in terms of style of expression;

while appreciating that style may be of greater importance in explaining art
than atoms, scholars should nevertheless be wary of an attitude that would
emphasize style over substance. Works that eschew substance for style
would fail to qualify as scientific.
Natural scientists are also sometimes sceptical of human science more
generally. It is worth noting, then, that the analysis of this book, and this
chapter especially, is human science analysis. That is, it has assumed that
science can be studied just like any other social process (see Ziman, 2000, 7).
It would be ludicrous to embrace the defence herein of the distinctiveness of
science while clinging to the belief that human science is a lesser brand of
science. To be sure human scientists, who must cope with intentionality, can
be differentiated from the bulk of natural scientists, who need not engage
intentionality, in the following way: human scientists have recourse to a
much wider set of theory types and methods than natural science (and engage
a different set of causal relationships, though there are important links
between natural and human phenomena). Only a narrow definition of
science could exclude human science on this basis.20
Note that the ability to respond to one extreme attitude toward the
definition of science is aided by the ability to respond to the other. Those
who wish to define science narrowly often fear the spectre of defining it so
loosely that it ceases to exist. But it has proven possible to define science
broadly, and yet still identify clear limits to science. And these limits accord
very well with what both the scientific community and general public tend to
think of as the boundaries of science.

2.3 Why?

Why does the scientist do what they are doing? The question can be
asked on two levels. The first deals with the goals of humans in general.
Why do people do anything? How do individuals judge whether any
behavior is good? Do individuals judge an act’s goodness in terms of their
own selfish desires, or do they take into account the concerns of others? And
by what standards would they judge in either case? This important set of
(philosophical) questions will be deferred until later in this chapter, when the
five broad types of ethical analysis will be discussed.

20
As was seen in chapter 3, human science is also somewhat less able to establish precise
causal relationships, due to the nature of individual thought processes. This point is often
emphasized by philosophers of human science (see Fox, 1997, 126). Nevertheless, “the
obstacles to a science of society are ordinary practical obstacles that can be and sometimes
are overcome” (Kincaid, 1996, 2).
5. Classifying Practice 177

Why do scientists do the sort of research that they do? Scientists are
guided by both the selfless desire to enhance human understanding and the
more selfish desire to have a successful career as a scientist. Valiela (2001,
268-72) suggests that scientists may pursue excitement, self-promotion, a
sense of duty, or profit. Ziman (2000, 158) refers to a principle of
disinterested scientific investigation such that scientists are guided by a
search for truth rather than personal gain; since scientists are human it is
doubtful that this principle is ever perfectly achieved. Hull (1989), though,
suggests that selfishness serves science well. Scientists’ inherent curiosity
and competitiveness drives them to make important discoveries. Their
concern with reputation guides them to identify errors in the works of others,
while limiting these in their own work. In science the penalties for outright
deception may be large enough to keep its incidence small (but growing?).
There does seem to be a strong predisposition toward honesty among
scientists. The fact that scientific works are made available to all causes
scientists to be wary of being fraudulent (Rota and Crants, 2000, 475). This
does not prevent scientists from exaggerating the strength of their argument,
or from ignoring contradictory evidence, but does guide them away from
outright lies. However, scientists driven by a desire to aid humanity, and
inclined to cooperate, may serve science better. If the institutions of science
do not flawlessly channel selfish behavior toward beneficial acts, there is
scope for encouragement of ethical behavior.
It is a simple fact of scientific life that the most brilliant of insights are of
no social benefit if they are not published or read or cited by others. They
may as well not have happened. Thus even a selfless scientist willing to
sacrifice career ambitions must pay heed to the standards by which research
is judged in their discipline or field. One of these standards has been
mentioned above: the tendency of each discipline to favor certain
phenomena, data, theories, and methods. No message emerges more clearly
from the classifications developed in this book than that each theory and
method has strengths and weaknesses, and thus that there are always
advantages to theoretical and methodological flexibility. Disciplinary
preferences must then place arbitrary limits on scientific investigation, and
this must slow the rate of advance in scientific understanding. The problem
is admittedly much smaller in those natural science disciplines that can lean
heavily on the experimental method, but exists even there. Against the
obvious cost of these arbitrary limitations, defenders of disciplines would
point to a huge benefit in terms of the maintenance of scientific standards.
By approving only statistical analysis and mathematical modelling,
economists thus ensure that they are capable of judging each other’s works.
To allow different methods would diminish their ability to judge. It is
doubtful that the benefits outweigh the costs (Szostak, 1999). But more
178 Chapter 5

importantly scientific communities can have the benefits without incurring

the costs. If they accept that particular questions are amenable to different
methods, then they need to identify subgroups of specialists that can evaluate
each other’s work. Different sub-communities of scholars can thus work
toward consensus on what their method says about a particular question.
Scholars without expertise in that method can draw on the sub-community’s
judgment – especially if all scholars and students were introduced to the
strengths and weaknesses of each method, as they could and should be (see
chapter 8). In terms of refereeing journal articles and books, multiple
methods pose no problem (at least once large bodies of users of each method
are identified). Even the use of multiple methods in a single article or book
should not pose huge difficulties once a large pool of adepts has been
identified, especially given the general practice of using two or more
referees. With respect to achieving a common understanding of what various
methods say about a particular question, this may be problematic if different
methods point in different directions (recall the discussion of triangulation in
chapter 3). But note that the choice faced by scientists here is between trying
to compare and contrast different types of evidence, or simply not knowing
what some methods have to say. Moreover, in cases where different
methods point in the same rough direction, such that scientists can
triangulate, they will receive the benefit with little cost.
Despite the claim that methodological (and theoretical) orthodoxy
protects standards, it is nevertheless true that the quality of scientific
contributions is hard to evaluate. Universities that wish to reward
researchers for their contributions to scientific debate often rely heavily on
quantitative evidence: how much was published, rather than how good it was
(this problem is perhaps less severe at the top universities, but quantitative
standards exist even there).21 Such an incentive structure encourages
researchers to produce lots of little insights rather than a few big insights.
This can only exacerbate the tendency toward narrow specialization. It must
also encourage “game-playing” by which scholars add minutiae to the
theoretical arguments of others in order to gain an easy publication. At
times, scholars may even play with a theory they dislike. The Pragmatist
Peirce heaped justifiable scorn on those who would prolong a scientific
dispute for the sake of argument alone; those who would prolong it for
personal gain deserve even greater scorn. Finally, since empirical work

21
Glassick et al. (1997, 19) discuss how legal developments and a tightening job market
encouraged universities to develop more formal evaluation procedures in the 1970s, at a
time when research was gaining prominence relative to teaching and service. While
resulting standards likely reduced the incidence of discrimination, they encouraged an
emphasis on that which could be quantified. This meant that teaching was either ignored
or treated only in terms of student questionnaires. As for service, “virtually no institution
has yet figured out how to quantify such work” (1997, 20).
5. Classifying Practice 179

generally takes longer than theoretical work, the scientific enterprise is

biased away from the ideal balance between theory and evidence.
When scientists do perform empirical work, they will be tempted to use
the method that guarantees the quickest results. Even at least one of the
developers of the Statistical Package for Social Science worries that many
scientists thereafter performed (often shoddy) statistical analyses because it
was easy, and thus turned away from other methods such as PO and
interviews (see Wellman, 1998). Disciplinary standards often come to
reflect a preference for that which suits the professional interests of members
of the discipline. Disciplines which continue to embrace more difficult or
time-consuming methods risk losing out in institutional competition for
resources. The scientific community needs first to recognize, and then
reward appropriately, both the differential costs and benefits associated with
different methods.22 Interdisciplinary scholarly communities in particular
must take care to avoid rigid theoretical or methodological preferences, but
to judge work only on the basis of how much it adds to scientific
understanding (Szostak, 2003b).
Scientists are concerned not just with speed. Especially early in their
careers, when tenure decisions often hinge on number of publications,
scientists are likely to avoid risky projects. Even later in their careers they
may be wary of research that might generate results that others would
disdain. The safe path is to make only the slightest adjustments to the
theories, methods, or data used by others. The cause of theoretical and
methodological flexibility, and scientific advance more generally, necessarily
suffers.
How do scientists choose research questions that lead to interesting
results? Moving scientists away from disciplinary and other biases will be
most advantageous if scientists are then guided to ask exciting and insightful
questions. It is all too common at present for researchers to start with
mastery of certain techniques, and look for some small problem amenable to
their use. This is a strategy that surely has a place but will not usually
produce major insights. Allowing scientists to pursue their curiosity is
necessary but hardly sufficient to ensure that interesting questions are
addressed. Being in touch with reality can be of great importance here – just
as confronting reality is critical for the evaluation of scientific arguments.

22
Economic theory would guide scholarly communities to aim for a world in which the
marginal benefit gained from the last minute expended by a scientific community on each
method is equalized (assuming, as seems likely, that the marginal benefit falls as the
method is used more and more). Economists do not apply this analysis to their own
discipline – in part because these benefits are hard to measure.
180 Chapter 5

Historians of science and technology increasingly appreciate that

technology influences science perhaps even moreso than science influences
technology (Mowery and Rosenberg, 1989, Valiela, 2001, 4-5). Scientists
are often motivated to ask questions about why a particular technology works
(or not). They also often rely on technology in performing their research.
Close links between scientists and those developing new technologies can
thus enhance the rate of innovation in both spheres. But science and
technology advance in quite different ways, with science depending on
public disclosure of discoveries, and technology driven by a desire to gain
private advantage. Linkages must be encouraged, then, without perverting
the incentive structure of either science or technology.
The role played by technology in encouraging advance in natural science
can be played by public policy in human science.23 Indeed, Bunge (1998)
refers to public policy as social technology. Dogan and Pahre (1990, 156)
discuss how public policy has and should inform social science; they note
that Keynesian economics reflected Keynes’ engagement with practical
policy issues. Scientists can usefully ask why particular policies work or not.
At present, though, the links between human scientists and public policy
makers are much weaker than is the case between natural science and
technology. Glassick et al. (1997) bemoan the attenuation of links between
the academy and wider society over the last century. They suggest that
applying knowledge should be one of the core duties of a professor, along
with creating, integrating and teaching. Where professors once considered
giving policy advice to governments or community groups a primary part of
their job, this is now only a marginal component of faculty evaluation. This
hurts society, in that decision-makers lack easy access to expert insights
(though note the criticisms of expert advice in chapter 2). It also limits
scientific advance, for the human science community lacks the same
intensity of feedback which technology provides for natural science.24
Scientists need not only an outlet for their research but inputs as well. The
need for research funding varies considerably by research question.
Problems arise when scientists receive funding from sources that have a

23
Lisl Klein and Eason (1991, xii) make a similar point about applying social science in
industry: “One way of looking at and summarizing this situation is to say that in the social
sciences the ‘Development’ part of the ‘Research and Development’ continuum has been
substantially neglected and there is a serious gap between the world of research and the
world of practice.” Theories are often found to be inappropriate or too general when
applied to a specific circumstance..
24
Scientists should, though, not view public policy as a playground for developing theory.
Nor should they disdain the insights of those public policy is supposed to serve. Herda
(1999, 12, 20), echoing the discussion of action research in chapter 4, urges a cooperative
enterprise in which scientists help citizens develop solutions to their problems.
5. Classifying Practice 181

vested interest in certain research outcomes (the problem arises most

seriously with respect to private funding but is possible with non-arm’s-
length public funding as well). All sorts of subtle pressures can be placed on
the scientist to get the right results (see Ziman, 2000, 162). Urging or even
requiring scientists to openly declare potential conflicts of interest can
alleviate, but hardly solve, this problem. Creating institutional mechanisms
that limit the opportunities for pressure is a more promising avenue. For
example, authorities could require that drug trials be performed by scientists
chosen by expert panels rather than by drug companies.
Hopefully by acquainting scientists with the advantages of
methodological and theoretical flexibility they can be encouraged to change
their behavior. Seale (1999) sounds a note of caution here; in the Preface to
his own work on methodology he emphasizes “the premise that
methodological writing is of limited use to practicing social researchers, who
are pursuing a craft occupation, in large part learned ‘on the job’ through
apprenticeship, experience, and trial and error, rather than by studying
general accounts of method.” Most scientists, that is, have little familiarity
or interest in broader issues of how science should be pursued, but instead
focus on how it is done in their area of specialization. They take existing
practices largely as given. This understandable but dangerous state of affairs
can be attacked in three ways. First, both undergraduate and graduate
education can be altered so that there is a greater familiarity with the various
classifications in this book. Second, the institutional reward structure can be
altered so that at least the penalties for using novel theories or methods are
less severe. Third, the ethics of scientific practice could be more clearly
communicated to scientists: the self-aware scientist should feel guilt if
playing games or obeying clearly misguided disciplinary standards.
Scientists, like any humans, have five ways of making decisions (chapter
3). Rational decision-making may be particularly important. It is thus
essential that the incentives facing scientists are appropriate, and that
scientists are encouraged to believe in socially beneficial goals. Historians
of science increasingly appreciate the role of intuition in scientific discovery.
This should underline the importance of scientists internalising appropriate
ethical attitudes such as honesty and openmindedness, and also seeking to
become self-aware of any subconscious biases they might possess. This
would benefit process or virtue-oriented decision-making as well. Both
institutions and attitudes will influence rule-based decision-making. Finally,
while the traditions of scientific research cannot be changed overnight, it is
certainly possible to encourage scientific communities to develop superior
traditions. Much research in the study of science is premised on the belief
that this can be done.
182 Chapter 5

2.4 Where?

Where does the scientist do science? The spatial location of the actual
research project was explored in the discussion of method in chapter 4. Here
the concern is with the institutional location of the scientist. Where does the
scientist exist institutionally? And how does this affect the quantity and
quality of scientific work? Scholars know very little about how differences
in institutions and cultures of science affect rates of scientific advance (Cole,
1992, 227). A necessary first step is to identify possible influences.
In the modern world, academic science is the basic form of science
(Ziman, 2000). This has the important benefit that science is largely
pursued, at least officially, in the public interest. Scientists intend for their
research to be freely available to all, and indeed to be revised by others.25
Universities receive much of their (especially research) funding from the
public, and seek to hire and reward professors who advance scientific
understanding. The results of university research are generally introduced
quickly into the public domain, rather than being kept secret for private
profit. Ziman is justifiably concerned that private funding of university
research is increasing in relative importance, and that this not only interferes
with the scientific pursuit of the problems of the greatest public interest but
that it also encourages secrecy. The very fact of this common complaint
serves as a symbol of the advantages of public research.
To be sure, it has already been seen that various subconscious biases
divert scientists from fully exploiting the potential for selflessly advancing
scientific understanding. And the tendency of most universities to evaluate
professors in terms of quantity of output rather than quality or impact serves
only to reinforce these potential biases. Other elements of the institutional
structure create additional problems. The reliance on individual professorial
curiosity has huge advantages, but there should be a greater place for large
coordinated research efforts: these could harness the efforts of those who
publish little while answering important questions that do not lend
themselves to individualized research.
Not surprisingly, philosophers debate the need for disciplines, and
whether disciplines tend to make good decisions (Fuller, 1993, 49-54).
Since scholars within a discipline emphasize what that discipline thinks
important, and tend to ignore what other disciplines think is important

25
In addition to being disinterested in rewards and desiring to share their discoveries with all,
Merton (1973) suggested two other scientific norms: universalism, whereby scientists do
not judge an idea by who voices it (he emphasized a distaste for ethnic discrimination), and
skepticism. Many scholars have suggested that Merton exaggerated the degree of
adherence to these values. The emerging field og value epistemology explores how
scholars should behave.
5. Classifying Practice 183

(Pahre, 1996, 206), interdisciplinary scholars can usefully encourage the

borrowing of theories and methods. One key insight of interdisciplinary
scholarship is that scholarly communities develop worldviews that embrace
(at least) ethical perspectives and ideas about how research should be
performed that both reflect and reinforce the community’s preferences with
respect to theory, method, phenomena, and data. Scholars, and especially
interdisciplinary scholars, need to interrogate each element of worldview.
The common division of universities into fairly autonomous departments not
only solidifies the disciplinary biases noted above but discourages the
interdisciplinary exchange of insight. As Glassick et al. (1997) stress, the
emphasis on research in the modern university exacerbates this problem:
scholars focused on teaching are less able to ignore the necessity of drawing
linkages across areas of specialization. In chapters 7 and 8 a couple of
initiatives will be outlined that could serve to counteract the tendency toward
over-specialization inherent in the present institutional structure of science.
Students of science must address not just what a particular researcher
does but whether this is noted by the community of scientists. Science
advances through a conversation among researchers. “Without an efficient
communication network, the process would not benefit from all the ideas
and information spread throughout the community” (Kantorovich, 1993,
220). If a particular work is ignored, its insights are lost. Fuller (1997, 68)
bemoans the fact that most published scientific work is never read. Hull
(1989) argues that in an evolutionary process like science it is inevitable that
only a minority of mutations (scientific arguments) will be selected for
further investigation; yet Hull appreciates that if some mutations are simply
ignored rather than evaluated, a charge of wastefulness could be made.
Scientific fields tend to be organized hierarchically, with a few stars
garnering most of the attention, an inner core of 1 or 2% of the community
producing a quarter of scientific papers, an outer core of some 20%
producing half, and a “transient” population of scholars producing at most a
handful of papers in their lifetime. Citations are even more skewed, with 1%
of papers garnering a third of citations (Collins, 1998). Collins has detailed
how networks of leading scholars determine both what are the interesting
questions to ask, and appropriate answers to these: scholars outside these
networks will have great difficulty being heard. The “stars” thus not only
have enormous power to exclude the work of certain scholars from
consideration, 26 but will often do so simply because these are not speaking to

26
Valiela (2001, 139) reports that those at more prestigious institutions are more likely to
reject articles written by those at less prestigious institutions. Cole (1992) notes, though,
that biases in judging grant proposals due to either gender or affiliation are slight. He
suggests that the bias operates in terms of ideas; elite scientists can determine which ideas
are approved. Haack (1998, 13) notes that in science, unlike in trials, the jurors have a
184 Chapter 5

the network’s conversation. Thus all scholars have an incentive to produce

works that fit the preferences of leaders in the field – both to advance their
careers and to have their insights considered. A less hierarchical and more
collegial scientific community would be less wasteful and more likely to
encourage innovative research: some transients might then prove particularly
innovative (see Dogan and Pahre, 1990, 23-4). Efforts to catalogue all
scholarly research (see chapter 7) would also advance these goals.
Who gains and who loses if scientific standards vary across types of
science, and across scientists? The thrust of this book is that scholarly
communities should be open to a wider range of theory, method, phenomena,
and data. While different research questions will naturally encourage
emphasis on some subset of the possibilities, the danger is that particular
communities will arbitrarily constrain their investigations. The beneficiaries
will be community members adept in the theories or methods emphasized.
Fuller (1997, 18, 101) suggests that consensus on certain issues in natural
science reflects the exclusion of viable alternatives (Szostak, 1999, makes a
similar point about economics). The losers will be the scientific enterprise
itself, and those scholars who question misguided disciplinary standards.
“Researchers who attempt to work outside of disciplinary and departmental
boundaries face internal as well as external alienation;” they may be shunned
by those they work with daily, and sidelined at conferences and in journals
(Palmer, 2001, 131). Palmer stresses the importance of interdisciplinary
centres and interdisciplinary organizations in giving interdisciplinary
scholars an institutional home.
In chapter 3, the ‘where’ question led to an appreciation of the need for
theories of varying degrees of generalizability. The analogous conclusion
here is that the scientific enterprise needs communities of scholars with
differing ranges of inquiry. Specialized research is necessary to carefully
gauge what particular theories and methods have to say about particular
causal links. Integrative research is necessary to draw connections across
specializations, and also to compare and contrast the use of theories,
methods, and data across specializations.
Academic science tends to be very competitive. The disinterested pursuit
of understanding can at times give way to spirited defences of opposing
positions. To be sure, scientific advance can result from such competition, as
scholars identify weaknesses in the arguments of others. But the danger is
that they only notice such weaknesses. And thus potential for synthesis is
ignored, and problems in empirical evaluation exaggerated. “If scientists
who disagreed worked together to resolve the dispute, I am inclined to
believe that more disputes could be empirically resolved” (Cole, 1992, 234).

vested interest in the outcome. Collins (1998) does not discuss in detail the factors that
determine elite status, but recognizes more than once that chance plays a role.
5. Classifying Practice 185

Whitley (2000) has proposed that scientific fields can be identified in

terms of how scholars compete for reputations. He notes that science is
unusual as a profession in that one’s prestige generally depends on how one
is viewed by others in the profession. In some fields a narrow elite controls
reputational recognition. In others, multiple elites exist; some of these may
even be outside the academy. In still others reputational recognition may be
quite diffuse. Whitley not only thus takes a laudable step toward identifying
different answers to ‘where’ but also argues that these can be understood in
terms of the attitudes of different fields toward theories and methods. In
fields that strongly favor one theory and method, the reputational reward for
a particular innovation is less uncertain than in fields open to multiple
theories and methods; as a result science is more routinized, and thus
hierarchical. If multiple theories and methods are allowed, personal contacts
among researchers, and diffuse discursive contributions, are more important.
Attitude toward theory and method does not entirely determine the degree of
interdependence between researchers; presumably the nature of the theory,
method, and phenomena are important here. Whitley develops a four by four
typology based on two different types of uncertainty and two different types
of interdependence. He argues that only seven of the sixteen resulting
classes are viable, and places various scientific fields in one of these
categories (noting that fields often change in type as they evolve). He
recognizes both that some fields may not fit cleanly into any one class, and
that his classification may not be exhaustive. The classifications developed
in this book are much more detailed, yet still manageable in scope. Arguably
they are also less subject to subjective bias: answers to the 5W questions are
often straightforward, whereas evaluation of how reputations or relationships
evolve in a field can be controversial. The classifications developed here
allow a more nuanced understanding of the interplay among theory, method,
and practice (see chapter 6). Still, scholars of science are well advised to
investigate further how both reputations and personal relationships vary
across types of science. Some of Whitley’s results – such as that
methodological flexibility is associated with less hierarchy – are consonant
with the analysis undertaken here. Other results are not: it has been argued
above that all scientific fields should strive for clarity of analysis, and that
theoretical and methodological flexibility can be pursued while maintaining
scientific standards.
Scientists and universities depend on government finance and policies in
a variety of ways. Scholars of science have naturally worried that
governments, directly or subtly, intentionally or not, influence the direction
and even results of scientific research. Research funding is the major
concern. Especially when large sums are involved, as with the space station
or linear accelerators, funding “raises troubling questions of whether
186 Chapter 5

political considerations dominate scientific value, and also of whether big

science is conducive to the exercise of scientific curiosity and especially
revolutionary scientific insights” (Fuller, 1997, 7-9). Governments play
other roles: notably they often subsidize or regulate scientific publications,
and now the Internet; questions can be raised of whether they thus encourage
the best means of scientific communication. In some fields governments
have an important influence on scholarly reputations (Whitley, 2000).
How is scientific insight communicated to the wider public?27 The need
for social scientists to be more active in policy making was mentioned above.
Natural science research is also of interest to the public, and often raises
troubling ethical or social questions. Scientists are understandably frustrated
by journalists who seek controversy rather than insight. Yet in failing to
admit that science is an evolutionary process, where today’s results may be
overturned by future research, scientists encourage scepticism of all
scientific results. A more nuanced understanding of the scientific process on
the part of the public would allow and encourage the public to evaluate
scientific results on their merits (Fuller, 1997, 6). Policy-makers who must
draw upon scientific advice should be especially adept at this. Scientists also
often fail to appreciate the legitimate concerns of the public. Often, these are
rooted in a different ethical approach than the consequentialism that
characterizes science. Citizens guided by alternative approaches to ethical
decision-making (see below), such as intuition or tradition, may, for
example, be much less likely to countenance environmental risks. An
appreciation of the five different ethical perspectives, plus the humility that
should come with an appreciation of the complexity of the scientific
undertaking as a whole, should encourage better communication from
scientists to the public, and the reverse.28 This should in turn encourage
greater public support for the scientific enterprise.

2.5 When?

How is a scientist’s work situated historically? If science works at all,

there should be some accretion to scientific understanding through time. Yet
history shows that theories that have seemed correct for a long time have

27
Stocklmayer et al. (2001) suggest that there are five reasons to communicate science to the
public: the economic (because science is intellectual property, whether public or private),
the utilitarian (because science is useful to know), the democratic (because citizens must
make decisions about, say, global warming), the cultural (because there is something
beautiful about science at its best), and the social (because if the public trusts and embraces
science this will enhance social cohesion).
28
Science has only become exclusive in the modern era: only scientists make contributions to
scientific journals. Yet scientists, unlike other public authorities, are not required to justify
their special status (Fuller, 1997, 36).
5. Classifying Practice 187

later been found to be misguided in at least some respects. This creates an

opening for sceptics to suggest that scientific progress is but a mirage:
advances that are cherished today will be overturned tomorrow. The ability
to harness nuclear energy or alleviate certain physical and mental illnesses
would suggest otherwise. While scientific understanding in any area is likely
incorrect in some respects, it would be a mistake to leap to the opposing
conclusion that it is necessarily incorrect in all respects.
If it is accepted that science advances, then scientists should be expected
to purposely build upon the work of predecessors. If, though, it is
recognized that received theory is likely misguided in at least some
particulars, then scientists should not accept blindly the authority of past
scholarship. This is a tricky balancing act, of course. It is, though, the same
balancing act that an openminded person should always pursue: accepting
certain beliefs that seem accurate as a necessity for daily life but being
willing to question any of these in the light of new evidence (see chapter 6;
Szostak, 2003). Histories of science could be more helpful here: they are
often celebrations of progress that imagine that scientists of the past were
focussed upon the same questions that motivate scientists today (Cussins,
2000, 343); they thus fail to warn practicing scientists that science
occasionally pursues the wrong path and needs to be redirected.
Thomas Kuhn (1962) famously suggested that the scientific community
does a poor job of achieving balance. Most disciplines develop a paradigm,
a loosely (and sometimes contradictorily) defined term that seems to
embrace a core group of theories, methods, and phenomena. Research
within the discipline is constrained to operate within the paradigm. This
allows research to proceed on various extensions of the basic paradigm,
while the core ideas remain unchallenged. Research results suggesting that
the core ideas are misguided will not cause these to be jettisoned
immediately. If enough contradictory evidence is compiled and an
alternative paradigm is in view, a scientific revolution may occur which
overturns the previous paradigm. Kuhn’s work demolished a vision of
science in which this advanced steadily by adhering to strict scientific
principles. Kuhn not only established that science needs occasional
revolutions, but that the link between paradigm and evidence is tenuous. 29 In
particular, the timing of revolutions is subject to a variety of forces (and
some scholars thus worry that revolutions may no longer be possible,

29
Some critics of science have treated Kuhnian paradigms as simply social constructs. But
Gower (1997, 245-6) notes that paradigms are by their nature an attempt to grapple with
reality, and thus necessarily influenced by reality. He notes that Kuhn himself was quite
hostile to the relativist viewpoint of such critics. He certainly did not see paradigms as an
entirely arbitrary construction.
188 Chapter 5

because of the entrenched power of a scientific elite, or because present

scientific ideas serve the status quo). Yet while many have seized upon
Kuhn’s ideas as a rationale for extreme scepticism, Kuhn himself does not
doubt that science advances, albeit unevenly.
Most of the time, new paradigms absorb important elements of old
paradigms. The theory of relativity did not erase Newtonian mechanics, but
left this as a reasonable approximation to reality in a wide range of cases.
Sceptics might argue, though, that scholars rewrite history after a paradigm
shift – emphasizing previous work that presaged the new paradigm – so that
science seems more progressive than it actually is (Cole, 1992, 9). This
would be especially likely if Kuhn’s idea of paradigm incommensurability
were correct. If paradigms define terms differently, it may be impossible to
communicate across paradigm boundaries. If, though, differences in
definitions are far from absolute, then some degree of communication should
be possible (Szostak, 2002a). Still, if paradigm shifts involve changes in
favored methods as well as theory, there may be no way to accurately
determine their relative strengths. Kincaid (1996, 32) stresses that paradigm
shifts occur gradually, and thus comparisons can be made along the way.
This, notably, is yet another argument for relying on multiple methods.
Since science necessarily builds on its past it can be conceived of as an
evolutionary process. Within generally accepted theories of genetic
evolution mutations are assumed to be random. As was seen in chapter 3,
evolutionary processes driven intentionally are also possible in diverse areas
such as culture and art. In such cases, scientists can hope to explain why at
least some mutations occurred, as well as why these were selected.
Simonton (1988), Giere (1988, 161-5) and Kantorovich (1993), following
Donald Campbell, suggest that, despite intentionality, scientific mutations
are almost random. In going beyond what is known, researchers must
necessarily search blindly. The community of researchers will thus generate
a wide range of ideas; these will be selected at two levels: the individual
researcher will develop only those that seem plausible, and the research
community will choose what seem the best ideas from those already selected
at the individual level (Simonton, 1988, Kantorovich, 1993). Moreover, in
attempting to answer one question, scientists often serendipitously stumble
upon answers to others, or discover surprising new questions to ask
(Kantorovich, 155, 159). [Note that their ability to recognize the importance
of such insights will depend on their familiarity with other realms of inquiry;
the classifications developed in this book could potentially acquaint
researchers with the full range of scientific inquiry, and thus aid the
interdisciplinary process of transferring insight across scholarly
communities.] Still Kantorovich (173, 186-9) does not argue that scientific
mutations are entirely random; he – unlike some others who prefer an
5. Classifying Practice 189

evolutionary approach – appreciates that the history of science is replete with

many examples of simultaneous discovery. These reflect the fact that the
time was ripe for generating a particular discovery (likely because the ideas
to be combined had been put in place) and/or for selecting it. Students of
science should thus be open to the possibility that certain discoveries reflect
serendipity, but others do not, and that there is a continuum between these
extremes.
For a scientific community to accept a conclusion, there must be a
confluence of a plausible theory and a set of supporting evidence. Often
there is a mutual interaction between theory and evidence, with evidence
shaping theory, and theory guiding the search for evidence. For consensus to
be achieved, the community must have enough interest in the question for
members to seriously engage both the theory and evidence.
While he speaks of researchers searching blindly on the frontier,
Kantorovich does appreciate that they will generally utilize certain theories
and especially methods in their searches (recall the heuristics of Langley et
al., 1987). He does not seem to appreciate how this simple fact will decrease
the randomness of scientific discovery. It has been seen, though, that each
method and theory type is best suited to asking and answering certain
questions. If scientists, even those striving to develop a new paradigm, do
not look beyond the methods and types of theory common in their field, then
scientific discovery will be truncated; mutations will occur only over the
permitted range. Scientists may also limit the range of causal links they
consider; this too will narrow the range of possible insights.
Ziman (2000) suggests that the Kuhnian picture is overdrawn, and that
there are always some scientists pursuing revolution, while others work
within the paradigm. However, the institutional rewards for work within the
paradigm, and penalties for deviation (including, it should be noted,
disincentives toward engaging in interdisciplinary research), may be so
strong as to seriously diminish the amount of revolutionary work. Scholarly
communities may thus stay with paradigms long after evidence points in
other directions, or even fail to do the sort of research that may point toward
an alternative paradigm. In chapters 4 and 6 the possibility that certain
methods may be biased in favour of certain theories is discussed: if so
disciplinary strictures against alternative methods will protect paradigms
against contradictory evidence. Also in chapter 6 a healthier attitude toward
paradigms will be advocated.
This book has suggested in various places that scholarly communities
should be open to a wider range of theory and method. Paradigms tend to
insist on a very limited set of theories and methods, and limit attention to
those phenomena and links that these seem to be able to explain to a
satisfactory degree. If, as is often argued, a paradigm can only be replaced
190 Chapter 5

by another that seems to have the potential for better explanation, then
opportunities for better explanation through recourse to a wider set of theory
will be eschewed even during times of revolution. In other words, scholarly
advance in at least some fields depends on overcoming the ideal of the
paradigm, which pretends that one or a handful of related and highly
generalizable theories can explain all that needs to be explained.
Previous sections have respectively explored, albeit briefly, the analogies
that could be drawn to the answers to the 5W questions in chapter 3: the
three types of intentional agent, attitudes and actions, the five types of
decision-making, and idiographic versus nomothetic outlooks. What can be
said here about time-path? The behavior of particular scientists will be
conditioned by whether their field is static, moving smoothly between
equilibria (making new discoveries, and incorporating these smoothly into
the corpus of theory), moving in a particular direction, or behaving in an
unpredictable manner. These four scientific time paths create different
opportunities, incentives, and constraints. The analysis of Whitley (2000),
discussed above, touches on some of these issues.
In asking ‘when?’ it is important to look not only at the state of science
itself but at the broader society. Scientists are human, after all, and can
hardly be immune to the cultural influences (and time path of societal
evolution) that affect others. Feuer (1995) describes how oscillations
between optimism and pessimism in the wider society have over the last
couple of centuries influenced the way scientists viewed the world. Krieger
(1997, ix) describes physics as a subculture whose theories reflect the
autonomy, alienation, and exchange that characterize the wider society.
Szostak (1999) discussed various cultural influences on the evolution of
economic theory and method in the twentieth century: economists developed
theories (and supportive methods) that provided comforting answers to a
range of cultural concerns. This does not mean that all of that theorizing and
empirical work was misguided; it does suggest that the scientific community
may have embraced certain results and procedures for reasons beyond their
scientific merits. It is thus a useful exercise for scientists to ask what cultural
biases might influence their evaluation of scientific research, and how the
theories and methods of their field may reflect past cultural pressures. In
addition to cultural influences, it should be noted that the institutions
discussed in the previous section also evolve over time (Stump, 1996, 450).
Cole (1992, xii, 30) suggests that societal factors are important on the
frontiers of science, but that core theories tend to reflect reality. He argues, in
effect, that for a theory to be accepted by an entire scientific community it
5. Classifying Practice 191

must be buttressed by very convincing evidence.30 Work on the frontier,

alternatively, cannot be judged as objectively (1992, 158-9); evidence, by
definition, is not overwhelming, and thus various biases can influence what
groups of scientists believe.31 When objectivity was discussed above, it was
suggested that in the long run appeals to real world evidence could overcome
various biases. But it was argued that this result could not be assumed, and
scholars should not accept Cole’s argument without question. Nevertheless
biases are likely more important on the frontier. Students of science should
keep this in mind when studying particular bodies of research.
The timing of particular research in a person’s life should also be briefly
considered. Polkinghorne (1996, 35-7) describes how even the greatest of
scientists accomplished little in later life. While scientific understanding
here is imperfect, it does seem that most scientists are most innovative in
their twenties or thirties, but most able to see the big picture in their fifties.
Institutional constraints may be at work here: it could be that scientists doing
integrative big picture research are only heeded if they have first gained a
reputation from specialized research. But if, as seems likely, the strengths
and weaknesses of researchers evolve with age, the institutional structure of
science should reflect this. One concern would be that in some fields
individuals are still in graduate school mastering the work of others while at
their most creative age. Another concern would be that powerful older
scientists may limit the degree of innovation by younger colleagues. Finally,
it should be noted that the relative importance of different types of mental
activity may vary across fields, and over time within fields; if so, career
progress expectations should vary as well.

2.6 How?

“How do scientists do research?” Much of this question has been dealt

with in chapters 3 and 4. The concern of this chapter would be with “How
might scientists produce research that leads them to misguided conclusions?”
The answer to such a question would naturally involve the limitations in both
individuals and communities of scholars, and in institutions. All of these
matters have been dealt with above.

30
Collins and Pinch (1993, 143) make a similar argument about “sciences”: the same degree
of certainty should not be expected in a new science as has (seemingly) been achieved in
physics.
31
In some cases, frontier work is quickly accepted by the scientific community. Cole (231)
suggests that this happens when the evidence is very strong or when the contribution both
answers and creates important questions within the research trajectory. Alternatively,
“Most new scientific contributions never have a chance to enter the core because they fail
to meet the latter criterion [of being important]” (1992, 15).
192 Chapter 5

At the start of this chapter, a brief list of some of the more common
questions asked of scientific practice was provided. This list in turn reflected
a reading of a variety of works on the practice of science. While there is no
pretence that the list was exhaustive, it is still worthy of note that each of
these questions was addressed as the 5W questions were answered. In some
cases it was possible to go some way toward answering these; in others it
was only possible to clarify questions for empirical investigation. These 5W
questions do seem, then, to once again provide a categorization that is both
exhaustive and convenient.

3. UTILIZING THIS CLASSIFICATION

In chapter 3, it was argued that all scientific theories should provide clear
answers to each of the 5W questions. A somewhat less strong argument can
be made here: the enterprises of history, philosophy, and social study of
science must grapple with each of the 5W questions, but not every research
project within these enterprises must do so. Even at the level of individual
research projects, researchers should at least be aware of the other possible
critiques, lest they ignore evidence of these at work.
Any comprehensive attempt to comprehend any particular innovative
process, whether individual or collective, should ask how the perceptual
limitations of all humans affected the process, how the personal
idiosyncrasies of the researcher(s) involved influenced research agendas, and
what the social influences on researchers were. How strong was the
evidence in favor of the theory(ies) that emerged from the innovative
process? (That is, how well did reality overwhelm these various biases?)
What rhetorical devices did researchers utilize to convince the broader
community? What incentives did researchers face (specialization, emphasis
on quantity of publications, profit, and so on), and how did these affect their
behavior? What institutional constraints did they operate under, and how
might these have constrained their inquiries? Did disciplinary boundaries
limit the investigation? Did the research enterprise benefit from interaction
with those working in technology or public policy? How did this research fit
within the broader history of science (what assumptions and preferences and
questions did the researchers start from)? In particular what paradigms did
scientists operate within or supersede? Did they accept the theories and
methods favored by others pursuing similar questions, or not? How did the
training of researchers (mis-) prepare them? What role did serendipity play?
And finally what type of science did the researchers do – what theories,
methods, and causal links did they engage – and how might a broader
perspective have changed the results?
5. Classifying Practice 193

The last question, and many of the others, will be difficult to answer
unless one can refer to other actual research agendas. That is, it is always
hard to know what has been lost due to any bias or constraint unless one can
point to other groups of scientists who operated in a different manner. The
classifications of earlier chapters at least allow room for some educated
speculation. If a research enterprise eschewed certain sorts of theories or
methods, scholars can know what the strengths of these are and can guess as
to the insights they might have provided.32 In this way, research in the study
of science might encourage research in the neglected directions. This in turn
should both advance the cause of scientific discovery while also testing the
accuracy of the original speculation.
The 5W questions were somewhat less mutually exclusive here than was
the case in chapter 3. Overlap is clearest with respect to the ‘Why?’ and
‘Where?’ questions, for the incentives faced by scholars largely reflect the
institutional environment in which they operate. As noted in chapter 1,
overlapping classes are a common problem in classification. Despite some
overlap, it has nevertheless been possible to identify distinct subsidiary
questions under each heading.
The 5W questions have also not so clearly guided an exhaustive
classification of possible critiques as they did of possible types of theory. Yet
the classifications of earlier chapters are quite useful in this regard. With
respect to ‘who’ recourse to the three types of intentional agency, and then
the use of Table 2.1 to identify all possible types of both individual bias and
societal influence, promises a fairly exhaustive classification. With respect
to ‘why’ the five types of decision-making are a valuable guide. With
respect to ‘when’ the investigation of different time paths may prove a good
way of evaluating the historical circumstances of any field of inquiry at a
point in time. Finally, with respect to ‘where’ considerations of
generalizability lead to an appreciation of the necessity of both specialized
and interdisciplinary research.

32
Szostak (2005) argues that economists, in studying the Great Depression, have relied on
only one type of theory. Ten turning points in the historiography of the Great Depression
are described, and it is shown that at each point alternative theories provided potential
solutions to puzzles thrown up by the dominant macroeconomic approach. The emphasis
by economists on mathematical models and regressions has encouraged them to ignore the
clear evidence that the timepath of technological innovation was highly unusual in the
interwar period. Technological explanations of the Depression can be viewed as
complementary to macroeconomic explanations, but require a much greater openness to
disequilibrium processes and one-off events (and interdisciplinary causal linkages).
194 Chapter 5

4. THE FIVE TYPES OF ETHICAL ANALYSIS

One of the moments of epiphany that the author enjoyed while writing
this book occurred when discussing decision-making processes in chapter 3.
In striving to outline the various ways in which an individual might make
any decision, it was realized that there were precisely five types of decision-
making process. These five types of decision-making process were, in turn,
identical to the five types of ethical analysis identified in Szostak (2002b,
2004); these five types of decision-making could thus be justified both
deductively and as inductively. At first surprising, this result now seems
obvious: in making ethical evaluations individuals are making decisions, and
thus it follows that each broad type of ethical analysis accords with a
particular decision-making process. This one-to-one correspondence has a
particular appeal, for the list of decision-making processes had been derived
in a fairly deductive manner (at least until it became clear that the two lists
were similar), but the list of types of ethical analysis inductively: by reading
widely in philosophy and attempting to classify. As noted in chapter 3 there
has been scholarly discussion of each type of decision-making, but of no
other type of decision-making. With respect to ethical analysis, philosophers
tend to emphasize the three “formal” types of ethical analysis:
consequentialism (where an act’s goodness is judged in terms of whether its
consequences are good), deontology (interpreted here as implying a reliance
on rules, and thus embracing, for example, the Golden Rule, Kantian
categorical imperative, and arguments from rights), and virtues (where
individuals are urged to live in accord with one or more virtues, such as
honesty; see Baron, Pettit, and Slote, 1997). In addition, individuals should
be aware of the frequent use of both intuition and tradition in everyday lives.
Examples of philosophers lauding these as bases for ethical behavior can be
found across all philosophical traditions. It proved straightforward to place
various philosophical arguments within one of the five broad types, with the
singular exception of the eclectic philosophy of existentialism, which urges
individuals to choose their own path (that is, “decision-making process”;
Szostak, 2004). There is thus good and varied evidence to support the belief
that these five types of decision-making process are mutually exclusive and
exhaustive. Notably, this discussion provides a powerful example of one
role for interdisciplinary analysis: only when one scholar casts their gaze
upon quite different questions can such convergence in understanding occur.
The fact that there are precisely five ways to make decisions raises the
question of whether these flow from the 5W questions. The questions “Why
is an act ethical?”, “When is an act ethical?”, and “What is an ethical act?”
are too similar to distinguish among types of ethical analysis, though one
could imagine that the best answer to “Where is an act ethical?” involves an
5. Classifying Practice 195

appeal to tradition, and to “Who is ethical?” [but note the different

formulation of the question] involves virtues. Rather than seeking some
artificial correspondence between the 5W questions and the five types of
ethical analysis, it is better to evaluate the five approaches to ethical analysis
– which can be thought of as ethical theories, a phrasing eschewed here to
avoid confusion – in terms of the typology of theory from chapter 3. While
ethical theory differs from scientific theory in an important respect – it
speaks to how the world should work rather than how it does work – the 5W
typology nevertheless gives valuable insight into the differences among the
five types of ethical analysis. It also, as with scientific theory, provides an
important means by which to recognize and classify distinctions within these
five approaches. In terms of agency, deontology (especially rights) and
intuition focus at the individual level, while tradition emphasizes a group
orientation. Consequentialism can apply to any type of intentional agent.
Virtues are appreciated at the level of individuals, though modern
communitarian virtue theory stresses that the choice of which virtues to
pursue occurs at the group level. In terms of action, consequentialism and
deontology stress actions while virtue theory stresses attitudes. Intuition
seems to stress attitudes, but may guide individuals most strongly when they
contemplate acts. Tradition applies to both acts and attitudes. With respect
to generalizability, consequentialism and deontology stress universal
guidelines, while tradition emphasizes society-specific rules. Some virtue
theorists argue for a universal set of virtues, while others believe that virtues
should be determined by groups. If one believes that human intuition is
grounded in genetics or “the gift of god” one will expect universal intuitive
behavior; if it is thought that intuition is grounded in experience one will not.
As for “when?”, consequentialism (and some types of deontology) accepts
that the rightness of an act depends on circumstances, while the other
approaches tend to argue that an act judged ethical in one circumstance will
always be ethical (though traditionalists and communitarians may talk about
how traditions evolve). The time path through which an act affects the wider
world is of direct importance to consequentialists, but at best of indirect
importance to others. The answer to the “Why?” question is the easiest of
all, for each type of ethical analysis is associated with a different type of
decision-making process.
It was shown in chapter 3 that consequentialism, virtues, deontology,
traditions, and intuition are the five ways in which any person might make
any decision. In terms of ethics, these five processes describe the five ways
in which any person might evaluate any act or outcome. Note that there is
no logical criterion by which it can be claimed that one approach to
evaluation is superior to others. Reasoned arguments can be put forward in
favor of each. Notably, when philosophers strive to justify the supremacy of
196 Chapter 5

one type of ethical analysis, they often have recourse to others. For example,
rights theorists refer variously to intuition, tradition, consequences, and
virtues (Almond, 1998). The five types of ethical analysis can be seen then
as equally valid and complementary. Recognition of this fact should guide
individuals to have greater understanding of those who reach different ethical
conclusions from their own. They can appreciate that ethical dilemmas
reflect disagreements among different ways of deciding what is right.33
While diversity can thus be embraced, individuals need not also embrace an
“anything goes” attitude. The selfish can be distinguished from those who
can provide reasoned justifications for their acts in terms of at least one type
of ethical analysis.
Moreover, individuals can strive where possible to act in ways that can be
justified in terms of all five types of ethical analysis. Such opportunities are
more common than is generally realized. Szostak (2004) evaluated some
150 ethical statements, each associated with a particular phenomenon from
Table 2.1. In roughly half of these cases all five types of ethical analysis
pointed in the same direction. To be sure, no ethical argument is
impregnable, and thus reasonable people could reasonably disagree with one
or more of these statements. This group of statements could nevertheless be
seen as an “ethical core” that people should not deviate from without reason,
though some may reasonably do precisely that from time to time.
Just as the classifications in previous chapters facilitate the exchange of
understanding across scientific fields, the identification of five independent,
complementary, and universally valid (since examples of each can be found
across diverse philosophical traditions) types of ethical analysis facilitates
the performance of a comprehensive and coherent ethical survey (Szostak,
2004). Since the five types of ethical analysis have applicability across all
phenomena, then they must necessarily apply to science in particular.
Scientists were urged above to emphasize the pursuit of understanding rather
than career goals.34 It can now be argued that such an attitude accords with
all five types of ethical analysis. In terms of consequences, the most obvious
counter-argument would be that scientific advance need not have beneficial
societal consequences; some might wish that nuclear physics or certain sorts

33
Polkinghorne (1996, 124) appreciates that scientists cannot disregard ethical analysis. He
bemoans the fact that “the absence in contemporary society of an agreed moral basis for
decisions has reduced ethical debate to the strident assertion of individual opinions.” Just
as society needs to be able to judge conflicting scientific arguments, it needs a means of
evaluating conflicting ethical arguments.
34
Hammersley and Gomm (2000, 164-6) urge researchers to pursue the production of
knowledge above all other goals. They stress that exaggeration has no place in scientific
inquiry. Max Weber had argued that scientists must pursue “truth” wherever it leads, even
if results are unpalatable (Hammersley, 2000, 33).
5. Classifying Practice 197

of biotechnology had never been imagined. But the fact that society devotes
billions of dollars to science is evidence of a profound belief that science
does more good than harm. And a scientific community dedicated to the
public good might tip the balance even moreso in that direction. Eschewing
personal gain to pursue understanding for society’s sake is almost
transparently virtuous. As well, such behavior easily accords with such
ethical rules as the Golden Rule or Kant’s categorical imperative: if
individuals would want other scientists to behave in a socially responsible
fashion they should do so themselves. And most/all traditions celebrate
scientists who generate useful ideas, rather than those who successfully
achieve promotion (at least for the most part).
Finally, selfish careerism should make a scientist feel (intuitively) guilty.
Few scientists are likely self-consciously selfish. Many, though, are able to
rationalize careerist behavior, and thus minimize their guilt. A host of
excuses are possible: “Science is impossible”, “Everybody does it”, “I’ll
change once I get tenure/promoted/retire”, “I am obeying the incentives of
my field and trust others to ensure their consonance with the social good”, “It
does not matter what any one scientist does”. Hopefully this book will have
exposed at least some of these excuses as false. Science is indeed possible.
But it is all too easy to perform bad science. Each scientist has a
responsibility of self-reflection, in order to ensure that they are not being
guided by baser motives. Science could be more successful than it is if there
was greater recognition of the strengths and weaknesses of different types of
science. Scientists who act upon this potential not only advance science in
their own right, but set an example for others. In sum, this book serves to
align the intuition of scientists with the other four ethical guidelines, and thus
urges them toward a more selfless approach to scientific investigation.
It was noted at the outset of this chapter that its subject matter has often
been neglected by scholars of science. This is true even today of ethics:
while institutional changes are (laudably) forcing scientists to reflect on the
ethics of using intentional agents in research, and of accepting funding from
external bodies that expect a particular outcome, scientists rarely engage
more general ethical questions concerning scientific practice, and even
philosophers tend to skirt the issue. Philosophy of science engages
epistemological questions (what can we know?), not ethical questions (what
should we do?). Yet the answers to the sorts of question posed here are
straightforward, and thus a greater attention by scientists to ethics would
yield great dividends.
Human scientists have even shied away from ethical evaluation in their
research. That is, they not only neglect ethical analysis of themselves but of
others. Wolfe (1999) outlines – and debunks – several reasons for this state
of affairs. Scientists were hostile to tradition, feeling that it hindered
198 Chapter 5

personal liberty. But a recognition that tradition is a valuable source of

ethical guidance – note that evolutionary arguments can be made that
societies will tend to select beneficial traditions over time (though power
relations, for example, may affect this outcome) – coupled with a recognition
that there is no one right way to make ethical decisions, should overcome
this bias. Scientists revelled in Hume’s fact/value distinction: what should be
cannot be deduced from what is. But, as many interpretivists have noted,
people have ethical beliefs which affect their other thoughts and acts;
scientists cannot hope to fully understand people without appreciating their
ethics. Scientists were so devoted to the ideal of appreciating (especially
cultural) diversity that they hesitated to evaluate ethical behavior. But
scientists cannot strive to help minorities unless they have some sense of
ethics. [Recall also that recognition of five valid types of ethical analysis
allows individuals to avoid the twin dangers of ethical bigotry and ethical
anomie.] Finally, scientists were suspicious of religion, on which many
ethical arguments are grounded. But one can respect both religion and
scientific inquiry, and certainly hope to comprehend the influence of religion
on people’s lives. Chapters 3 and 4 urged scientists to appreciate theoretical
and methodological flexibility. Here, an engagement with ethical issues
must first be encouraged. Then openness to each of the five types of ethical
analysis can likewise be encouraged in scientific research. Scientists can
thus aspire to ask of an individual or group whether their ethical beliefs are
grounded in only one of these, or are more broadly based. Scientists can also
aspire to assess the consistency of ethical beliefs. And scientists can
generally exhibit respect for diversity, while maintaining an ability to
criticize the selfish and the self-deluding.
Scientists who consciously avoid ethical reflection may well fail to
appreciate subconscious ethical biases in their work. Do they seek results
that will justify cherished ethical arguments? Root (1993) argues that ethical
biases are almost inevitable in human science. Ethical biases can also assert
themselves in various branches of natural science. Scientists were urged
above to question their motivations in terms of each of the five types of
ethical analysis. They can also be urged to ask questions about the content of
their research. Are there hidden ethical presumptions in their work? For
example, Root discusses uses of the word “development” at both individual
and societal levels, and notes that there are often hidden ethical presumptions
regarding the desirable outcome of development. Will the research justify
actions or policies that are consonant with only some ethical perspectives? If
so, can the scientist’s ethical preferences influence results? Root
recommends involving subjects in research design in order to overcome such
biases; the analysis here suggests instead the importance of introspection.
Chapter 6
DRAWING CONNECTIONS ACROSS THESE
CLASSIFICATIONS

1. THEORY AND METHOD

There are strong similarities in the answers to the 5W questions with

respect to theory and method. This might lead scholars to expect that certain
theories should be tested with certain methods. That is, if a theory posits a
certain form of agency, a method that is particularly good with that form of
agency should be used. There is an important kernel of truth in this idea. In
the natural sciences, theories involving non-intentional agency are most often
tested with experiments, a method with few drawbacks when dealing with
non-intentional agency. Yet even here it was seen that experiments are
imperfect, and that recourse to other methods is advisable. And this leads to
an important general insight: given that no method is perfect, scientists
should not limit any form of inquiry (including the test of any theory) to just
one method.
It might still be argued that only those methods believed to be of
particular utility for the theory in question should be used. Again, this is a
valuable insight. Surely, natural scientists focused on non-intentional
phenomena can eschew methods such as interviews. But note that there are
five key questions to be asked of any theory. For most theories, several – in
human science, often all – methods will be well-suited to at least one of the
key characteristics of that theory. And even if a particular method does seem
admirably suited to the exploration of a particular theory focused upon a
particular causal link, important insights can generally be gained by
investigating related links. In human science, given that every phenomenon
is causally related to every other, there must thus be some scope for the
200 Chapter 6

application of every method to any human science question. As well, recall

from chapter 2 that insight may be gained into phenomena/links by
examining different types of data: this provides a further rationale for
methodological diversity.
Even if a particular method were to seem admirably suited to testing all
aspects of a particular theory, scientists should still be wary of excluding the
use of other methods. Theories, after all, can only effectively be tested
against alternatives. Methods well-suited to certain types of agency or
process may be poorly suited to detecting evidence of alternatives. And thus
by limiting the methods employed, scientists inadvertently limit the
possibility of recognizing alternative theories. Economists favor the use of
mathematical modelling and statistical analysis, methods that (at least with
the formulations commonly used in the discipline) are well suited to
individual agency, analysis of actions, rational decision-making, and
generalizable equilibrium processes. These, notably, are each characteristics
of the rational choice theory that dominates the discipline. While negative
results with this method might encourage a search for alternative theories (in
practice this seems rarely to happen), alternative theories involving group
agency, attitude formation, irrationality, unique historical transformations, or
disequilibrium processes can scarcely be considered if the discipline clings to
this one method. This is especially the case given the limited inductive
potential of this method. By emphasizing one method, then, the discipline
necessarily biases research results toward theory elements the method is
particularly good at elucidating. The discipline can hardly avoid denigrating
the importance of one-off events, or disequilibrium processes, or attitudes, in
such a situation. Since all disciplines have methodological preferences, the
problem is ubiquitous.1 Hall (2003) thus worries that the field of comparative
politics has come to embrace theories of path dependence and strategic
interaction that are at odds with assumptions commonly made in regression
analysis: he thus advocates increased use of comparative case study research.
Again, though, the problem is likely less severe in those natural sciences
where experiments are well suited to the questions at hand.
Interdisciplinarians can provide a powerful antidote to the tendency of
disciplines to rely on methods that cause them to “see” only their favored
theory. The discussion in previous chapters of the tension between
nomothetic and idiographic approaches provides one example of this general

1
Freidheim (1982, 5) notes how the use of quantitative methods guides researchers to
develop and apply theories that focus upon variables that can be quantified. Elsewhere,
she finds that paradigms (see below), or clusters of theory/method that guide large groups
of researchers, are most easily identified by method: this suggests that the methodological
preference of the group guides their theoretical choices. Freidheim hopes that these
paradigms will evolve to become focussed instead on certain sets of phenomena.
Emphasis on one method is ill-suited to such a transformation.
6. Drawing Connections Across these Classifications 201

problem. Case study methods are well suited to identifying particularities,

but can also be used to point toward generalizations. Statistical analyses are
best suited to identifying generalities, though they can also identify outliers
worthy of further exploration. Scientists should be seeking to identify the
limits of any generalization: when does it apply or not apply, and what other
phenomena affect the strength of the relationship? Exclusive reliance on one
sort of method or another will lead to either over- or under-estimates of the
extent of generalizability.
Another key example is provided by the discussion in chapter 3 of the
interpretivist emphasis on attitudes and the positivist emphasis on actions.
Positivists favor methods that elicit supposedly objective data. Interpretivists
argue that one can only understand the meaning that agents attach to
situations by putting oneself in their place: they thus favor methods that
focus on subjective attitude formation. Happily, there has been some
recognition by both interpretivists and positivists of at least some of the
disadvantages of methodological inflexibility. Some from both camps have
appreciated that their methods are imperfect, and that the methods favored by
the other camp can compensate. For example, the potential for subjective
bias in interpretivist methods can be countered by using more quantitative
methods as well (Jones, 1985, 98). The fact that certain methods may be
better suited to some elements of their favored theories than others is less
well appreciated. And the argument that scientists should embrace inductive
methods that will guide them to evidence outside the purview of their
favored theory is scarcely recognized at all. As a result, even researchers
who embrace multiple methods tend to emphasize either qualitative or
quantitative methods (McNeill, 1990, ch.6).
The tendency of different disciplines in human science to favor different
methods is undoubtedly a key source of the differing conclusions that these
disciplines reach with respect to public policy issues. Of course, different
theoretical perspectives also play a role, but the preceding analysis suggests
that these different theoretical perspectives both support and are supported
by the differing methodological preferences. The tendency of disciplines in
natural science to reach complementary theoretical conclusions may owe
much to the common reliance on experiments (though experimentation is
generally impossible in many areas of geology or astronomy). If so, it is
worth exploring whether the experimental method has biased research in
certain directions.2 It may be that the natural sciences are becoming more

2
Szostak (2003) suggested that the complementarity of theory in natural science could be
attributed to the fact that natural science disciplinary boundaries were more logical,
reflecting for the most part a division in terms of degree of aggregation. That analysis
provides a much more benign view.
202 Chapter 6

methodologically pluralistic; if so theoretical inconsistency may become

more of a problem.
Rule (1997) suggests that the major problem in human science is a
tendency to focus on theory-testing rather than explanation. Scientists do not
collect all evidence relevant to a particular problem or question, but only that
which facilitates the testing of a particular theory. The evidence is collected
and reported in terms of tests of a particular theory, and thus tends to be
forgotten when the theory loses favor. Rather than scientific advance, then,
after bursts of effort focussed on particular theories, the scientific enterprise
starts from scratch in pursuing the next theory. Rule, it should be
emphasized, recognizes that science can only advance through development
of theory. His problem is not with the use of theories but with an exclusive
focus on the testing of these. [Note here that much of the problem outlined
by Rule is due to a tendency of scientific communities to focus on theory
clusters with disparate elements, and reject these as a whole, rather than
identifying the range of applicability of those theories that seem realistic.] He
wants scientists to be primarily focussed on explanation [of causal links].
They will thus be open to all types of evidence. And they will report this in a
way that is useful even to those from a different theoretical perspective.
Strict methodological preferences will obviously be a barrier to this solution:
adherents of a particular theory will favor particular methods, and these will
limit the range of evidence consulted (Szostak, 2004a).
Note that the problem identified by Rule will be greater for those methods
that tend to be deductive rather than inductive. As argued in chapter 4,
inductive and deductive approaches are complementary at every stage of
scientific analysis. Exclusive reliance on deductive methods must severely
limit the development of theories.3 Care must be taken here: scientists in
reporting their research often emphasize the deductive aspects, and give
scant attention to the inductive processes by which they developed and
refined their hypotheses. They may thus leave a false impression that the
process relied exclusively on deductive methods. Experiments or statistical
tests that failed may have guided the researcher to change the research plan
in various ways. Often, the researcher will have had recourse to non-
reported inductive methods: note in this regard the frequent role of mistakes
in which laboratory scientists fail to control for some aspect of an
experiment, and the experiment then turns out to have surprising results (as
when Pasteur accidentally let some of his cultures die, only to find that these

3
“An openness to new ideas, a willingness to try out unpopular techniques, and a spirit of
passionate searching after the unfamiliar, Feyerabend argues, contribute more to the
progress of science than commitment to cool reason and accepted methods” (Klee, 1999,
201). This book embraces the flexibility inherent in Feyerabend’s philosophy of science,
while avoiding his scepticism about the very existence of a separate realm called science.
6. Drawing Connections Across these Classifications 203

proved the source of vaccines). The experimenter has recourse to a variety

of trial-and-error methods, and may try experimental setups not suggested by
theory. Nevertheless, a scientist running an experiment or performing
statistical analysis will be limited by the data they can readily combine, and
this can severely constrain the exploration of connections not posited by
theory. The problem of excessive reliance on deductive methods alone is
thus not just a matter of appearances, but a real brake on scientific advance in
at least some disciplines. Interdisciplinarians can again provide an antidote,
as long as they make use of inductive methods.
Despite the apparent similarities in typologies of theory and method, then,
this book does not support the idea that certain theories should only be
investigated with certain methods. While recognizing that different methods
have different strengths, and even that some methods may be of little or no
use for a particular inquiry, scientists are guided to pursue methodological
flexibility. Rather than assuming that one or two methods hold the key to a
particular investigation, scientists are advised to start from the opposite
assumption that many/all methods can shed some light on any theoretical
proposition. Communities of scientists and/or interdisciplinary scholars
should examine each of the dozen methods in turn and ask three questions:
does the method shed light on some aspect of the theory in question that
existing methods fail to address?; does it act to overcome biases inherent in
existing methods?; or does it perhaps serve to illuminate alternative theories
that existing methods are unlikely to highlight? Of course, there are costs to
applying multiple methods to a problem: smaller communities of scholars
may lack the resources to adequately evaluate results from multiple methods,
and effort devoted to one method must mean less effort devoted to others.
Nevertheless, if the answers to these three questions are positive, the benefits
of methodological plurality will likely outweigh the costs.

2. THEORY (METHOD) AND PHENOMENA (DATA)

It was argued in the previous section that scientists should use multiple
methods with respect to any theory. A similar argument can be made with
respect to method and phenomena: multiple methods should be used in
analyzing any phenomenon or causal link.4 This is especially the case given

4
Potter (2000, 69) suggests that methods are specific to particular subject matter. While it is
true that the strengths and weaknesses of methods vary by subject matter, students of
science should be wary of casually dismissing methods in any investigation. Goldenberg
(1992) argues that a researcher should not know what variables are relevant at the outset of
an investigation, and thus cannot know which method is best (note that even he seems to
underestimate the advantages of multiple methods here).
204 Chapter 6

that in understanding any one causal link it is generally helpful to understand

related links. As well, scientists should often wish to explore more than one
type of data, and this implies the use of multiple methods. Again the case of
(most of) the natural sciences lends itself particularly well to experiments,
but even here recourse to other methods is advisable.
What of the relationship between theory and phenomena? To be sure,
more than one theory should be explored, if only because theories can only
be tested in opposition to alternatives. And scientists can expect that over
most causal links more than one theory will have insight to provide.
Nevertheless, much more than was the case with method, when looking at a
particular link scientists can often exclude certain types of theory from
consideration: certain causal links may clearly involve a particular type(s) of
agency, a particular type(s) of action, a particular type(s) of decision-making
process, and/or a particular type(s) of dynamic process. Scientists must, of
course, be careful of assuming away alternatives, and should thus justify
their decisions to eschew certain types of theory. But when analyzing the
link between personality dimensions and art it makes sense to focus upon
individual intentional agency. Recall, though, that in studying any causal link
useful insights may be gained from the study of related links. In the case of
personality and art, the wider society will condition individual actions, and
thus theories that focus on other types of agency may be of use.
While certain types of theories can be excluded from a particular
investigation, care must be taken that relevant theory types are not excluded
arbitrarily. McNeill (1990) suggests that there is a mutual influence among a
community of scholars’ theoretical preferences, topics of interest, methods,
and various practical considerations (such as what data are readily available).
Interdisciplinary scholarship (see Klein, 1996) often speaks of a disciplinary
worldview that embraces such mutually supportive elements. It has already
been seen that scientists often needlessly constrain methodological
flexibility: with respect both to the theories and phenomena of interest,
openness to the widest range of methods would better serve science. This
problem will be exacerbated if scientists arbitrarily constrain the set of
theories that they apply to a particular set of phenomena.
The typology of theory can guide groups of scientists toward theoretical
plurality. First, scientists need to be clear about what set of phenomena they
are concerned with. Then they need to ask the 5W questions of the theory or
theories that they apply to this set of phenomena. From this they can identify
what types of theory are being ignored. And they can then ask ‘why?’ If
group agency is ignored, is this justified by the subject matter? (Can they
imagine ways in which group agency might play a role?). If attitudes are
eschewed for a focus on action (or vice versa) is this reasonable? Is it
reasonable to assume equilibrating mechanisms, or are elements of
6. Drawing Connections Across these Classifications 205

disequilibrium clearly present? Such an exercise will not be costless, but

will alert the relevant scientific community to unnecessary gaps in their
understanding, and encourage research along lines previously ignored.
Scholars in different disciplines (at least in human science) often tackle
the same causal links in ignorance of each other’s work. Different groups of
scholars, then, use different theories and methods to answer the same
questions. If scientific inquiry were catalogued by causal link (and then
within causal links in terms of theory and method; see chapter 7), scholars
would be less ignorant of related work. And then when the exercise outlined
above of identifying the range of applicable theory types for a particular set
of phenomena was performed, the efforts of different disciplinary groups
could be integrated. In analysing trade agreements, for example, economists
emphasize rational individuals and equilibrium outcomes, while political
scientists tend toward group or relationship analysis, diverse sorts of
decision-making, and dynamic processes. This sort of exercise would move
the scientific community as a whole some distance toward the goal of
theoretical and methodological flexibility – and also toward the unity of
science. It would help scientists to identify the relative importance of
different theories. Since all disciplines tend to arbitrarily constrain the range
of theory and method applied, though, even combining disciplinary efforts
will likely still cause some theories to be ignored unduly.
It was emphasized above that scholars should first of all be clear about
what set of phenomena they are discussing. This was not meant to prohibit
scholars from exploring the possibility that their favored theory has
applications beyond any particular set of phenomena. Rather, it was meant
to guide researchers to be clear about what links they are discussing at any
point in time. The purpose of science is to identify the range of applicability
of each theory.5 This must mean evaluating the theory link by link, with the
inevitable result that it will be found to be of great importance for some
links, lesser importance for others, and none for still others.
The process of scientific analysis can usefully be approached from
different vantage points. If individual causal links are taken as the starting
point, the best strategy is to first identify relevant theories, and work toward
identifying the relative importance of each of these. This in turn will likely
depend on the realizations of other phenomena:6 certain theories will be of

5
Ziman (2000, 127) maintains that natural scientists are increasingly moving away from an
emphasis on laws toward recognition of the context in which these operate.
6
Ragin (2000, ch.6) suggests analysis in terms of fuzzy sets such as “fully democratic”,
“mostly democratic”, and “not democratic.” Unpacking of phenomena (in this case state
decision-making systems) provides a better strategy. As well, Ragin argues that the length
of time a state has existed has important causal implications. But this is only important
because it affects various attitudes toward the state, the maturity of institutions, and other
phenomena: the best strategy is thus to examine the realizations of these phenomena.
206 Chapter 6

great import to a particular causal link in some circumstances (when in the

mountains, or unemployment is high, and so on) but not otherwise. In
evaluating the relative importance of differing theories, scientists should
have recourse to a wide range of methods.
If instead theories are taken as the starting point, the goal is to identify the
range of applicability. Over some causal links they may always be of great
importance. For others they may be of some import in certain circumstances
(realizations of other phenomena) and not in others. For still other links they
may be of no importance under any circumstances. Scientists could
reasonably aspire to map the results of this sort of inquiry onto Table 2.1
(and a similar table for natural science), with different sorts of arrows
capturing the strength and scope of the theory over different causal links.7 In
establishing the range of applicability of a theory, scientists should have
recourse to the widest range of methods. It is important to appreciate here
that scientists cannot establish the range of applicability of any theory
without contrasting this with alternative theories. As well, scientists will
often gain insight by examining links outside the scope of the theory. If this
exercise is performed for all theories, the scientific community will have the
same result as in the previous paragraph: a link-by-link understanding of the
relative importance of different theories. Some scholars (for reasons of
personality and experience) may choose to specialize by theory, others by
causal link, and still others by method; as long as the common goal is clear,
and other scientists are able to integrate the work of specialists, scientists can
work together toward the common goal. The goal, it should be noted, is
messy, but so too is the world with which science grapples.
While openness to diverse theories has been urged, the importance of
operationalizable theory should nevertheless be stressed. Since all
phenomena can be operationalized, at least indirectly (see chapter 2), any
theory that carefully specifies the relevant phenomena can also be
operationalized to some degree. But theories must also specify causal
mechanisms between phenomena. Since such linkages might not be
observable, it would be inappropriate to forbid theories on the grounds that
they can only be tested indirectly in terms of whether posited correlations
among realizations of phenomena occur. Still the advantages of theories that
posit observable relationships among phenomena deserve emphasis.8

7
Philosophers are coming to appreciate that theories are much like maps: they represent a
more complex reality, are social institutions, are abstractions that classify and simplify, are
functional, and require skilled interpretation (Ziman, 2000, 126-7). While perhaps not
always possible, Ziman suggests that mapping theories is generally desirable.
8
Seale (1999, ch.7) hails grounded theory (which grounds theory in empirical observation)
for stressing the link between theory and empirical research; he argues that this should be
the hallmark of all good research. He urges theorists to provide real-world examples of
6. Drawing Connections Across these Classifications 207

Philosophers of science worry that data can be theory-laden: that the

theory applied in an investigation will constrain what researchers observe.
Valiela (2001, 9-10) describes how various scientists over the centuries
“observed” Halley’s comet, guided by different theories of the heavens.
Boylan and O’Gorman stress that this concern is not a “tired old cliché” but
reflects the fact that human capacities to perceive and to theorize are linked
(1989, 92). They nevertheless argue that scientists both can and have over
time used data to distinguish among theories: scientific theories do not so
constrain what scientists perceive that they can never detect contradictory
evidence. Kincaid (1996, 33-4) identifies three causes of theory-ladenness
and suggests that they all can be overcome. First, theories impose certain
meanings on the world. As seen in the previous chapter such meanings can
be imperfectly translated so that scientists can draw comparisons. Thus,
Boylan and O’Gorman (1989) describe how the phlogiston theory of early
chemists was eclipsed even though it shaped how chemists perceived
chemical reactions. As long as scientists confront real world evidence,
alternative theories can be found to have better explanatory power. Second,
theories determine the questions asked of data. Note first that this is not
always so. Moreover, as long as scientists explicitly compare different
theories, a range of questions will be asked of any set of data. Third, every
test requires background assumptions about the data, and these may be
influenced by the favored theory of the researcher. Kincaid urges scientists to
strive for independence between background assumptions and the theory
being tested. Again, explicitly testing theories against alternatives can
counter this problem.
Following the simple practice of both applying different theories to the
same data and testing each theory against different types of data will provide
the best insurance against theory-ladenness. Only by comparing and
contrasting such results can scientists be sure that their favored theory has
not guided them to both the data and its interpretation that verify the theory.
As an example, Simonton (1988, 20) reports how early geologists developed
different theories depending on whether they lived in areas of sedimentary or
metamorphic rock; likewise psychologists develop different theories
depending on whether they study humans or animals, the normal or
abnormal. Exposing such theories to a wider body of data will guide
scientists to change and/or synthesize them.

their theoretical arguments. He notes, though, that grounded theory cannot help scientists
with unobservables.
208 Chapter 6

3. BETTER PARADIGMS

The word “paradigm” has diverse definitions. Even Thomas Kuhn

(1962), the modern popularizer of the term, used it in several different ways.
For present purposes, though, it can be loosely defined as some combination
of theory and method, and perhaps also phenomena and philosophical
perspectives, which is widely accepted within a particular discipline, such
that the vast majority of research engages only these. “Paradigms point
scholars in a given direction, suggesting what is and what is not important”
(Dogan and Pahre, 1990, 155). Kuhn famously argued that “normal science”
operates within paradigms, with scientists gradually establishing various
implications of core theories through the application of core methods.
Revolutionary science, in contrast, involves the conscious decision to
develop or draw upon new theories or methods. Revolutionary science will
occur (though the timing is unpredictable) when some scientists feel that the
existing paradigm faces too many difficulties in coping with reality.
At various points above, scientific communities have been urged toward
flexibility with respect to the range of phenomena to be studied and the types
of theory, method, and philosophical perspective to be applied. Is this book,
then, a call for permanent revolution, the displacement of normal science
forever? Not at all. As Kuhn emphasized, normal science is valuable. Being
able to take certain theoretical and methodological precepts for granted,
communities of scholars are able to readily comprehend and build upon each
other’s research. They can collectively explore various implications of the
paradigm. While scientists should not ignore contradictory evidence, neither
should they abandon a paradigm at the first hint of this: the evidence might
be wrong, or the paradigm might require only minor adjustments. When
revolutions come, they need not sweep away all that has gone before (but
can; recall the discussion of Rule above), but often absorb the older paradigm
into a newer one (see chapter 5). Relativity theory, for example, did not
displace Newtonian mechanics, but recognized it as an accurate explanation
of our daily reality. Physicists have incorporated the insights of the old
paradigm as a (important) special case within the new.
Paradigms also have psychological advantages. It was noted in the
previous chapter that scientists dislike admitting to the inevitable
imperfections of all theories and methods. Paradigms allow them to perform
valuable research while assuming that the theories and methods they employ
are perfect for the task. The danger, of course, is that in ignoring these
imperfections they come to believe that their paradigm copes better with
reality than it actually does. And thus scientific advance will be slower than
it could be. Fuller (1997, 19-20) criticizes Kuhn for suggesting that normal
scientists be shielded from the insights of history and philosophy of science,
6. Drawing Connections Across these Classifications 209

so that they could believe wholeheartedly in their paradigm. The thrust of

this book is that in facing up to the imperfections in theory, method, and
practice, and thus seeking the range of applicability of all theories, scientists
can effectively combine the best of normal and revolutionary science.
Scientists can have the psychological advantage of knowing that scientific
progress is possible without being wedded to an unnecessarily limiting vision
of how to proceed.9 If, of course, specialized researchers fail to appreciate
the advantages of looking outside the paradigm, interdisciplinary researchers
may play a crucial role in developing new paradigms.
The trick is to enhance the advantages of revolutionary science without
losing the benefits of normal science. On the one hand, the scientific
community as a whole must embrace theoretical and methodological (and
philosophical) flexibility. On the other hand, groups of scholars must
continue to collectively explore the implications of particular theories and/or
methods. At present, paradigms provide a very uneven coverage of the
possible combinations of theory, method, philosophical perspective, and
phenomena. That is, certain phenomena are studied with certain theories and
methods, even though others are clearly applicable. To solve this problem,
scientists could collectively either increase the number of paradigms (or,
more precisely, paradigmatic groups) within the world of science or enhance
the internal flexibility of paradigms. There is likely scope for both. Though
by some counts there are already some 10,000 discernible scientific fields of
inquiry, there are also many scholars whose talents are presently under-
utilized by the scholarly enterprise.10 However, simply expanding the
number of scholarly communities cannot overcome the danger that within
paradigms theories will be judged on an unnecessarily shallow empirical
basis.11 There is thus scope for advantageously broadening paradigms so as
to not artificially constrain the combinations of theory and method explored.
Paradigms should be defined primarily in terms of subject matter rather than
preferred theory and method. Such flexibility comes at a cost: the diminished

9
Seale (1999, 19, 30) appreciates that scientists dislike facing up to the uncertainties of
science. He urges them, however, to embrace fallibility, appreciate that there is no perfect
theory or method, and strive for a consciously self-critical research community.
10
Dogan and Pahre (1990, ch.3) suggest that small academic communities are more
productive than large. In part, this reflects the law of diminishing returns: as more
scientists enter a field the average contribution falls. Large groups often also become
focussed on jargon and theoretical hair-splitting. The authors could have noted the
countervailing advantages of largeness. If science advances through synthesis of ideas,
larger communities should generate more ideas to be combined. Of course,
interdisciplinary scholarship provides another source of combinations.
11
An example of this tendency is Silverman (2000, 99), who is sceptical of triangulation
because he feels that a scientific community needs to start from one theoretical perspective
and choose only methods and data that give structure and meaning within that perspective.
210 Chapter 6

ability of scholars to understand what others working on the same problem

are saying. And thus a flexible paradigm may likely in practice involve sub-
paradigms in which different groups of scholars specialize in different
theories and/or methods. As long as there is substantial overlap in sub-group
membership, communication within the larger paradigm should proceed
admirably.
Stephen Turner (2002) criticizes Kuhn for exaggerating the difference
between normal and revolutionary science. The Kuhnian approach limits
scholarly understanding of the gradual evolution of scientific insight.
Science, that is, proceeds by evolution as well as revolution. Communities
of scientists, no less than any other community, cannot be expected to
perfectly share any set of presuppositions. Individual scientists can thus be
expected to make subtle changes to a paradigm. If the paradigm thus evolves
through time, the need for a dramatic revolution may be forestalled or
eliminated. The arguments in this book can be taken as encouraging this sort
of evolution in scientific thinking.
Greater communication across paradigms should also be urged. It was
suggested above that paradigms be open to multiple theories and methods;
paradigms should also be careful of arbitrarily ignoring key phenomena. Yet
some limitation of subject matter will be necessary (or science would be one
big paradigm). Since causal links are ubiquitous, the understanding of
particular phenomena within one paradigm will inevitably benefit from
familiarity with the role these play within others. With respect to theory and
method, while individual paradigms should strive for theoretical and
methodological flexibility, they will still benefit from familiarity with how
well theories and methods and data fare in the analysis of other phenomena.
Dogan and Pahre (1990, 158) suggest that competition among paradigms is a
key source of scientific progress, but is limited by communication barriers.
Scholars from outside a paradigm often ask penetrating questions that do not
occur to those within. These can point normal science in interesting
directions, or pave the way for paradigmatic revolution. The sorts of useful
questions that Dogan and Pahre refer to would naturally include insights
about how the theories, methods, phenomena or data embraced by one
paradigm might shed light on matters of concern within the other.
Dogan and Pahre also suggest that paradigms can often usefully be
synthesized; after two or more groups of scholars have learned much about
related questions, these insights might be combined into a larger whole
(1990, 51). Collins (1998), like Dogan and Pahre (1990), has argued that
fragmentation and synthesis among scholarly communities is an essential
part of both scientific and philosophical advance. Specialized scholars
pursue fragmentation; interdisciplinary scholars strive to put the pieces
together: both are needed. Note here that specialization of a sort is both
6. Drawing Connections Across these Classifications 211

possible and desirable within the realm of interdisciplinary scholarship:

while some interdisciplinarians must pursue the broadest sort of integration,
others may specialize in the study of particular theories or methods or data,
and act to ensure that all communities that should be using these are, and are
aware of relevant experience elsewhere.
Along with their psychological advantages, paradigms have public
relations value. Much of the public credibility of science stems from the
existence of paradigms: accepted bodies of “knowledge” on which scientists
build (Ziman, 2000, 258). Yet we have seen that the public is all too aware
that scientific research often overturns previous results. The public is often
frustrated by scientific disagreements; they would prefer that there would be
one right answer to questions of global warming or breast cancer prevention
or psychological counselling. Willinsky (1999) argues that if the public could
appreciate the basis of particular disagreements they would become much
more supportive of the scientific enterprise as a whole (see chapter 7). The
point to be emphasized here is that a greater flexibility in paradigm structure
would not weaken, but rather enhance, public appreciation of the scientific
enterprise. The public knows that scientific understanding is fallible. Rather
than transmitting a false image of scientific perfection, the public can be
better impressed by articulating how scientists and the wider public can best
cope with this fallibility.
In performing any piece of research, scholars must take some
“knowledge” as given. The world is a complex place and scientists can
hardly hope to understand it if they always start from a position of complete
ignorance. In performing other pieces of research they may well question
the very theories and methods that they applied in the first instance. This
“web of belief” (see Rachels, 1993) approach, in which scientists accept
certain things as true today, but are willing to question any of their beliefs
tomorrow, is a valuable guideline for everyone.12 Individuals may find it
useful, for example, to believe that “Fred is honest”, if acting on this belief
seems to make their life easier than constant doubt. But if Fred is observed
acting dishonestly they should be prepared to change this belief. Scientific
theories are in this respect like the schemas used in everyday life, though
written down and generally more complicated; they provide an internal
image of the external world (Giere, 1988, 6). A scientific community acting
in accord with a “web of belief” approach would thus be both setting an
example and acting in a way that all should appreciate. And the “web of

12
The “wide reflective equilibrium” of Winkler and Coombs (1993) and “pancritical
rationalism” of Bartley (see Szostak, 1999) provide similar guidelines. Morton (1997), in
summing up the basics of the theory of knowledge, emphasizes human fallibility, and
urges individuals to assess the risk of a belief being wrong versus the risk of mistakenly
thinking it to be wrong.
212 Chapter 6

belief” approach provides the only solution to the problem of infinite regress
identified by Wittgenstein: if a scientist claims that the double helix of DNA
contains cytosine, they can be asked how they know; the response would
involve laboratory techniques for isolating and identifying components of
biological samples; questioning of these could be answered in terms of other
theories that implicate these techniques; and so on (see Moser et al., 1998,
ch.5). At some point, scientists must accept some statements as true, even
though these cannot be proven, or the scientific enterprise becomes
impossible (see Cussins, 2000, 336).13 The sort of flexible paradigms
suggested above accord much better with a “web of belief” approach than do
existing narrowly-defined paradigms.
Scientists should not err in the opposite direction and cast away theories
or methods at the first sign that they do not work. Science being a fallible
process, there may be problems with the evidence gathered rather than the
theory or method in question. Moreover, recall that scientists should be
striving to identify the range of applicability of both theories and methods.
A finding that a particular theory or method does not work in a particular
situation need not and should not imply that it is useless everywhere.14
In other words, scientists need to be more self-conscious and self-aware
about their paradigms. They need to remember that they are a convenience,
not a gift of God. They need to appreciate that all research requires a set of
assumptions, but that these in turn should not go unquestioned. They need to
appreciate the value to scholarly communication of shared assumptions, and
also the danger of these becoming articles of faith rather than questions for
further investigation. If scientists find paradigms to be psychologically
comforting, they need to force themselves from time to time to question
these. They should, in particular, investigate the possibility that a more
flexible paradigm would advance understanding more quickly. They should
be as explicit as possible about the core elements of their paradigm, and what
sorts of evidence would and should cause them to question it. And they
should gaze at alternative paradigms from time to time to see what they have
to offer.

13
Of course, the very statement that “science is possible” is itself unprovable. When
Feyerabend made this point, Lakatos responded that it was possible for scientists to
proceed on the assumption that science was possible, but be ready to critique this position
(see Motterlini, 1999, 13).
14
“Most new and contradictory ideas prove to be of little value;” it is better to reject some
good ideas than casually destroy consensus (Cole, 1992, 199). Many scientists likely share
this view, and there is obviously a kernel of truth to it. Given the danger that scientists
thus too readily dismiss novelty – it is all too easy to rationalize narrow-mindedness within
this perspective – it would be useful to seek to establish the empirical validity of this point
of view.
6. Drawing Connections Across these Classifications 213

4. CRITIQUES OF SCIENTIFIC PRACTICE: ARE

THEY CORRELATED WITH OTHER
CLASSIFICATIONS?

Throughout this book, and especially in chapter 5, it has been argued that
science is imperfect yet possible. It is thus important that scientists know the
potential biases that can afflict scientific practice, in order to combat these.
While all scientists should be aware of all potential imperfections, particular
scientists likely need to be more aware of some than others. This section
explores whether the particular problems discussed in chapter 5 are of greater
importance with respect to certain phenomena, data, theories, or methods.
Since the typology of critiques is quite different from the other classifications
in this book, the analysis here must be more detailed than that in previous
sections of this chapter. Note that the very question of “how are certain
biases or errors distributed across different types of science?” has hardly
been asked heretofore. While many correlations seem obvious there is much
scope for further research:
We can proceed through the various biases and errors identified in
chapter 5, asking of each how they might be associated with different
phenomena, data, theories, and methods. The various human perceptual
limitations will afflict all of science. As noted in chapter 2, however, some
phenomena are more observable than others, and thus on average less subject
to perceptual difficulties.
As noted in chapter 4, some methods (such as PO) are inherently more
subjective than others (such as statistical analysis), though researcher biases
can affect the results from the application of any method. Awareness of
potential biases, and efforts at replication, are particularly though not
exclusively important when more subjective methods are used. It is likely
also true that some theories invite greater subjectivity than others:
psychoanalysis invites reflection on unseen mental processes, and
phenomenology is deliberately vague.
Researchers may be influenced by the social structure or power relations
of their society, and guided to results that support the status quo. This type
of bias will obviously only matter for research topics that have implications
for the status quo. Broadly speaking, human science is in greater danger here
than natural science. However, there are exceptions: scientists studying
either the environment or human genes will not be immune from such
influences. And history shows that even studies of the earthly orbit can
offend the powers that be.
A similar point can be made with respect to the possibility that scientists
will pursue answers that reflect their ethical predisposition. Such a bias is
likelier to operate in the study of intentional agents. However, as argued in
214 Chapter 6

Szostak (2004), ethical evaluation of non-intentional agents such as

institutions is also valuable: do these reflect and/or encourage ethical
behavior? Even when ethical analysis does not have a place in scientific
research itself, it can still be important in the communication of scientific
results to the public: scientists tend to have a consequentialist orientation,
and must be prepared to speak to those who would evaluate scientific
research in terms of one of the four other types of ethical analysis.
Disciplinary biases, by which researchers are guided to ignore the
phenomena, methods, and theories outside their discipline, affect all
scientists. Some disciplines are perhaps more interdisciplinary in orientation
than others. Sociology, for example, embraces a wider range of theory and
method than economics (though sociology is characterized by subfields with
a less open outlook). The subject matter of some disciplines is better suited
to the application of only one or two theories or methods than is the subject
matter of others. In particular, the natural sciences have likely divided up
phenomena more logically – to a large extent in terms of level of aggregation
– than have the human sciences (see Szostak, 2003). Most natural sciences –
those that focus exclusively on non-intentional agents – must rely on a subset
of theory types, and can lean more heavily on one favored method.
Nevertheless, even natural scientists should be open to differing theories and
(especially) methods, as well as links to other phenomena.
The incentives toward overspecialization, and the production of lots of
acceptable work rather than smaller amounts of great work, afflict all of
science. If, as suggested above, natural science has a more logical structure,
it may be somewhat better at tying specialized research together. But there is
a crying need across science for greater efforts to synthesize diverse
research. The next chapter will discuss one stratagem for alleviating this
problem. More generally, this book has pointed to various roles to be played
by interdisciplinary research, and the concluding chapter will outline several
ways in which this book’s classifications could aid interdisciplinary
scholarship. Hopefully, theoretical and methodological flexibility will also
encourage “big thinking” by specialized researchers. Learning more than one
theory or method is time-consuming but mind-expanding. It thus encourages
a shift in attitudes toward quantity versus quality in publications.
The (related) temptation to use the easiest method – that which takes the
least time and money – afflicts all of science as well. The effect of this bias
will depend on whether the easiest method is well suited to the question at
hand, and how well suited the harder methods are. In chapter 4 the
possibility was raised that the popularity of statistical analysis of secondary
data in social science may reflect its ease of use, especially relative to PO.
The prominence of textual analysis in the study of literature, art, and
philosophy (the humanities) likely also reflects in part the ease of use of this
6. Drawing Connections Across these Classifications 215

method. While textual analysis and introspection may be the only access
scholars have to certain questions in this area, there is clearly scope for the
application of a wide range of methods to questions concerning the social,
cultural, or political causes and effects of changes in art, literature, or
philosophy. The prominence of experiments in natural science may also in
part reflect ease of use, though there are clearly other good reasons for this.
Scientific debates are not won anywhere by the facts alone: rhetoric is
everywhere important. Scholarly communities that exalt style over substance
are most likely to be swayed by rhetorical flair. Clarity of expression in
theorizing was urged in chapter 3: the assumptions – that is, realizations of
all relevant phenomena necessary for a result to hold – and causal
argument(s) should be clearly articulated. Even – perhaps especially –
assumptions taken for granted by a particular community should be stated.
Some scientific disciplines encourage clarity more than others. Scientists
must nevertheless everywhere be on guard for unstated assumptions. In
some areas, such as the analysis of how art moves individuals, there may be
great scope for rhetorical flair in the provision of supporting evidence, but
the theory itself should still be expressed clearly.
All sciences benefit from interaction with the real world. Natural
scientists benefit enormously from interaction with technologists. Much
interchange of this sort occurs, though there are concerns that certain
institutional arrangements limit the free flow of scientific understanding.
Human scientists, however, have more limited contact with policymakers,
and this likely limits the effectiveness of both social science and social
policy.
Research requires money. Those who fund research can thus introduce
biases into the sorts of questions investigated, and the answers found.
Governments in developed countries have – laudably – established arms-
length granting agencies to reduce this bias. Various government agencies,
private firms, and non-profit foundations also provide research funding.
These naturally fund endeavors that interest them. These need not be the
research questions that society as a whole would wish to see studied (though
in the case of government agencies, it could be hoped that they are indirectly
expressing the public will). Even more dangerous is the possibility that
funding agencies will make it clear that they will reward researchers who
find the right answers. This danger is perhaps most obvious in the case of
drug trials financed by drug companies. But the natural sciences more
generally are at greater risk than human science: research in natural science
tends to be more expensive, and is more likely to have implications for the
profitability of private firms. Finally, note that cost affects the type of
research performed: scientists with limited funds may not be able to pursue
the research strategies they think best; conversely as universities increasingly
216 Chapter 6

reward researchers for the amount of research funding they bring in,
researchers are given a perverse incentive to favor expensive research
projects.
Problems associated with the hierarchical nature of science are ubiquitous
(though Whitley, 2000, argues that fields differ in the type of hierarchy).
Scholarly specialization provides at best a partial solution, and carries its
own costs. All scholarly communities are wasteful (see Hull, 1989), and all
discourage certain sorts of innovation. A partial solution to this problem will
be suggested in the next chapter.
Paradigms are important in most disciplines. Some disciplines are viewed
as pre-paradigmatic: these may embrace an attractive array of theories and
methods, but at the cost of there being no core of ideas around which debate
can concentrate. For disciplines that have paradigms, the question to be
asked is whether the disciplinary preferences for theory, method, and subject
matter are strong, and if so to what degree they limit access to evidence that
would threaten the paradigm. This in turn will depend on many of the
questions asked above, especially that of how well suited a particular theory
or method is to a particular area of inquiry. Interdisciplinary scholars may
have a clear role to play in showing what a discipline loses by closely
adhering to a particular paradigm.
Finally, what is science? (this was the ‘what?’ question in chapter 5).
While different biases are of different relative importance for different
research questions, all of natural and human science deserves to be called
“science”. Four points should be emphasized here:
x No type of science is perfect.
x Some types of science are subject to fewer sources of error and bias than
others.
x No type of science is impossible. There are strategies for coping, albeit
imperfectly, with all biases. There are likely, nevertheless, limits to how
far understanding can progress.
x While natural science is less subject to most biases than human science,
this is far from always being the case.

These points have been made with respect to “types of science”. They
could be reworded such that they spoke instead of theory types, methods, or
phenomena.
Chapter 7
CLASSIFYING SCIENTIFIC DOCUMENTS

The archives of modern science are of Byzantine complexity. Even the

most elaborate classification schemes cannot provide pointers to every
scrap of knowledge that might be applicable in a particular context.
Ziman, 2000, 261

The classifications developed in this book can serve many purposes.

Preceding chapters have emphasized the advantages of identifying the full
range of scientific possibilities, and then identifying the strengths and
weaknesses of different ways of doing science. Both research and teaching
can benefit directly from these classifications. From time to time it has also
been hinted that these classifications could be used to classify scientific
papers and books, whether in traditional libraries or in on-line archives, so
that researchers would be better able to find documents relevant to their
research interests.1 If it is useful to enhance the productivity of science, then
there is great value in enhancing access to scientific information, for this lies
at the heart of the scientific project (Palmer, 2001, xii).
Many studies have found that scientists seriously under-utilize subject
catalogues and indexing and abstracting services. Weinberg (1988)
attributes this unfortunate state of affairs to the fact that such classifications
do not adequately detail what is novel about a work. They provide
information of use to the novice concerning what a work is “about”, but

1
Marcella and Newton (1994,3) identify the purpose of classification with being able to find
things. While classifications have other uses, this use is undoubtedly of great import.
Journal articles rarely receive the classificatory treatment allocated to books, being judged
to be “component parts of larger documents” (Svenonius, 2000, 24); this severely limits
the ability of scholars to locate relevant research in other fields.
218 Chapter 7

generally say little about the ideas contained in a work. Researchers looking
for documents that contain applications of a particular theory or method, or
studies of a particular causal link, thus rely on bibliographies and
recommendations from other scholars, while realizing that they thus likely
miss important works relevant to their research interests. Weinberg (1988)
recognizes that the theory employed in a work is a key piece of information
rarely noted in subject catalogues. She also mentions method, and cites
favorably Swift et al. (1978), who had argued for classifying works with
respect to both theory and method (Palmer, 1996, also advocates
classification with respect to methods). Yet Weinberg makes no
recommendations for change, feeling that any effort to classify documents by
theory and method would necessarily be too complicated.2
Weinberg details some of the costs she had incurred in her own research
due to deficiencies in subject catalogues. It took her much longer to track
down ideas that she had encountered elsewhere, though she could not
remember exactly where; sometimes, this would prove impossible, and
incomplete citation would result. She could not be entirely sure that a
particular idea she was writing about had not been treated previously. If she
could find some works on similar topics, by asking colleagues for
suggestions, she could not be confident that she was not missing important
works, particularly if these appeared in different fields.
Limitations in systems of document classification do not just hinder the
individual researcher, but hobble the scientific enterprise. “Progress [in
scientific understanding] depends on cumulative scholarship, which in turn
depends on scholars’ ability to access all that has been created by the human
intellect” (Svenonius, 2000, 29). Researchers frequently “re-invent the
wheel” by duplicating research that they were unaware of. The loss of
valuable research, which is not retrieved because it is not classified in terms
of its novel elements, is a significant cost. Moreover, science advances by
carefully establishing the range of applicability of its theories and methods.
If classification systems make it difficult to retrieve and thus compare works
that apply a particular theory or method to quite different questions, then this
key form of scientific progress is rendered exceedingly difficult. Finally, the
scientific community is often accused of amassing many bits of information,
but making little attempt to tie these together. While scientific incentives,
which favor new research over integrative research, are partly to blame for
this state of affairs (see Glassick et al., 1997), difficulties in searching for
relevant literature are also at fault. “As research and knowledge become

2
Weinberg notes that full-text searching is often of limited utility in searching for theories,
as theoretical ideas can often be paraphrased in numerous ways. As full text searching
becomes increasingly possible, a common vocabulary of theory, method, and phenomena
becomes increasingly valuable.
7. Classifying Scientific Documents 219

more interdisciplinary, the academic subjects represented in our research

libraries become increasingly ill-suited to the conduct of research” (Palmer,
1996, 66). Interdisciplinary research in particular requires the ability to
search widely for relevant documents, an ability that is seriously constrained
by present document classification systems (Bates, 1996).
Weinberg’s suggestion that it is not practicable to classify documents in
terms of ideas must seem plausible at first glance: given the costs of not
doing so, librarians would surely have done this if they could see a way
forward. Weinberg’s recognition of the centrality of scientific theories and
methods is, however, suggestive of how information scientists might
proceed. Of course, until now a sound basis from which to classify the
seemingly endless variety of scientific theories and methods has been
lacking. This chapter will discuss how the classifications developed earlier
in this book can allow the development of a system for classifying scientific
documents that would greatly enhance the efficiency of the scientific
enterprise by solving or alleviating the various problems noted above.

1. FACETED OR ENUMERATIVE?

Information scientists debate the relative advantages of “enumerative”

versus “facet-based” systems (with most now favoring a facet approach;
Marcella and Newton, 1994, 17). The Dewey and Library of Congress
systems are predominantly enumerative: they are based on a hierarchical list
of subjects, with several major categories subdivided into several layers of
constituent subjects. Such systems must be extremely detailed if documents
on every subject are to be classified precisely: the Library of Congress
classification is summarized in some thirty thick volumes, which must be
regularly updated as new subjects are recognized. While the Dewey and
Library of Congress systems are primarily enumerative, faceted elements
have crept in: common notations for such characteristics as language used,
time period covered, and geographical region covered have been developed
and can be added to any call number within the enumerative scheme.3
The more recent classification systems have depended primarily on facet
analysis. The generally recognized first mover in facet analysis was
Ranganathan: in various works he suggested that five types of facet should

3
Call numbers also contain author and/or title specific notation to distinguish works that
would otherwise receive the same call number, and also generally note edition and year of
publication. While the concern of this chapter is with classifying works so that subject
catalogues are more useful, the addition of new dimensions of classification will reduce the
size of classes and thus the need of cataloguers to distinguish works by author or title .
220 Chapter 7

be recognized: personality, matter, energy (primarily concerning the process

under study), time period, and place. He then went on to develop (that is,
enumerate) notation for lists of these five types of facet. A facet-based
system has one huge advantage over enumerative schemes: it can better deal
with documents that combine two or more subjects (Marcella and Newton,
1994, 18-20). New areas of study tend to emerge as combinations of old:
biochemistry combines elements of biology and chemistry for example. As
seen previously, both Dogan and Pahre (1990) and Collins (1998) have
argued that fragmentation and synthesis are key elements in the evolution of
science. Strictly enumerative schemes can only cope with novelty by adding
the new field to the existing scheme (with often troubling questions of where
it best fits: with biology or chemistry?). Facet-based schemes can instead
simply combine the notation for the facets combined: biology plus chemistry
equals biochemistry (though there is still a question of which goes first,
which will affect whether it gets classified closer on the shelf to biology or
chemistry). As we shall see, a similar benefit can be gained by linking
notation for phenomena when designating causal links. Of course, faceted
schemes must still be hierarchical: different types of biochemistry require
separate identification and notation just as in an enumerative scheme.
Though it is not widely recognized in the literature, Ranganathan’s five
types of facet accord with the 5W questions utilized above. When he speaks
of “personality” he is asking ‘Who?’: what individuals or groups or objects
are the subject of the work in question? When he speaks of time period and
place, he is obviously engaging questions of ‘When?’ and ‘Where?’, though
not in quite the way these were addressed in chapters 3 and 5. His “matter”
facet accords well with a simple ‘What?’ question, though again somewhat
less well with the particular way in which this book has approached ‘What?’
His “energy” facet, concerned as it is with the sort of process involved,
reflects reasonably well a ‘Why?’ question. It is noteworthy that the
literature often presents Ranganathan’s five facets with little justification. A
system more firmly grounded in terms of these five questions would be more
easily justified. It would also be easier to recall and to utilize: both of these
are qualities highly valued in a document classification system.
The classifications outlined in previous chapters combine elements of
enumerative and faceted approaches: the list of phenomena and the list of
methods are enumerative, while the typologies of theory and scientific
critique are faceted. Since all extant document classification systems
combine elements of these, this is no bad thing. Indeed it opens the
possibility that these classifications can form the basis of a superior
document classification scheme. In what follows, then, elements of
enumerative and faceted schemes will be blended.
7. Classifying Scientific Documents 221

2. CLASSIFYING BY PHENOMENA STUDIED

All document classification schemes must decide which characteristic of

a document is to be noted first. This decision determines which documents
will be located in adjacent spots on the (literal or figurative) shelf. Literature
could be classified first by type (prose versus poetry), time period, or
language: this will determine whether eighteenth century French poetry is
shelved beside eighteenth century English poetry or nineteenth century
French poetry or eighteenth century French prose. There are good reasons to
suggest that subject matter deserves pride of place in universal systems of
classification (where universal means widely-applied as opposed to a special-
purpose scheme for a particular collection). This decision is made by most
existing classification schemes, and reflects a sense that customers are
generally more likely to seek works on a particular subject than works from a
particular country or author or time period (though exceptions may be made
in the order of priority when treating works of history or literature). This is
certainly true of the general public, and it is useful to strive here for a system
that serves both these and scientific researchers. Scientists may often wish to
search primarily by theory or method, but will generally have a particular
topic in mind as well.
Existing classification schemes have usually adopted a discipline-based
rather than topic-based approach to subject treatment (Marcella and Newton,
1994, 33). As Bulick (1982) has described in detail, this characteristic has
caused great confusion as disciplinary boundaries have shifted over the
decades, and as interdisciplinary fields have blossomed. Three broad types
of problem emerge: subjects that are studied by more than one discipline
receive multiple classifications, subjects that are inherently interdisciplinary
have no obvious place, and subjects that combine existing subjects result in
arbitrary classification decisions. “In a properly designed hierarchical
classification, each subject should have one and only one place where it fits
into the system” (Hunter, 1988, 35). All three of these problems will limit the
ability of both scientists and the general public to survey the literature on a
particular topic. One likely outcome will be that researchers will be
unfamiliar with works on a particular topic that emanate from disciplines
other than their own. These will not be shelved beside works on that topic
from the home discipline, nor will networking with disciplinary colleagues
alert one to their existence.
A superior system is possible. In the second chapter a hierarchically
organized list of the phenomena of interest to human scientists was
reviewed. A similar list could be developed for natural science, though
notably the three problems listed above are much less severe with respect to
subjects in natural science. Each phenomenon has one and only one place in
222 Chapter 7

such a list. Searches by phenomenon will thus uncover all relevant works
regardless of discipline. Moreover, each phenomenon has a precise
definition. Culture, for example, comprises the hundred or so subsidiary
phenomena listed, and no other phenomena. Scientists are often frustratingly
vague or contradictory in defining terms. See, for example, the hundreds of
definitions of “culture” or “religion” encountered in Szostak (2003). The
hierarchical list of phenomena would encourage greater clarity, and thus ease
the task of classification. It is noteworthy in this regard that one task that
interdisciplinary researchers want information scientists to perform is the
translation of jargon across disciplinary boundaries (Palmer, 2001, 131).
While information scientists commonly compile thesauri to this end, it would
be even better to strive for common usage of key terms.
As scientific understanding advances phenomena once viewed as similar
may be found to be different in important ways, and vice versa (Hjorland,
1994). It is thus advisable that documents be defined in terms of phenomena
defined functionally (that is, in terms of the role they play), so that
classification systems not need continual revision. This is, of course, the
guiding principle pursued in chapter 2. Ironically, Hjorland goes on to argue
that since different disciplines treat the same phenomenon from different
perspectives, it is best to classify documents by discipline. If documents are
classified by phenomena, however, these differences can be readily
compared and contrasted, rather than pursued in isolation.
A simple notational scheme whereby causal links are denoted by the first
letter of the category name was employed in Szostak (2001) and (2003): a
link wherein some cultural phenomenon influences some element of social
structure is thus denoted by an arrow from ‘C’ to ‘S’. A more complex
system would be needed for document classification, in which either letter or
number codes were attached to second and third level phenomena: a
particular occupation might be designated SO97, where O stands for
occupation, and 97 represents the place of the particular occupation on some
list of occupations. Note that notation serves to indicate where in the
hierarchical structure of phenomena a work fits; Marcella and Newton (1994,
48-54) urge notations that reflect hierarchy. The use as notation of the first
letter of categories, and where possible phenomena lower in the hierarchy,
also makes the classification evocative and thus facilitates recall.
As noted with respect to occupations, the list of phenomena in Table 2.1
would have to be fleshed out by developing lengthy lists of subsidiary
phenomena. Fortunately, lengthy lists of such phenomena already exist
within existing classification schemes. The classification of phenomena
would simply determine where such elements fit.
While existing document classification schemes could be drawn upon in
fleshing out lists of phenomena, the result would nevertheless be much less
7. Classifying Scientific Documents 223

complicated than these. One benefit has already been discussed: there will
no longer be multiple entries for a single phenomenon. Existing schemes
also presently treat many causal links between phenomena as subjects. By
listing hundreds of phenomena, and providing appropriate notation for these,
it becomes possible to identify thousands of causal links by joining the
notation of appropriate phenomena. For example an arrow from IPE to SO97
could indicate how the personality dimension of extroversion-introversion
(within the category of Individual Differences) affects performance within a
particular occupation. Note that in so doing a technique from faceted
schemes (linked notation) is borrowed and applied to an enumerative
scheme. Marcella and Newton (1994, 60) applaud systems that allow “the
expression of compound subjects via notational synthesis.” And Rowley
(1992, 240) argues that “... any tool for the organization of knowledge must
[among other things] take into account the relations between subjects.”
Most scientific research investigates how one or more phenomena exert an
influence on one or more others. Researchers interested in the link between
IPE and SO97 will want to be able to find all such works without necessarily
being overwhelmed by every scholarly reference to either IPE or SO97. The
proposed scheme allows the researcher to readily distinguish works focused
on the causal link from those describing one of the phenomena involved. In
other words, it identifies the subject matter of a work much more precisely
for the researcher than do existing schemes. Note that while existing
schemes often classify causal links as subjects, there is no notational signal
of the difference between a work that studies one phenomenon and a work
that investigates relationships among two or more phenomena. The system
proposed here clearly differentiates these quite different types of work.4
The proposed scheme would be grounded in a logical structure focused
upon the subjects scholars study. These, unlike disciplinary boundaries, do
not change, or at least not much: in human science all relevant phenomena
have likely been identified, while in natural science any new phenomena
discovered are generally components of long recognized phenomena (chapter
2). Thus, interdisciplinary fields such as biochemistry pose no problem for
classification: the information scientist simply asks what it is that
biochemists study and classifies their works accordingly. If an
interdisciplinary area asks how one phenomenon affects another it is
classified around that causal link.5

4
If works were also classified in terms of the data employed, researchers could readily
ascertain which data have been used to investigate which phenomena or causal links.
5
Since flexibility is hailed as one of the key advantages of faceted schemes (e.g. Rowley,
1992, 179), it is worth noting that this classification of phenomena allows the development
of a flexible enumerative scheme (see below).
224 Chapter 7

Most existing classification schemes have been developed in an inductive

fashion. Classificationists – those who develop classifications for classifiers
to use – surveyed the works that they needed to classify and went from
there.6 Additions have been made to all schemes as new works were
encountered. Information scientists would prefer a theoretical justification
for classification schemes, to reduce the arbitrariness of decisions regarding
where particular materials should be placed. Rowley (1992, 159-60),
recognizing that a basic problem of subject classifications is that different
people define subjects differently, thus praises the effort to build languages
or schemes which depend upon some theoretical views about the nature and
structure of knowledge.7 One key advantage of a scheme based on a list of
phenomena such as that in Table 2.1 is that this logical structure reflects a
great deal of scholarly consensus concerning where phenomena fit best. It is
not an ad hoc construction but one grounded in scientific understanding.
While it is likely imperfect in some ways, and may itself require some
adjustment over time, such adjustments will be much less frequent than with
a system that has no theoretical basis.
One advantage of any hierarchical classification system is that it will place
related subjects together, both on the shelves to some extent and in the index
(Rowley, 1992, 163, Marcella and Newton, 1994, 7-10). Obviously this
advantage will be accentuated by a logical structure relative to an ad hoc
compilation. Scientists should not only be able to search by phenomenon or
causal link; they should be able to quickly identify material relevant to
closely related phenomena or links.
Works would be classified for “shelving” purposes in terms of the
primary phenomenon or causal link studied. Information scientists appreciate
that only so much can be accomplished with shelving coding: in this case
material on the same or similar causal links is put together (and it can be
hoped that those interested in a particular link from C to S will peruse the
corresponding link from S to C for relevant information). Much more can
and must be done with coding for the subject index (Hunter, 1988, 44-5).
For the purposes of the subject catalogue, it would be useful if all
phenomena and causal links studied to an appreciable degree were
referenced. Authors and/or publishers should be encouraged to provide
guidance to bibliographers here.

6
Experts refer to “literary warrant” to justify developing classifications around existing
works. Nevertheless, some 80% of books tend to fall within some 20% of existing classes
(Marcella and Newton, 1994,34).
7
Langridge (1992, 18) notes that scientists classify what they study, not what they produce.
This last task is performed by bibliographers, whose duty is not to produce knowledge but
“to organize the knowledge produced by specialists so that it may be available for
whosoever requires it”. The system outlined here would provide bibliographers with a
scientific basis from which to proceed.
7. Classifying Scientific Documents 225

3. CLASSIFYING BY THEORY USED

After having classified subject matter, a classification scheme should

proceed to code works in terms of first theory and then method. Coding for
theory and method would be a major innovation. It would further aid the
task of simplifying document classification, for existing schemes often treat
(some) theories and methods as “subjects”. General works on a particular
theory or method could be classified under the most relevant category, or
within a generalia class. Theories or methods applied to a particular
phenomenon or causal link would be classified under that phenomenon or
causal link. Nevertheless, in the subject catalogue all works using a
particular theory type or method could be easily identified. Since theories
and methods are only coded for general works on theory or method, this is at
present not feasible in almost any library or electronic archive.
At present, researchers interested in locating works using a particular
theory or method in other fields are forced to rely upon a variety of flawed
research strategies. Most likely, they will seek advice from colleagues in
these other areas, while appreciating that they will inevitably miss much
(Weinberg, 1988). Or, once aware of related research in another field –
perhaps from reading about this in a general publication such as Science –
they will search for works by names they have heard of in that field (Palmer,
1996). Or, they may locate one or two relevant works, and then use citation
indices to locate related works: this strategy is advocated by Bates (1996)
and White (1996). All of these strategies are both time-consuming and likely
to miss important relevant works. Note in particular that they largely guide
search of one field at a time, and thus limit the range of research that one
scholar will locate.8 Information users will only search for what they think is
likely to exist (Mann, 1993): if a researcher does not think that a particular
theory or method has been applied in a particular discipline, they will not
look there. Yet the most useful information is often that which surprises the
researcher, and thus providing scholars with easy access to information they
did not expect may be particularly valuable (Palmer, 2001)
As noted in chapter 3 it is straightforward to place all major scientific
theories within the typology of theory types. Importantly, it was found that

8
Fiscella (1996) thus advocates the publication of interdisciplinary bibliographies. In some
areas of inquiry these do provide a very inclusive guide to the relevant literature. Yet these
are time-consuming to produce due to the problems with classification systems noted here.
Even these bibliographies may miss key works as a result. And huge numbers of
bibliographies (regularly updated) would be needed. Likewise, Palmer’s (2001) suggestion
that information scientists be included on interdisciplinary research teams is both more
costly and less effective, especially for individual researchers, than having more
comprehensive classification systems in the first place.
226 Chapter 7

the points that commentators emphasize when describing a theory are, in

fact, answers to the 5W questions. There are only a handful of answers to
each of the five questions. And some combinations are impossible: non-
intentional agents cannot change attitudes nor make decisions. Thus the
number of cells in the typology is manageably small, and could be captured
in terms of one or two notational spaces. This is especially the case if no
attempt is made to code for degree of generalizability.
With phenomena, it would generally be possible to use the first letter of a
phenomenon for notational purposes. This will be less feasible for theory
type, since each notation must necessarily refer to some combination of
answers to distinct questions (unless four or five notational spaces are
devoted to theory type). Moreover, it will take some time for scholars to
familiarize themselves with the typology. While they will welcome the
possibility of searching by theory, they will often wish to search in terms of a
particular theory (such as action theory or rational choice theory), rather than
a cell in the typology. Necessarily, then, notation will not aid the recall
function, until a particular researcher has learned to associate a particular
symbol with a particular theory type.
Some sort of thesaurus would be needed to translate individual theories
into theory types. In some cases this would be easy: rational choice theory
provides clear answers to each question. In other cases, different versions of
a particular theory fit in different cells of the typology. This information
itself should be of great interest to the researcher, for it will alert them to
differences that exist within a particular theoretical camp. Without the
typology to guide them they could easily be bewildered by conflicting
interpretations of a theory.
One obvious advantage of this approach is that it would alert researchers
to similar theoretical efforts across diverse disciplines, even when these
operate under different names in different disciplines. And thus researchers
could much more readily ascertain the range of applicability of a particular
theory. This advantage in turn flows from the key advantage of the typology:
it provides a means (and likely the only means) for allowing thousands of
theories to be captured in a classification of manageable scope. It will
nevertheless be desirable to distinguish particular theories within a theory
type. This should be done theoretically rather than semantically. That is, it
should not be enough that theories go by different names. In order to merit
distinct notations there should be some identifiable criterion by which the
theories are to be distinguished. Different types of agent might be
distinguished in detail, for example, in terms of the phenomena identified in
chapter 2. Information scientists may be able to make a significant
contribution to the scientific enterprise by inductively establishing different
7. Classifying Scientific Documents 227

types of action or attitude in relation to the ‘what?’ question, or different

varieties of the four time paths in response to ‘why?’
It would also be useful to researchers if some distinction were made
between works about a theory and works applying a theory. This may not be
absolutely necessary, as works applying a theory would be filed under the
relevant phenomenon or causal link, while works about theory would be filed
under a category or generalia class. Those familiar with the details of the
classification scheme might thus be able to distinguish these two types of
work by the subject notation. This distinction should, nevertheless, at the
very least be made in the subject catalogue.
Theories evolve over time. One further advantage of the typology is that
theories are not blindly followed as they morph into a quite different
theoretical animal. Rather, if advocates of a particular theory come to
provide a different answer to one of the 5W questions, this would be
reflected in the classification of their work.9 As noted above, thesauri would
need to be complex enough to guide researchers searching under a particular
theory name to all relevant theory types.
Finally, as with phenomena, while a document should be classified in
terms of the primary type of theory utilized, the subject catalogue should
reference all theory types utilized.

4. CLASSIFYING BY METHOD USED

There being only a dozen methods, it would be straightforward to provide

each of these with a simple symbolic designation. As with phenomena,
seeking to use the first letter of each method would aid both intelligibility
and recall (though some methods would have to be renamed, as these often
start with the same letter). Lower case or Greek letters might perhaps be
used to distinguish this notation from that of subject, though notational order
would do so in any case. While each distinct method can be captured with
one notational place, another notational place or two might well be devoted
to distinctions within methods. At least some methods can usefully be
unpacked: with respect to statistical analysis cross-sectional analysis could
be distinguished from time-series analysis and various statistical techniques
identified. Flexibility would be needed here as new techniques are
occasionally developed.
While classification codes are available for some methods at present,
these are generally used to classify works about a method rather than works

9
Dogan and Pahre (1990, 151-2) warn researchers borrowing a theory from another field to
keep abreast of changes within that field, and note that researchers can cling to a theory
that has been rejected in its home field. They should at least know why this was the case.
228 Chapter 7

applying a method. Moreover, methods (and even moreso theory types) are
not classified in an exhaustive and coherent manner in any existing system of
classification. Researchers can be blissfully unaware of insights from the
application of a particular method to other phenomena. Scholarly efforts to
comprehend the strengths and weaknesses of methods are thus blunted.
As with theory, it would be useful to distinguish works about methods
from works applying methods. Again, the fact that works about method
would be classified under categories or generalia rather than particular
phenomena or causal links provides an imperfect solution to this problem.
With such a system of classification in place, a scholar interested in applying
a particular method to a particular phenomenon or causal link would be able
to quickly identify any work that had previously applied that method to that
phenomenon or link. If desirable, the researcher could readily expand the
search to embrace closely related or similar phenomena or causal links.
Alternatively it would be straightforward for the researcher to ascertain what
methods had been applied to the particular causal link in question. Likewise,
a researcher could easily ascertain what methods had been used to evaluate a
particular theory type, both in general and with respect to particular causal
links. A researcher with a more general interest in a method could instead
find out to what range of phenomena or causal links that method had been
applied. This alone would tell them something useful about the range of
applicability of the method (and they could learn more by studying the
difficulties encountered by researchers in these varied applications of the
method).10 The value of all of these types of search should be obvious. So
too should be the fact that they are presently impossible.

5. LINKS AMONG SUBJECTS

Ranganathan discussed various ways in which a document might link two

or more subjects (see Marcella and Newton, 1994, 27). The first type of link
is the sort of causal influence discussed above. When phenomena were
discussed, it was shown that causal links could be captured with the use of an
arrow to join the notations for the relevant phenomena. This solution bears a
close similarity to that of Ranganathan, with the arrow serving to indicate
that a causal link is being addressed.
Another type of link between subjects occurs if understanding in one
subject area is applied to another. This sort of connection should be captured
with respect to the theories and/or methods used in the document. That is, in

10
This is as true of classification as any other method. Bowker and Star (1999) note that
those classifying in different realms face similar problems but do not interact; the result is
“pluralistic ignorance.”
7. Classifying Scientific Documents 229

applying “understanding” from one area to another, scholars would almost

always be applying the theory and/or method commonly employed in one
area to another. When economists tackle questions in the realm of sociology
(say, decisions within the family) or political science (say, voting behavior),
they almost always apply rational choice theory and/or statistical analysis.
This sort of link can be captured with much greater precision in the
classification scheme advocated here.
The final way (though Ranganathan broke this into three components) in
which subjects might be linked involves the discussion of similarities or
differences among subjects. Such works always pose a problem for
classification, as it is not clear which subject should be given priority for
shelving purposes. In principle this problem is not too severe as long as the
subject index provides guidance with respect to each subject compared. The
only advantage the system developed above would provide here is a more
precise and exclusive classification of subjects, whether these are
phenomena, theories, or methods.
Dahlberg (1994) has identified five types of interdisciplinary research,
each of which poses difficulties for library classifications. Some of these
reflect the sorts of linkages identified by Ranganathan. Yet Dahlberg also
points to the fact that some disciplines – such as education – are informed by
many other disciplines, and some topics of study – such as energy – are
analysed from diverse disciplinary perspectives. In such cases, present
classificatory systems inevitably classify related works in a dispersed
fashion. The proposed scheme would not.

6. TYPE OF WORK, LANGUAGE, AND MORE

Having identified subject matter, theory type, and method, the proposed
scheme could then proceed to code for type of work (encyclopaedia etc.),
language, time period, and geographical area (as necessary) in a manner
similar to that of other classification schemes. [An additional notation will be
suggested for survey articles and meta-analyses below.] Classificatory
notation for different editions/versions of the same work could also be
considered (Smiraglia, 2001). The introduction of novel classifications of
theory and method above will of course further disperse documents covering
common times and places. But this is a small price to pay to allow scholars
and students to locate documents by theory and method. Marcella and
Newton (1994, 56-8) note the huge advantages of developing a common
coding for such characteristics as time and place (a practice that has been
followed to some extent by all large systems of classification); the proposed
scheme extends these advantages into the realm of theory and method.
230 Chapter 7

7. CRITIQUES OF SCIENTIFIC PRACTICE

Some works are critiques of scientific practice. They point to what are
believed to be inherent limitations or biases in the scientific enterprise (see
chapter 5). While scientists are infamous for being unaware of a variety of
such critiques, many practising researchers will (or at least should) have
interest from time to time in particular types of such critique. Students and
the general public will also have an interest in pursuing certain critiques of
scientific practice.
Existing discipline-based systems of classification tend naturally to
disperse such critiques widely. Some will be housed with general works on
(natural) science. Some will be classified as sociology of knowledge. Some
will be treated as history of science. Some might be viewed as epistemology
or philosophy of science. And those that apply a particular critique to a
particular subject will likely be shelved with that subject. Interdisciplinary
scholars may carry certain critiques across disciplinary boundaries, but even
they will likely only be familiar with a subset of the applications of a
particular critique. This dispersal would not be so problematic if a common
classification of such critiques had been developed, so that the subject
catalogue could guide users to appropriate documents wherever these might
be housed. But, as with theory and method, such a common system has yet
to be employed.
Of course, a typology of such critiques was developed in chapter 5. As
with the typology of theory types, this provides a very useful mechanism for
classifying critique-based documents. However, as was noted in chapter 5,
each element in this typology represents a causal link toward the
phenomenon ‘science’. While general critiques might readily be classified in
terms of such causal links, some special coding would be necessary for
documents applying critiques to a particular theory, method, phenomenon, or
causal link. Such documents should be housed with the primary theory,
method, phenomenon, or link discussed, but these (and general critiques)
should be readily discovered by searches by critique. In all cases the coding
of scientific critiques would follow the other relevant notation. It might be
useful to put this notation in parentheses both to avoid confusion and to
signal that the work is a work of critique.11
At present, if a researcher becomes aware of a particular type of critique,
and wonders if it has been applied to their area of expertise, they have no
good means of proceeding. Those interested in pursuing critical analyses of
a particular field are likewise hobbled. Those with a vague unease about a
particular scientific theory or method cannot now easily check what sorts of

11
The Universal Decimal Classification attempts to classify some works according to “point
of view”. A theoretically derived typology of critical points of view is provided here.
7. Classifying Scientific Documents 231

critique have been levelled against it. The system of classification of

critiques outlined here, in conjunction with the above classifications of
phenomena, theory, and method, would render all of these searches not only
feasible but straightforward.

8. CLASSIFYING PHILOSOPHY

As noted in Szostak (2000) and (2003), and briefly in chapter 5,

philosophical concerns often exist on a plane above the scientific study of
causal links. Librarians would undoubtedly want to maintain a class for
general works of philosophical interest. Arguably, though, philosophical
evaluations of particular theories or methods are best placed with scientific
analyses of these, rather than in a separate part of the library. More
controversially, philosophical (primarily ethical in this case) evaluations of
particular phenomena or causal links might also be better placed with the
relevant scientific material than with other works of philosophy. Just as with
critiques of scientific practice above, a notation that signified philosophical
analysis is desirable: this too should be at the end and in (a different type of)
parentheses. In any case, the subject index should identify such
philosophical works when searches of relevant phenomena, links, theories, or
methods are performed. Likewise, subject searches of particular
philosophical theories should produce all works applying that theory
wherever they might be housed.
Five broad classes of ethical analysis were identified in chapter 5. These
in turn reflect the five types of decision-making discussed in chapter 3. Each
of the five types of ethical analysis has been advocated by philosophers
across diverse times and places. Some key sub-classes of these types of
ethical analysis were identified by Szostak (2004). Most philosophical
schools, and many complex philosophical arguments, borrow from more than
one of these philosophical approaches; works can still be classified in terms
of the dominant approach. It is possible, then, to provide a classification of
ethical analyses (theories) which has a theoretical basis and which places
each type of ethical analysis in but one place in the classification. As with
the classification of scientific theories, this would classify similar approaches
with different names together, while separating different approaches
operating under one name. Similar efforts could likely be undertaken in
other philosophical areas such as epistemology. Note that, while notation for
philosophical theory would come after the notation for scientific theory,
these two types of theory are otherwise treated in a quite similar manner.
232 Chapter 7

9. EVALUATING THE SYSTEM

The novel elements of the proposed system of classification rely on

familiarity with the 5W questions, the list of phenomena discussed above,
and a list of some dozen methods. Despite this simplicity the proposed
system codes much more information than existing systems. One common
concern with faceted systems is that the notation can be very long and thus
hard to comprehend. Despite adding new elements, the proposed system of
classification remains manageable. The initial classification of a work’s
subject matter in terms of the most important causal link could involve up to
ten symbols if two narrowly defined phenomena were involved, but would
generally involve less. Method requires only a couple of notational spaces
and theory type two or three more. Treatment of time, place, language, and
publication type proceed as with other schemes. Philosophy and critiques of
scientific practice may add a couple of additional notational spaces for some
documents. The total length of the complete notation is thus similar to that of
existing systems of classification.
Hunter (1988, ch. 8) identifies a handful of qualities to be sought in a
classification system. The first is that each document should have a unique
place. This the proposed system achieves to a much greater degree than
existing discipline-based systems. The second is simplicity. As noted above,
this system rests on a handful of key concepts. The third is brevity. The
proposed classification could be described more easily than existing systems.
The fourth is hospitality to new concepts. While this classification system is
based on hopefully exhaustive classifications, they are nevertheless flexible
such that any new discoveries could be added. And there would most likely
be an obvious place for any new elements to be introduced. The integration
of existing fields into new fields, hailed by Dogan and Pahre (1990) and
Collins (1998), bedevils existing classification systems, but would be
handled with ease by the proposed system. An optional fifth quality is
flexibility such that individual users can change the system to suit their
purposes. Note here that some research libraries might find it advantageous
to code first by theory or method. The final quality is expressiveness, so that
notation reflects the structure of the system. The logical structure of the
notation, plus the possibility of often using the first letter of a phenomenon12

12
One downside here is that main categories may not be ordered in terms of similarity: It
might be desirable for Culture and Social Structure to be adjacent, but C and S are
separated by E for Economy, P for Politics, and several others. While the Dewey system
has been criticized for separating Language (400) and Literature (800), there is a general
feeling that order within main classes is much more important than order across main
classes (Hunter, 1988, 76-7, Marcella and Newton, 1994, 35-6). There also seems to be a
sense that there are or should be some forty main classes; the proposed system, even with
the addition of natural science classes, would have perhaps half as many.
7. Classifying Scientific Documents 233

or method as notation, together suggest that this system could excel in terms
of expressiveness.
Information scientists devote considerable attention to the development of
thesauri, or guides to the terms used in a classification system. They thus
hope to aid users who might search under a synonym for a word used in the
particular classification system. One beauty of the proposed system is that
an overview can be provided in a handful of tables: of phenomena, methods,
theory types etc. Note that these classifications in turn can be derived from
the 5W questions asked in chapter 1. The organizing structure is transparent
enough that any user should be able to find what to search for without having
to guess what precise term to search under. Nevertheless, thesauri should
still be provided, especially with respect to individual theories and critiques
of scientific practice.
Rowley (1992) discusses two ratios by which the effectiveness of a
subject index can be judged. The first, the recall ratio, compares the number
of relevant documents retrieved relative to the number of relevant documents
in the system. The second (and easier to measure), the precision ratio,
compares the number of relevant documents retrieved with the total number
of documents retrieved. By identifying all of the phenomena, links, methods,
theories, and types of critique in a document, the proposed scheme should
aid recall enormously. Likewise, by avoiding ad hocery and a discipline
basis, while aiding recall, it should greatly enhance precision. Researchers
will be able to quickly identify a search that will produce all relevant
documents, regardless of whether these documents were produced within
diverse disciplines or by interdisciplinary scholars. Rowley proceeds to
considerations of the time and effort expended by the user, and ease of use.
Again, by coding in terms of what people wish to look for in a work, ease of
use should be greatly enhanced. Rowley closes by mentioning specificity,
exhaustivity, and degree of error. By basing the system in hopefully
exhaustive classifications, each of which has a specific place for each
element, the system should excel in these respects as well.

9.1 Coping With “Information Overload”

A document classification scheme along the lines above would have

enormous advantages over existing schemes for librarians, researchers,
students, and the general public. By taking one further step, the benefits to
these latter three groups could be greatly enhanced. The classification
scheme makes it easy to identify works by causal link, theory, or method (or
type of critique or philosophical perspective). While this has huge
advantages, it will often mean that investigators retrieve a huge array of
documents. While retrieving too many documents is a less serious problem
234 Chapter 7

than getting too few, it should not be ignored. This problem will arise in any
large library; it will be overwhelming in the emerging set of Internet web
sites which attempt to gather all published work in some area of study (and
even more so in some likely future web site(s) which attempt universal
coverage).13 Note that some 100,000 scholarly papers and 30,000 books or
book chapters are produced annually in the United States alone.
This problem of information overload can only be solved by scholarly
efforts to summarize research. These take two forms. Survey articles
summarize scholarly research on particular phenomena, links, theories, or
(less often) methods. Meta-analyses (chapter 4) look at similar types of
research, such as experiments involving a similar treatment, and attempt to
discern why results may have differed. At present, both types of research are
less valued within the academy than original research. The result is a huge
mass of research with only sporadic and incomplete efforts to synthesize it.
Yet a good survey article spares later scholars from a very difficult
information gathering process (Palmer, 2001; she particularly advocates on-
line surveys with links to works cited). Palmer (1996, 184) suggests that the
main implication of her examination of interdisciplinary research practices is
the key importance of general works such as survey articles, textbooks, and
generalist journals. Willinsky (1999, 165-7) argues that it would be sensible
to devote some fraction of government research funding to such studies.
Palmer (2001) feels that information scientists should participate in the
writing of survey articles, partly because they face greater professional
incentives to do so, but mostly because they will ensure a greater breadth of
coverage. The proposed classification system, especially once adopted by
websites or libraries with reasonably exhaustive collections, could solve the
problem of imperfect coverage: scholars wishing to perform a survey or
meta-analysis could easily locate all relevant documents. They would simply
need the incentive to do so. Greater appreciation within the academy of the
value of such work is one possible source of such an incentive. Greater
appreciation outside the academy of the value of such guides to the wider
literature (and both non-profit and for-profit websites14 might be a powerful

13
Examples include the Educational Resources Information Centre (ERIC) which presently
contains almost a million document abstracts, and hopes to soon have whole articles (note
that a fee is charged for downloading by document), a physics website which receives
35000 hits per day, and various medicine sites which attempt to provide doctors with
access to all research on particular diseases or treatments (see Willinsky, 1999).
14
Willinsky (1999), though somewhat tongue-in-cheek, describes a potential non-profit
organization that would provide a vast storehouse of scientific documents, as well as
commissioning surveys (and more focused research where a particular problem was
identified). Willinsky notes that there is a crying need for especially social science
research to be presented to the public in such a fashion. This would enhance democratic
decision-making, and thus public support for science.
7. Classifying Scientific Documents 235

influence here) is perhaps a more likely force. In both cases, the use of a
classification system which can identify the huge mass of research performed
on (or with) particular phenomena, causal links, theories, or methods, should
highlight the advantages of, and indeed crying need for, works which survey
the literature.
Having recognized the need for scholarly works that synthesize, a
notation for survey articles and meta-analyses should be included within the
classification of “types of work”. Both researchers and the general public
will generally find it useful to read such works first.
The public is often frustrated by conflicting scientific results. Even those
who appreciate that science advances through debate wonder why scientists
do not seem to try very hard to settle disputes (Willinsky, 1999, 31). They
will hardly be heartened to find that scientists need not even be aware of
conflicting results achieved by other scholars. While a better classification
system can alleviate such ignorance, survey articles and meta-analyses can
take researchers an important step toward appreciating the implications of
the research of others. The goal of meta-analyses and review essays is not to
establish a winner among competing interpretations, but rather to outline the
strengths and weaknesses of different pieces of research. It can be hoped
that the scientific community will strive to answer questions raised in survey
articles. Writers of surveys need to overcome the tendency to stress their
favored view and point out only the weaknesses of alternatives. For
questions with public policy implications, the public could be left to judge.
Both writers and readers of surveys of research on particular causal links will
be better able to critique alternatives if they first have an understanding of
the inherent strengths and weaknesses of different theory types and methods.
The typologies developed in chapters 3 and 4 – and document classification
by theory and method – are important here; survey articles on particular
theories and methods would also be of great use.
Both Collins and Pinch (1993, 144-8) and Meltzoff (1998, xv) bemoan
the fact that media reports generally provide only the results of competing
studies. The public, they note, can hardly hope to distinguish among
competing results unless acquainted with some idea of the methods [and
theories] employed by researchers. And as noted above the public must
appreciate that no method is perfect, but neither is any method useless. If the
public had ready access to (regularly updated) survey works that would
compare different studies, both they and the media would likely delve deeper
into scientific controversies.
Oakley (2000, 312-24) celebrates one example of meta-analysis. The
Cochrane collaboration is an international effort to draw together for
purposes of comparison controlled experimental research studies in the area
of obstetrics. The collaboration was motivated by a realization that neither
236 Chapter 7

researchers nor physicians had an easy way of judging the mass of research,
and especially the contradictory results that different studies might generate.
The managers of the collaboration then went on to provide a guidebook for
expectant mothers that described interventions that seemed beneficial, those
requiring further investigation, those with unknown effects, and those with
documented ill effects. These (regularly updated) results have been used
both by health care decision-making bodies and by activist groups. The
collaboration now operates across diverse fields of health science, and
maintains an updated library of research studies. Oakley documents the
emergence of a similar effort in the area of social work, and demands for
these across science (especially in education). Criticisms of the Cochrane
collaboration seem to focus on the unofficial nature of the collaboration, its
exclusive focus on controlled experiment,15 and the questionable value of
some of the research summarized. It would seem that there is little objection
to the principle of such summaries, and widespread appreciation of the value
of such efforts. Note in particular the use made of this endeavor by both
physicians and expectant mothers. As argued above, improvements in
scientific practice can be directly beneficial to the wider public.
Drawing on the experience of the Cochrane Collaboration, the Campbell
Collaboration strives to survey experimental results across human science. It
judges the reliability and validity of studies on particular questions, and then
attempts to combine the results of the better studies. It has unearthed some
11,000 examples of controlled experiments in human science. Early projects
of the collaboration treated the value of driver education programs and
“scared straight” programs for crime deterrence. In both cases, experiments
suggest that these programs are less wonderful – indeed likely harmful – than
have studies using other methods. The analysis in this book would guide
both scholars and the public to take such results seriously, but not eschew
insights to be gleaned from other methods.
The large-scale efforts to survey literature that were documented by
Oakley (2000) and Willinsky (1999) have tended to be organized on a
disciplinary basis. The classifications above would encourage and allow
these and other efforts to span disciplinary boundaries. The surveys would
thus be more comprehensive, and therefore would likely encompass studies
reflecting a wider range of perspectives.

15
Though Oakley herself takes pains to emphasize the many advantages of experiments (and
provides many examples where these provided more accurate answers than casual
qualitative evaluation), she is a believer in methodological flexibility and would likely
approve of efforts both to summarize (at least some) other types of research and to contrast
results across different types of research.
7. Classifying Scientific Documents 237

9.2 Visualizing Argumentation

Buckingham Shum (2003, 186) argues that researchers need answers to

questions like “How does the expert community perceive this theory, model,
language, empirical result?,” and thus need “tools to track ideas and results
in a field.” To this end, he and other contributors to Kirschner, Buchingham
Shum, and Carr (2003) have developed techniques for diagramming
scholarly argumentation. While their focus is on “mapping” narrowly
defined questions, such maps would be entirely consistent with the broader
map of the scholarly enterprise urged in this book. Moreover, the sorts of
diagrams they advocate – for example, Horn (2003) diagrams scholarly
exploration of the question “will computers ever be able to think?” – would
be a useful component of the survey articles urged above. Horn shows how
certain major issues of relevance to this question, and several more minor
issues, are related (through the use of symbols that mean “is supported by”,
“is disputed by”, and so on), and identifies differing points of view. Note
that some of the issues investigated by Horn would be of relevance to other
broad scholarly questions (such as what is consciousness): there would thus
be advantages to showing diagrammatically how the subject of one survey
article was related to another.
Kirschner, Buchingham Shum, and Carr (2003) note that such diagrams
are useful for students: they allow students to readily grasp how different
arguments are related, and thus to see how arguments they develop might fit
within the broader debate. These diagrams are also useful for scholars and
decision-makers. Critically these can grasp (and remember) the entire
debate: otherwise there is a tendency for each individual to emphasize the
part of the larger puzzle with which they are most familiar. And thus key
arguments cannot casually be forgotten, no matter how convenient this might
be. Finally, the mapping helps to de-personalize debate, and establish that
every complex issue benefits from multiple perspectives.
Buckingham Shum has developed software for identifying the key
arguments in a paper. He and other contributors hope to develop software
for classifying different types of argument, and the relations among these.
Obviously, these tasks will be easier if the scholarly community can first
agree on classifications of phenomena, theory, and method. Such software
may also provide an inductive means of better identifying key types of causal
arguments.

9.3 The Opportunity

Information scientists have over the last century or so developed a

handful of comprehensive classification schemes, as well as countless
238 Chapter 7

schemes with narrower purposes. While libraries focussed on one or very

few subjects have sometimes availed themselves of novel classifications,
general purpose libraries – despite the evidence that interdisciplinary
research in particular requires improved methods of classification – have
been very slow to adopt apparently superior systems. This is likely because
of the huge costs associated with reclassifying existing holdings, and
retraining both staff and customers in a new system (Marcella and Newton,
1994, 17, 110). The advent of the Internet, and electronic cataloguing more
generally, provides an opportunity for new classification systems to take hold
in a new market. Notably, subject searches have become much more
common with the advent of electronic catalogues; Svenonius (2000, 46) joins
others in hailing this as a paradigm shift in document classification. Efforts
since the 1980s to develop a protocol for searching across libraries that use
different classification systems (Baeza-Yates and Ribeiro-Neto, 1999, 416)
provide a further opportunity. The Dublin core project
(www.Dublincore.org) is of particular note. It provides guidance on how
various elements (“meta-data”) – title, author, publisher, subject, and so on –
could be coded. It would be relatively straightforward to add to such a
system guidance as to how works might be classified by phenomenon,
theory, method, or critique.
Still, a new system, if it is to be utilised widely, must have some
demonstrable advantages over those, like the Dewey Decimal system and the
Library of Congress system, that have existed and been regularly enhanced
for more than a century. In this regard, Palmer (1996) notes that a more
interdisciplinary scientific enterprise requires more than just tinkering with
existing classification systems. And Svenonius (2000, 69) reminds librarians
of Cutter’s categorical imperative: “the convenience of the user must be put
before the ease of the cataloguer.” While there are costs associated with
change the long-term benefits of a superior classification may be immense.
Hopefully, it has been shown that there are huge advantages to classifying
documents in terms of phenomena, causal links, type of theory, method, and
also type of critique or philosophical perspective as appropriate. Likewise, it
has been argued that it is quite possible to achieve manageable
classifications of each of these elements. To be sure, only the broad outlines
of these classifications have been described. Much additional work would be
required to flesh out the details. The logical structure of each classification
provides a framework in which this detailed work can proceed. And it was
noted that information scientists can often borrow where necessary (while
simplifying) the detailed lists contained in other systems of classification.
Chapter 8
CONCLUDING REMARKS

It is useful to begin this chapter by summarizing the major points made in

the preceding chapters. This is done under four headings: the classifications
themselves; the advantages that flow from these; the insights from
comparisons across classifications; and the various Golden Mean arguments
(see chapter 1) that have both supported and been supported by these efforts
at classification. Three broad questions are then addressed: “Why should
science be believed?”; “How might science be improved?”; and “How
should science be taught?” In answering these questions the roles of both
specialized and interdisciplinary science are discussed.

1. THE CLASSIFICATIONS

The core of this book consists of several classifications: of science itself

at first and then of various elements of science. These classifications each
manage to avoid or minimize the various problems that can plague efforts at
classification; these potential pitfalls were outlined in chapter 1. Many of the
classifications in this book are based on or informed by the 5W questions:
Who?, What?, Where?, When?, and Why? The myriad advantages of the
5W approach were also discussed in chapter 1.
The major classifications developed are bulleted in what follows. Less
extensive and subsidiary classifications are italicized in the text. In the order
in which they were presented the classifications are:
x Science itself. Science was “unpacked” into its four constituent
components in chapter 1. These are phenomena/data (including empirical
generalizations; see chapter 2), theories (including hypotheses; see
chapter 3), methods (including tools and techniques; see chapter 4), and
scientific practice.
240 Chapter 8

x Phenomena. A hierarchical list of the phenomena of interest to human

scientists was outlined in chapter 2, and summarized in Table 2.1. The
derivation of a similar classification for natural science was also
discussed, and key elements identified.
x Data. These could be classified in terms of the five human senses. In
chapter 2 a more useful distinction was suggested in terms of the
character of data themselves: humans can observe objects, beings, events,
and (through introspection) attitudes. A classification of the quantitative
character of data was also reviewed. And the four types of epistemic
knowledge were discussed.
x Grand Theories. A tentative list of grand theories currently in vogue in
human science was provided at the start of chapter 3. The list was not
exhaustive, but it was suggested that it would be possible to construct an
exhaustive list for both human and natural science.
x Theory Types. More usefully, a 5W typology of theory types was
developed in chapter 3. Several leading genres of theory were then placed
within this; see Table 3.1. In answering the 5W questions, brief lists of
the six types of agency, three types of “action”, five types of decision-
making, two ends of a generalizability continuum (nomothetic versus
idiographic), and four types of time path were developed. A list of the
five types of causation was reprised, and it was shown how these were
related to the typology of theory types.
x Methods. A list of the dozen methods used by scientists was provided at
the outset of chapter 4. The 5W questions allow the comparative
identification of the key strengths and weaknesses of each method. A
typology of the strengths and weaknesses of methods is summarized in
Table 4.1. That chapter also reprised a list of the four criteria for
identifying a causal relationship, and a typology of mixed method studies.
x (Critiques of) Scientific Practice. At the outset of chapter 5, some major
critiques of scientific practice were summarized. This list is not
exhaustive. A 5W typology of critiques of scientific practice was then
developed.
x Approaches to Ethical Analysis. Chapter 5 also argued that ethical
decision-making, like all other types of decision-making, must rely on
one or more of the five types of decision-making identified in chapter 3.
There are thus five broad types of ethical analysis.
8. Concluding Remarks 241

2. ADVANTAGES OF THESE VARIOUS

CLASSIFICATIONS

To avoid repetition, advantages that flow from the classifications in

general are discussed first. These advantages have usually been noted in
several previous chapters. Advantages that are specific to particular
classifications are then discussed in the order these were developed.
Science, quite simply, advances through the examination of carefully
specified theories with the use of carefully delineated methods, and with
respect to carefully specified phenomena (both those involved in a causal
relationship, and those that affect this) and data. It depends on possible
errors in scientific practice being overcome or limited in effect; this is only
possible if these too can be identified. The classifications in this book will
both aid and encourage individual researchers in precisely identifying their
phenomena, data, theory, and method. This is often not done at present,
especially in human science. When it is done, it is often done in terms of
jargon incomprehensible to the outsider. These classifications will allow the
scientific community as a whole to better place each piece of research in a
broader context, and see how it compares or contrasts with other research.
In particular, by classifying all key elements of science, a strong
argument can be made about scientific concepts: that these are of very
limited (and perhaps negative) scientific value unless they can be precisely
defined in terms of phenomena, type of theory, or method (chapter 2).
The classifications, being exhaustive, provide scientists with a complete
list of the possibilities for scientific investigation. They thus establish that
science is essentially bounded, though the number of research possibilities is
very large. The recognition of boundedness should alleviate one source of
postmodern despair of the possibility of enhanced understanding: that the
world is too complex to comprehend. Scholars should be able to embrace
the complexity of the scientific enterprise without succumbing to despair.
Science is not only bounded but interconnected. An essential unity is
provided to science by the twin facts that causal links connect all phenomena
studied (at least indirectly), and that the full range of theory and method has
a place within the scientific enterprise.
By alerting scholars to what types of science are possible, the
classifications should guide them to investigate all possibilities. Scientific
investigation should not be arbitrarily constrained through ignorance or
prejudice. Scientists can use the classifications to identify links that have
received little attention, or theories or methods that have not been applied to
a particular link despite having unique insights to provide. An appreciation
that all methods, theory types, and data types have differing strengths and
weaknesses should encourage a greater willingness to employ diverse
242 Chapter 8

theories, methods, and data in investigating any scientific question. So too

should a change in gestalt as the various classifications both allow and
encourage scientists to appreciate that their research interests are part of a
larger picture. Scientists should recognize in particular that all methods are
biased in favor of supporting theories that provide similar answers to the 5W
questions. In all of these ways, scientists should be encouraged to recognize
that science advances through a conversation among different viewpoints.
The various classifications greatly facilitate the evaluation of the
strengths and weaknesses of different types of science in terms of exhaustive
5W typologies that can claim a considerable degree of objectivity. In
chapters 3 and 4 the strengths and weaknesses of different theory types and
methods were evaluated. Chapter 5 discussed various critiques of scientific
practice; in chapter 6 these were related to types of science. Chapter 6 also
discussed how the strengths and weaknesses of theories and methods varied
in relative importance across phenomena and causal links. Chapter 2
discussed how some phenomena are more operationalizable than others.
Chapter 4 investigated the advantages and pitfalls of triangulation,
whereby different methods are applied to a particular question, in order,
hopefully, to obtain a more accurate answer. It was noted that triangulation
could also be pursued with respect to theory, phenomena, and researchers.
The classifications provide the basis for a far superior system of
document classification. This would greatly increase the efficiency of the
scientific enterprise, and also enhance public access to scientific
understanding (see chapter 7). In particular, it would become straightforward
to ascertain whether any combination of theory, method, phenomena and/or
philosophical perspective had previously been investigated. This would
improve citation practice, and eliminate the all-too-common practice of
reinventing the wheel.
These classifications would both encourage and support a greater
scholarly effort with respect to survey articles and meta-analyses (see chapter
7). These would be much easier to perform if documents were classified in
terms of these various classifications.
By creating a common vocabulary of phenomena, theory types, methods,
and critiques (and ethical perspectives), interdisciplinary communication will
be facilitated. The various classifications also serve to advertise the
necessity of interdisciplinarity. Indeed, one basic lesson of this book is that
science depends on the interplay between specialized and interdisciplinary
research. Interdisciplinary scholars have a key role to play in tying isolated
pieces of research into a broader whole, both with respect to developing
integrated views of complex problems, and tying the entire scientific
enterprise together. Interdisciplinarians have a central role to play in
identifying the range of applicability of theories – and thus the different
8. Concluding Remarks 243

theories that may speak to a particular link. In doing so, they will often need
to contend with disagreements across communities of scholars. They also
can expose specialized researchers to the benefits of extending their gaze
over a wider range of phenomena, theory, and/or method (informing them of
related research undertaken by other communities), and remind them of the
various biases and errors that might afflict their work. In particular they can
note that communities of scholars often choose methods that support their
theories, and more generally a worldview of mutually supportive elements.
Interdisciplinary scholars can also identify and investigate combinations of
phenomena, theory, and method that have been ignored by specialized
researchers. Systems of document classification that were not grounded in
disciplines would further encourage interdisciplinary interaction. If
communities of specialized researchers were to become more theoretically
and methodologically pluralistic, the role of interdisciplinarity would shrink,
but would remain essential to the scientific enterprise.
The classifications have huge pedagogical value (see below). Students
can be provided with an overview of the entire scientific enterprise. They
will be able to place any topic covered in their coursework within this
broader structure. They will also be aware of the gaps in their education, and
have an idea of where to look to fill those gaps. A curriculum grounded in
these classifications would facilitate the acquisition of critical thinking skills
as well. Note in particular that students would come to appreciate the
strengths and weaknesses of all theory types and methods. In terms of both
skills and perspectives, students would be better prepared for lifelong
learning. And the academy itself would be better placed to respond to
numerous criticisms of a liberal education.
The classifications underpin a definition of “science” itself, which
escapes most of the problems that have plagued previous efforts to define
science (chapter 5). The definition involves the study of phenomena or
causal links through the use of theories that answer each of the 5W questions
and at least one of the dozen scientific methods.
Turning now to advantages of particular classifications, the classification
of phenomena (chapter 2) serves some important purposes in its own right.
It will encourage scholars to display a greater balance in their coverage of
causal links in their research, teaching, and policy advice; at present it is
often too tempting for scientists to focus on the links that they understand
best. The recognition of the number of causal links should extinguish any
sense that one grand theory can explain all; mapping grand theories onto the
list of phenomena should show that all grand theories have more to say about
some causal links than others. Attempting to organize diverse material in
terms of one grand theory or method must necessarily oversimplify; the list
of phenomena serves as an antidote. It shows the value of unpacking
244 Chapter 8

phenomena into constituent phenomena (without embracing reductionism),

and of being clear about what phenomena are being investigated. And it
shows that almost all phenomena both influence and are influenced by all
other phenomena. Most centrally it guides scholars to appreciate that
scientific understanding must be developed link by link (through the
integration of generalizable and link-specific theories); an attitude that is
complementary to the emphasis below on establishing the range of
applicability of theories. As noted above, one key role of interdisciplinary
research is to tie diverse link-based analyses into a coherent whole.
The classification of data (chapter 2) guides scientists to appreciate that
no phenomenon is perfectly operationalizable, though there is a continuum.
Scientists are guided to seek multiple types of data (and given guidance on
the types of data that exist), and warned of the danger of reifying a data
source simply because it is easy to analyze. Scientists should be as
transparent as possible in justifying a particular operationalization. They are
guided to appreciate that no type of data is perfect. And they are alerted to
the twin dangers of either insufficient unpacking, or failing to disentangle the
multiple influences on a particular set of data.
The classifications guide scientists to seek to identify the range of
applicability of a particular theory; especially in human science this
important task is often neglected (see chapter 3). Theoretical progress can
then be defined as any theoretical or empirical work that helps scientists to
do so. The 5W typology of theory (chapter 3) should encourage theorists to
clearly engage each of the 5W questions. Much theoretical debate at present
reflects the fact that some of these questions have not been addressed. The
typology allows scientists to readily identify different theory types within a
particular theoretical camp, and also to identify similarities across theories
that bear different names (a major problematic for interdisciplinary research).
It thus brings a much-needed order to the thousands of theories that exist in
science. In particular, it allows both scholars and students to quickly identify
the key sources of theoretical disagreement. The typology also allows the
word “theory” to be defined better than heretofore. And it guides scientists
to evaluate theory clusters theory by theory – this approach is
complementary to the strategy of evaluating the applicability of a theory link
by link. Theories, notably, should be evaluated in terms of clarity, coherence,
how well they provide answers to the 5W questions, and how supportive
analysis using multiple methods is. The typology also indicates where each
of the five types of causation fit within the scientific enterprise. Of particular
importance, the typology alerts researchers to the existence of theory types
that might shed additional light on particular questions. And it shows that
human scientists must in general have recourse to a wider range of theory
types than natural scientists.
8. Concluding Remarks 245

The list of the dozen methods used by scientists (chapter 4) allows a

simple yet powerful definition of “scientific method.” The 5W typology of
the strengths and weaknesses of methods (evaluated in terms of reliability,
validity, and representativeness; chapter 4) not only encourages
methodological flexibility (at present most methods are applied to only a
subset of relevant questions), but guides scientists to appreciate that they
could and should be more careful in their application of many of these.
Efforts to overcome common flaws in the application of various methods
would encourage a greater recognition of the value of applying multiple
methods to any question. The importance of replication, and comparisons
across studies, was also highlighted. The typology also shows that in general
human science must have recourse to a wider range of method than natural
science.
The 5W typology of critiques of scientific practice encourages scientific
introspection; if aware of the various biases that can afflict them, individual
researchers will be more likely to combat these. In particular, scientists
should be aware of the rhetorical strategies that they and others employ to
buttress their arguments. The typology also suggests various sorts of
institutional reform that would enhance the efficiency of science. By
providing a handy means of organizing diverse critiques, and arguing that
none of these renders science impossible, the scientific community will
hopefully become more open to these various critiques; at present scientists
are too easily offended by extreme statements of particular critiques, and
thus miss out on valuable guidance.
Recognition that there are five complementary and equally valid types of
ethical analysis (chapter 5) allows both individuals and societies to steer a
course between ethical orthodoxy and anomie. Individuals are guided to
respect the views of others, without embracing an “anything goes” attitude.
Individuals are also guided to introspection, in order to ascertain why they do
what they do (and in particular to beware of excuses commonly used to
rationalize unethical behavior). And individuals are guided to seek behaviors
and policies that can be justified in terms of all five ethical approaches.

3. DRAWING CONNECTIONS ACROSS

CLASSIFICATIONS

Further insights are gained by juxtaposing two or more of the above

classifications:
Phenomena and Data (chapter 2). Categories of phenomena embrace
different types of data. Moreover, many phenomena have no direct data
246 Chapter 8

counterpart. Even with respect to phenomena that do have a direct data

counterpart, other types of data may provide important insights. This is
especially the case with respect to causal links. Scientists are thus guided to
employ different types of data; this in turn encourages methodological
flexibility.
Phenomena and Types of Theory (chapter 3). Types of agency map in a
complex manner onto the list of phenomena. That is, most categories
contain multiple types of agent. This supports an argument that disciplines,
at least in human science, should apply more than one type of theory.
Types of Theory and Types of Causation (chapter 3). As noted above, it
is possible to identify for what types of theory each of the five types of
causation is applicable.
Theory and Method (chapters 4,6) Given the similarities in typologies of
theory types and strengths and weaknesses of methods, the use of particular
methods will bias results (generally) toward theories that emphasize the
types of agency, decision-making process, and so on, that the method is good
at investigating. Thus, when comparing theories multiple methods should
generally be applied.
Theory, Method, and Phenomena (chapter 6). The importance of the
various strengths and weaknesses of a particular theory varies according to
the methods used and the phenomena or causal links investigated. A similar
argument can be made with respect to methods. These variations reflect the
degree to which theory, method, and phenomena reflect the same sorts of
agency, action, decision-making process, and so on. Scientists are guided to
ask what light a particular theory or method can shed on a particular
phenomenon or causal link. Nevertheless they are urged to appreciate both
theoretical and methodological flexibility, and to strive to identify the range
of applicability of particular theories and methods.
Theory and Data (chapter 6). Scientists should strive for
operationalizable theories. The best protection against the theory-ladenness
of data is to apply different theories to the same data, and to test theories
against different types of data.
Theory, Method, Phenomena, and Critiques of Scientific Practice
(chapter 6). Particular critiques of scientific practice are of greater
importance for some types of science than others. The perceptual limitations
of humans are of greater importance with respect to phenomena that are
relatively difficult to observe. Some methods leave greater scope for
subjectivity than others. Social biases and power relationships (and ethical
biases) are more likely to be reflected in the study of intentional agents,
though there are important exceptions. Disciplinary biases may be less
powerful in natural science. The need for theoretical and methodological
flexibility is less in those natural sciences focused on non-intentional agents,
8. Concluding Remarks 247

in large part as fewer theories and methods are applicable. Dangers of

overspecialization and emphasis on quantity of publications are ubiquitous,
as is the temptation to use the easiest method. Rhetoric matters everywhere,
but particularly in fields that emphasize style of presentation. The natural
sciences benefit from interaction with technologists more than the social
sciences benefit from interaction with policymakers. But natural scientists
are under greater pressure to investigate questions and disclose or not
disclose results that reflect private profit motives. All sciences could afford
to be less hierarchical, and provide scope for a larger number of, and more
flexibility within, paradigms.
Paradigms (chapter 6). It was argued that specialized scientific
communities could and should be more open to a wider range of phenomena,
data, theory, and method. They should define themselves in terms of subject
matter, and investigate all relevant and feasible theories and methods. This
may require the formation of sub-communities with expertise in particular
theories or methods. Conversation, it was suggested, is always possible
across different types of science. This is particularly important so that
scholars can be aware of how theories and methods have been applied
elsewhere.

4. GOLDEN MEAN ARGUMENTS

It was noted in chapter 1 that extreme positions on many philosophical

questions would bring into question various elements of the enterprise of this
book. Most obviously, if it were accepted that science was either perfect or
impossible, there would be little reason to classify it. However, if one backs
slightly away from extreme points of view, both the possibility and
desirability of the enterprise are restored. It was thus argued that an
appreciation of the value of the Golden Mean – a belief that the truth
generally lies somewhere between extreme positions – would provide a basis
from which this project could proceed. In later chapters, there were many
opportunities to note that various classifications served to justify Golden
Mean conclusions. These Golden Mean arguments, then, serve in general as
both justifications for and results of the classificatory enterprise:
x Science is neither perfect nor impossible (chapters 1 and 5 especially).
x Scientific explanations are influenced by both the real world and by
various sociocultural phenomena (chapter 1).
x Conversation is possible across different paradigms or types of science,
but communication may be imperfect (chapters 1,6).
x There is an external reality, but human perceptions of this are imperfect
(chapters 2,4).
248 Chapter 8

x Both individuals and societal aggregates exert causal influence (chapter

3).
x Science should investigate both intentional and non-intentional agency
(chapters 3,4).
x Scientists must analyze both actions and attitudes (chapter 3).
x Scientists should thus be open to both explanation and interpretation
(chapter 3).
x Human science is neither identical to nor entirely distinct from natural
science; rather it relies on a broader set of theory types and methods
(chapters 3 and 4), and is subject to a different mix of biases and errors
(chapters 5,6).
x Science needs both nomothetic and idiographic theories (chapter 3).
x Science, when dealing with intentional agents, should investigate each of
the five types of decision-making (chapter 3).
x Science should investigate equilibrium, non-equilibrium, and stochastic
time paths (chapter 3).
x Science requires both empiricism and rationalism (chapters 3, 4).
x Science should embrace both qualitative and quantitative approaches
(chapter 4).
x Likewise science should embrace positivistic and interpretivist
approaches (chapter 4).
x Science should examine all phenomena or causal links that can be
operationalized, rather than make an arbitrary distinction between
observable and non-observable (chapters 2, 4).
x Science should use both induction and deduction (chapters 1, 4).
x Science can be (imperfectly) performed in both natural and artificial
settings (chapter 4).
x Scientists should ask the 5W questions, and use ‘How?’ as a check on
these (chapters 1,4).
x No one method is better than all others for all questions (chapter 4).
x Science is neither perfectly objective nor entirely subjective (chapter 5).
x Scientific results may thus reflect a variety of biases, but need not only
reflect these; there are pressures that minimize bias (chapter 5).
x Science has many commonalities with non-science, but also important
differences (chapter 5).
x Science does not display flawless advance in understanding; nor does it
display no advance whatsoever (chapter 5).
x Neither ethical orthodoxy nor ethical anomie is ideal (chapter 5).
x Scientists should treat some statements as “true”, for research depends on
this, but should be ready to criticize any statement in the light of new
information (chapter 6).
8. Concluding Remarks 249

5. WHY SHOULD SCIENCE BE BELIEVED?

It has been argued above that science is neither perfect nor impossible.
Moreover, science has been distinguished from non-science: science involves
clarity with respect to theoretical formulation in the specific sense of
answering each of the 5W questions, careful specification of which
phenomena or causal links are being investigated, and exposure of theoretical
propositions to real world evidence in terms of at least one of the dozen
methods. While biases and errors can still affect scientists, the facts that
scientists are clear about theoretical propositions and seek evidence to bolster
these provides limitations to the degree to which scientific conclusions
reflect bias or error. This is especially the case at the level of scientific
communities: if these regularly confront evidence from a variety of sources,
and are careful in their statement and evaluation of theories, scientific
understanding in the long run should be largely shaped by the external reality
it aspires to represent.
Importantly, trust should not be placed in all who would call themselves
scientists. Those who bury their arguments in incomprehensible prose
immunize these arguments against controversion by the evidence. Those who
state theories such that these can be interpreted in multiple ways achieve the
same end. Those who eschew or misuse scientific methods, or rely
exclusively on an unexamined intuition, likewise shield their ideas from
reality. Only in identifying the boundaries of science, and recognizing the
limitations of certain allegedly scientific practices, can a justification be
maintained for belief in the scientific enterprise more broadly. Scientists
cannot command trust simply because they are employed in a university or
laboratory; both they and the communities to which they belong must act in
accord with the definition of science in order to deserve the public’s trust.
This does not, of course, guarantee that they will be right. Even scientists
and scientific communities pursuing the best of scientific practices will often
be wrong. This book has urged scientists to a “web of belief” approach
(chapter 6) in which confidence is placed in statements that seem reliable,
but where scientists are always ready to question all statements in light of
new evidence. Humans are not omniscient, and must make their best
judgments on the basis of inherently imperfect evidence. Only by seeking a
variety of evidence can confidence be increased in a particular proposition.
The public should neither abandon trust in science nor accept every scientific
pronouncement as fact. Scientists should be careful of claiming a greater
confidence than the evidence justifies. And the scientific community should
strive to evaluate all relevant information in reaching conclusions. Ideally,
the public will have access to such evaluations, so that they can make their
250 Chapter 8

own judgments as to the confidence with which any proposition should be

rewarded.1
Awareness of the limitations of every theory and method should guide
scientists to be careful of their conclusions. But nor should they embrace
dismay, especially given that different theories and methods exhibit different
limitations. By using multiple theories and methods (and types of data),
confidence can be greatly increased in any scientific statement. Still, there
are undoubtedly limits to how much humans can ever understand of the
world. The analysis in various chapters suggests, moreover, that these limits
may be reached more quickly for some scientific questions than others.2

6. HOW CAN SCIENCE BE IMPROVED?

The efficiency of scientific practice, the rate of scientific advance, and the
public’s faith in science can all be enhanced by a variety of reforms to
scientific practice. As stressed above, scientists should carefully delineate
the phenomena, data, theory, and method employed in their investigation,
and ideally do so in a manner that is easily comprehended beyond their
community: the classifications both aid and encourage this practice. In
particular, scientists are urged to carefully define “concepts” in terms of
phenomena, theory, or method. Scientific communities should employ a
wider range of theories and methods, analyze a wider range of types of data,
and perhaps draw on multiple ethical perspectives in their approach to the
analysis of any phenomenon or causal link. By recognizing that all
phenomena are causally interrelated, they should be guided to widen their
gaze and include a broader set of phenomena in their analysis. They might
in particular choose to investigate links that have been relatively ignored by
the scientific enterprise. The scientific enterprise as a whole should seek to
understand every causal link. Scientists should seek the range of
applicability of all theories, and recognize that multiple theories may have

1
Cussins (2000, 354) reviews the advantages and disadvantages of public scrutiny of
science. Scientific curiosity should not be constrained to pursue only those topics
suggested by governments or public opinion. Yet given the authority that at least western
societies cede to science, it is desirable that the public has some oversight capability to
ensure that scientists do not pursue self-interest or other biases unhindered.
2
It is noteworthy that postmodernists, who tend to advocate the pursuit of idiographic
research, nevertheless tend also to state their critiques of scientific practice at a general
level. They thus risk erroneously extrapolating from areas in which science is particularly
difficult. This may in part explain the popularity of postmodernism in the Humanities.
Note also that the approach outlined in this book allows the sort of detailed research
preferred by postmodernists to be tied into a big picture, without reliance on the grand
theories that they are justifiably suspicious of.
8. Concluding Remarks 251

something to say about a particular causal link.3 Recognition that science is

imperfect, and that every theory, method, and type of data is also necessarily
imperfect, should drive scientific communities to broaden their approach.
This does not mean, of course, that every scientist ceases to specialize.
There are huge advantages to specialization. The scientific enterprise can
have these benefits, while avoiding the pitfalls, if scientific communities
encourage the application of different theories, methods, and data, and also
encourage regular efforts to synthesize the literature so that specialists are
aware of the implications of alternative lines of research. Previous chapters
have discussed how such a research program can be embraced without any
diminution in scientific standards.4 If communities of specialized researchers
would define themselves in terms of an expansive but necessarily
circumscribed set of phenomena (which might, of course, evolve over time),
sub-communities could develop expertise in each relevant theory or method.
These would benefit from (and support) efforts of interdisciplinary scholars
to connect the use of theories and methods across communities, but would
work with others in their community – some scientists in the community
would need to have expertise in multiple theories or methods – to develop
the best understanding of the causal links in question.
Scientific communities tend to be inherently conservative. They can point
to evidence of scientific progress as justification for “staying the course.”
The argument here is not that science is presently useless, but rather that it
can be made much more productive. Those who benefit from the status quo
may nevertheless be suspicious of reforms. In chapter 5, but also elsewhere,
it was argued that scientists should be self-conscious and self-aware. They
should be aware of the possible biases that may afflict their work, and strive
to overcome these. By providing a manageable list of critiques, and arguing
against extreme versions of these, it is hoped that a greater degree of
introspection will be encouraged. Introspection is not only important for

3
Longino (2000) surveys many cases in biology where several partial explanations of the
same result exist: “the variety of causal pathways to a similar end forces an emphasis on
one aspect of a causal process …” Scientists should indeed try to understand particular
causal mechanisms both in isolation and in concert. Longino notes, though, that many
biologists use “methodologies that of necessity show only what they are designed to show”
[a danger noted in chapter 4], and then evaluate results by criteria that are also biased [a
danger addressed in chapter 5].
4
“Transdisciplinarity” is a term favored in Europe. It has diverse interpretations, some of
them quite close to North American concepts of interdisciplinarity. For some it implies
doing away with the very idea of expertise, in recognizing that any individual at best
knows only a piece of a larger puzzle. It would seem that it would be more accurate,
though less controversial, to speak of rethinking the definition of expertise. The tendency
of transdisciplinarians to favor integration in terms of themes is also troubling; this can
look too much like the all-too-frequent attempts to unify science via the route of grand
theory.
252 Chapter 8

improving an individual’s practice of science, but also for earning their

support of institutional reforms – such as encouraging methodological
diversity or comprehensive survey articles – that would further enhance the
productivity of science. The vast majority of scientists see themselves as
primarily involved in the selfless pursuit of knowledge; many of them,
however, all too readily rationalize practices that limit the rate of advance of
understanding. The greatest academic sin may be insecurity, for it forces
academics to cling to both theories and institutions that they know well. If
scientists were to accept that science is imperfect but possible, they should
become more open to beneficial changes.
At present, most scholars are familiar with only one or two methods.
They can thus too easily dismiss evidence from other methods. It has been
too costly in the past to acquaint all scholars (or students) with the strengths
and weaknesses of all methods. Moreover in the case of methods, and
especially theory types, it was not even known that such an enterprise was
possible. The benefits for scientific teaching will be addressed below; the
point to emphasize here is that there is no longer any excuse for not exposing
future scholars to an overview of the strengths and weaknesses of various
theories and methods. Scholars familiar with this material should not
casually dismiss evidence from methods with a clear relevance to a particular
question. And they should be more aware of the potential flaws in their own
favored method, and strive to limit these in practice. They may even be
guided to employ multiple methods themselves: the effect should be even
greater in future generations as it becomes common to gain exposure to
multiple theories and methods as students. Scientists who are intimately
familiar with the classifications should also recognize situations in which an
appeal to other theories, methods, phenomena, or data would be useful.
From these general lessons flow a number of more specific proposals for
the improvement of scientific practice, such as the argument for the
relaxation of paradigms (chapter 6), the suggestion that social scientists
interact more with policymakers (chapter 5), and the suggestion that science
be less hierarchical and more cooperative (chapter 5). Indeed, virtually all of
the advantages of classifications that were listed above can be rephrased as
proposals for the improvement of scientific practice.5 Likewise, science

5
Bunge (1998, 45-6) lodges several complaints against scientific practice in the area of
social psychology: the study of individuals is not integrated with the study of societies,
replication is rare, standards for judging research are poorly articulated, hypotheses are
often unclear, and important questions [causal links] are often eschewed for less important
and often vague questions, certain methods with obvious applicability are ignored,
contradictory evidence is suppressed, jargon is excessive, and questions of the validity and
generalizability of results are ignored. This book has argued against all of these practices.
Likewise, Kincaid (1996, 261-2) proposes several improvements to social science practice:
social scientists should seek complementarity across theories, use experiments more,
8. Concluding Remarks 253

would benefit from an understanding of precisely how each classification is

related to others, and an appreciation of the various Golden Mean arguments
contained in this book. Finally, the broad lessons drawn above are also
supportive of some reforms that have received only passing mention in these
pages, notably the argument that drawing more women and minorities into
science might serve to balance certain subconscious biases.
It should be emphasized that while the classifications developed in this
book can be used to improve scientific practice, they do not themselves limit
scientific exploration in any way. These classifications are intended to
embrace every possible method, theory type, phenomenon, critique, and
philosophical approach (and if any have been missed they could be
integrated into these classifications). One of the dangers inherent in
classificatory efforts is that important elements will be ignored; exhaustive
classifications cannot be guilty of this sin. Another danger is that
classifications will arbitrarily lump disparate elements together. The logic of
the 5W typologies minimizes this danger. So too does the logic behind the
five types of decision-making. This, as well as lists of phenomena and types
of data, benefit also from a great deal of scholarly consensus. The
classifications, then, provide a very accurate guide to scientific possibilities.
Rather than constraining scientific inquiry, these classifications should
encourage a much greater diversity in forms of scientific inquiry.
While it would be a mistake to limit the substance of scientific inquiry, it
has been argued that style should be constrained to be the servant of
substance (chapter 3). To be sure, style not only plays an important rhetorical
role but can contribute to understanding, especially in areas such as the study
of art. However, if style swamps substance the reader is unclear about what
theory, causal link, and even method or data are being discussed. Science
will advance sluggishly at best in such an environment. Shipman (1988) thus
lists clarity as a fourth criteria for judging scientific research – along with
reliability, validity, and representativeness (see chapter 4) – for without
clarity scientists will be unable to evaluate the other three. This book has
focused on one type of clarity, insisting that researchers be as clear as
possible about their theory, method, phenomena/links, and data. In addition,
there are various “sins of expression” that can limit comprehension of any
research report. Valiela (2001, 108-25) discusses a lengthy but non-
exhaustive list of these, including excessive use of jargon, inaccurate use of

investigate ceteris parabis assumptions, test competing hypotheses, express theories

clearly, recognize data weaknesses and seek new data sources, not assume that correlation
means causation, specify causal claims clearly, move beyond abstract grand theory, not
confuse heuristic models with explanation, and not accept functional arguments
unquestioningly. This book places these diverse proposals in context and suggests how
many could be achieved.
254 Chapter 8

words, using long words when shorter words would suffice, imprecise
phrasing, and poor paragraph structure. Some of the time these practices will
reflect sloppiness on the part of the researchers. Other times, they reflect a
desire – whether subconscious or conscious – for obfuscation. “It is a far
more difficult task to express complex thoughts in simple language than to
express simple thoughts in complex language” (Halperin, 1997, 51).
Scientists and scientific communities can strive to do better.
Finally, what of disciplines? The word has not often been uttered in these
pages (which may itself be symbolic of the fact that disciplines are not an
essential characteristic of the scientific enterprise). By suggesting that
scientific communities be open to multiple theories, methods, causal links,
data, and ethical perspectives, it has nevertheless been argued, at the very
least, that disciplines should be much less insular than at present. In
suggesting that the scientific enterprise needs more, and more flexible,
paradigms in chapter 6, a further argument was made for the redesign of
disciplines. And at various points in this book, many advantages that would
flow from increased communication across communities of scholars have
been noted. The scientific enterprise requires the exercise of multiple forms
of expertise – by phenomenon, theory, method, data, and perhaps ethical
perspective – and needs to operate through communities of scholars that
share a certain basis of understanding. The ideal form of such communities
would seem to be much more flexible than is the case with present
disciplines. And they should be much more open to the exchange of ideas
with other scientific communities than is presently the case, especially in
human science.6

6.1 Interdisciplinarity

While specialized scientists should be open to a wider range of

phenomena, data, theory, and method, there will nevertheless always be
scope for interdisciplinary scholarship.7 Put simply, the scientific enterprise
as a whole depends on both specialization and synthesis. There is a natural
tendency for most scientists to focus on one or the other, especially at

6
“A second cultural function of disciplines is that of preventing knowledge from becoming
too abstract or overwhelming. Disciplines legitimate our necessarily partial knowledge.
They define what it is permissible not to know...” (Abbott, 2000, 130). But scholars can
have the advantages of specialization without ignoring the big picture. And scholars should
not mislead themselves about their ignorance. Szostak (2003) discussed how human
science disciplines might be better organized: each could focus on links to and from a
particular category of phenomena but remain open to diverse theories and methods.
7
Many disciplines are at present fragmented into subfields with limited communication
between them. “Interdisciplinary” scholarship as discussed here may thus at times be
necessary even within disciplines (see Dogan and Pahre, 1990).
8. Concluding Remarks 255

particular points in their careers. Among the key advantages of the

classifications developed in this book is that they both highlight the
importance of and facilitate both specialized and interdisciplinary
scholarship (see above).
How should interdisciplinary scholarship proceed? A twelve-step process
for interdisciplinary scholarship was developed in Szostak (2002). This
process was developed in part inductively, by combining the diverse
suggestions that emerged from a debate about the nature of
interdisciplinarity, and in part by building on processes developed
deductively by previous scholars, notably Bill Newell and Julie Thompson
Klein. Of course, individual interdisciplinarians, and even teams of
researchers, will not be able to perform every step for each research project
that they undertake. The twelve-step process nevertheless warns them of the
costs of eschewing any particular step. More centrally, it provides guidance
to communities of interdisciplinary scholars on what they should collectively
strive to achieve. Indeed, just as previous chapters developed classification-
based definitions of scientific theory, scientific method, and science itself,
the twelve steps could provide a superior definition of interdisciplinarity.
Space prevents a detailed description of each step here. It is, however, useful
both to list these steps and to discuss briefly how the classifications
developed in this book aid the performance of virtually all of these.
1. Start with an interdisciplinary question. This means a question that
involves phenomena, data, theories, and/or methods from more than one
scholarly community. While these questions can emerge in a variety of
ways – deductively and inductively, rationally or intuitively (Szostak,
2002a) – one important strategy has been outlined in this book: the
classifications can be utilized to identify combinations of theory, method,
phenomena, and data that have either been ignored by the scientific
enterprise, or treated by multiple communities of scientists.
2. Identify the key phenomena involved, but also subsidiary phenomena.
Without comprehensive classifications, such as were outlined in chapter
2, it is all too easy for interdisciplinary scholars to neglect relevant
phenomena. The temptation is to continue to address the same
phenomena as previous scholars.
3. Ascertain which theories and methods are particularly relevant to the
question at hand. Again, without comprehensive classifications in place,
it is all too easy to continue to ignore relevant theories and methods that
have been neglected in previous research.
4. Perform a detailed literature survey. As discussed at length in chapter 7,
present systems of document classification are particularly ill-suited to
interdisciplinary research. Superior systems of classification could be
grounded in the classifications developed in this book.
256 Chapter 8

5. Identify relevant disciplines and disciplinary perspectives. This book has

touched on the vexed question of disciplinary perspectives (worldview)
more than once. This key insight of interdisciplinary scholarship – that
communities of scholars choose a mutually reinforcing set of theory,
method, phenomena, and philosophical perspective (at least) – has been
borne out in this analysis. In particular, it was noted how certain methods
will be biased toward support of certain theories (chapters 4,6). Not
surprisingly, communities of scholars tend to favor methods that support
their theories, and moreover choose to emphasize phenomena, data, and
even ethical perspectives that are congruent with these. Interdisciplinary
scholars have an important role to play in showing how disciplinary
perspectives limit research. Yet disciplinary perspective or worldview has
proven a frighteningly vague concept (Szostak, 2003a). As with concepts
more generally (chapter 2), worldview needs to be defined more precisely
in terms of phenomena, data, theory, and method. The identification of
the key elements of science in chapter 1, and their exploration throughout
the book, should aid interdisciplinary researchers immensely in
identifying worldview both in general and with respect to particular
communities of scholars.
6. If some relevant phenomena, theories, or methods have received little or
no attention, the researcher should try to perform or encourage the
performance of such research. In performing such research, the
interdisciplinarian behaves much like the specialized researcher. As has
been seen, specialized researchers also benefit from these classifications
in a variety of ways. Note again that the very identification of neglected
areas of research is greatly facilitated by the classifications.
7. Evaluate the results of previous research. As noted above,
interdisciplinary researchers cannot just take specialized research for
granted. Familiarity with the strengths and weaknesses of different
theories and methods, and the biases and errors that are likely to afflict
these, can be invaluable here. The interdisciplinary researcher should
also ask if important phenomena have been excluded from previous
analyses, and will benefit from comprehensive lists of phenomena in
doing so.
8. Compare and contrast results from previous research. The first task here
is to ensure that differences of opinion are not just semantic: if different
scholars define terms in different ways, they may produce seemingly
different results as a result.8 The classifications provide a base for a

8
Salter and Hearn (1996) identify three key problems faced by interdisciplinary research
[and teaching]: a translation problem as different disciplines develop their own
terminology, a language problem as different disciplines attach different meanings to the
same word, and a reception problem as discipline-based journals eschew work which
8. Concluding Remarks 257

common scholarly vocabulary; in the meantime they provide a base from

which to evaluate the terminology used in different scholarly
communities. If different results reflect more than semantics, it should
then be asked whether different scholars have actually investigated
different aspects of a common problem (as when economists focus on
economic repercussions of a trade deal, and political scientists on
political effects). The interdisciplinary researcher needs then to place
specialized research on a map of causal links (see chapter 2). The
typology of theory types will also be helpful here, for some research may
have identified different types of agent, action, and decision-making
process from others.
9. Develop a more comprehensive integrative analysis. The recognition that
the goal of science is to identify the range of applicability of different
theories is important here: rather than seeking one all-encompassing
grand theory, the interdisciplinary researcher is guided to seek a complex
amalgam of different theories.9 Some of these might be nomothetic,
others idiographic. These theories will be best justified by recourse to
multiple methods. The typologies of theory and method can guide
researchers as to which of these are best suited to different aspects of the
question at hand. The interdisciplinary researcher may also be guided by
the classification of phenomena to add phenomena to particular theories.
10. Reflect on the results of integration. Having a list of potential biases to
refer to can be invaluable as the researcher interrogates their own results.
The twelve-step process summarized here provides a further check, in
causing researchers to reflect on the likely costs of the steps omitted.
11. Test the results of integration. Are there implications of the integrative
framework that can be empirically evaluated? The typology of strengths
and weaknesses of method can be helpful here. The researcher should be
careful of choosing only methods likely to be supportive.
12. Communicate the results. Clarity of terminology is again critical, both to
communicate to diverse scientific audiences and to the public.

borrows heavily from other disciplines. The acceptance of a common list of phenomena
would go some way toward solving the first two problems (and suggestions of better
scientific standards in chapter 5 addresses the third). There are literally thousands of
definitions of the word “culture” in the literature; not one is as precise as that provided by
Table 2.1 (see Szostak, 2003).
9
The interdisciplinary researcher should nevertheless be open to the possibility, stressed by
Newell (2001), that there may be emergent properties not obvious in any of the earlier
theories.
258 Chapter 8

6.2 Evaluating Faculty

Along with urging introspection by individual scientists, this book has

also urged institutional reforms that would give the correct incentives to
scientists. Glassick et al. (1997) have also engaged the question of how best
to reward scientists (or at least those employed by universities). Notably,
they are motivated by a concern that at present scholars are rewarded almost
exclusively for the “scholarship of discovery”, while the scholarships of
“integration”, “service”, and “teaching” are severely undervalued. With
respect to integration, they emphasize the need to overcome the isolation and
fragmentation of disciplines, and to make connections within and across
disciplines; integration allows bits of specialized insight to become more
meaningful to non-specialists, and brings new insights to specialized
research. This book has argued for the importance of the scholarship of
integration, and also for one important sort of service: interaction between
scholars and policymakers. The ways in which these classifications can
enhance teaching effectiveness will be discussed below.10
Glassick et al. (1997) develop six criteria by which they hope that all four
types of scholarship could be evaluated. The first involves the statement by
the scholar of clear, realizable, and important goals. This book too has urged
clarity at many points, and has also encouraged scholars to see how their
research fits into a larger enterprise. The second criterion is adequate
scholarly preparation, including the understanding of existing scholarship
and the development of necessary skills. The classifications developed in
this book facilitate the mastery of relevant literature and also a broad
approach to the definition of “necessary skills”. The third criterion is the use
of appropriate methods. This has been discussed with respect to both
scientific discovery and the scholarship of integration in chapter 4; the best
teaching methods will be discussed below. The fourth criterion is the
achievement of significant results. These classifications should enhance the
collective productivity of the scientific enterprise, and in particular stop
scientists from reinventing the wheel. The fifth criterion is effective
presentation. This again involves clarity; an appreciation of rhetorical
strategy was encouraged above. The final criterion is reflective critique,
which involves critical evaluation and response to the criticisms of others.
Introspection, a “web of belief” approach, awareness of the biases that can
affect scientists, openness to diverse viewpoints, and an appreciation that
science advances through dialogue, have all been urged in this book;

10
Glassick et al. (1997, ix) note that the emphasis on research in modern universities
encourages excessive specialization. Those focused on teaching students are more likely
to see the need to tie diverse bits of understanding into a coherent whole.
8. Concluding Remarks 259

reflective critique was the penultimate step in the process of interdisciplinary

scholarship outlined above.
Hopefully, it has been shown that the arguments in this book are
complementary to those advanced by Glassick et al. (1997). The
improvements to scientific practice that have been advocated here would
cause scientists that embraced them to fare better under the criteria advocated
by Glassick. In turn, adoption of such criteria by universities would
encourage at least some of the practices that have been endorsed in this book.

7. HOW SHOULD SCIENCE BE TAUGHT?

This book will hopefully be of interest to a wide range of academics.

Many of these are sceptical of suggestions that universities and colleges
should change the way that they teach and/or increase the importance of
teaching in their evaluation of faculty. They worry that such proposals are
only put forward by those who have been unsuccessful at research, and that
teaching is inherently more difficult to evaluate than research. As with other
extreme arguments engaged in this book, there is a kernel of truth, but also a
danger of unquestioned acceptance. Scepticism should not prevent reasoned
efforts to enhance the student learning experience. In any case, the focus here
will be on redesigning curricula rather than pedagogy, though it should be
noted that the revised curricula advocated here would encourage diverse
pedagogical approaches. Not only is there great value in enhancing college
curricula, but it is often as easy to measure curricular reform initiatives as
research outcomes. Moreover, as we shall see, the curricular reforms
envisaged here will enhance the ability of scholarly institutions to assess how
much their students have learned.
The curricular reforms that flow from the analysis of this book can be
succinctly stated (see Szostak, 2003a, for more detail). In their first year of
college, students should be exposed to the classifications of phenomena,
data, theory types, methods, critiques, and approaches to ethical evaluation
that have been discussed in this book. In the case of theory types and
methods, the students would be introduced to the strengths and weaknesses
of different methods and theory types. [Note that some of this material might
be usefully introduced in high schools; it would, in particular, provide a
useful structure for often incoherent social studies curricula.] This material
need not be memorized all at once. It should be reinforced throughout a
college education: instructors in introducing a new topic should review what
phenomena, data, theory types, methods, critiques, and/or philosophical
perspectives are to be engaged. In the case of theory and method, instructors
260 Chapter 8

should address how the strengths and weaknesses of these come into play for
the question at hand. Note that instructors could do this by making only
minor changes to existing course content (though the classifications may
encourage them to engage a wider range of phenomena, theory, and method
in their teaching). The students could thus master the classifications in the
course of applying them throughout their college career. At the same time,
they would be able to draw connections within and across courses, for they
would be able to see where each topic covered fits within the broader
scientific enterprise. Along with the classifications developed in earlier
chapters, students could usefully be exposed to the twelve-step process for
interdisciplinary research outlined above, for this process provides guidance
to any individual or group on how to approach complex issues. A capstone
course could draw students from diverse areas of specialization together: a
focus on public policy issues such as environmental protection would be
particularly useful here. Students could each provide insight from their
individual areas of expertise; other students could use their understanding of
the strengths and weaknesses of different theories and methods to analyze the
views of others, and work toward synthesis.11 The curriculum in general, and
capstone in particular, would prepare students for lifelong learning: they
would be able to place any piece of information within a complex organizing
structure, they would be aware of those elements of the scientific enterprise
that they had not engaged (it is all too easy at present for students to
exaggerate the comprehensiveness of their education), they would know the
full set of theories, methods, and so on that they might wish to apply to any
novel problem, they would have a firm basis from which to critically
evaluate the arguments of others, and they would appreciate that no
argument is perfect but that some are better than others.
It might be thought that exposing students to the full range of method,
phenomena, types of theory, and ethical perspectives would be either
impossible or inordinately time-consuming. However, these important
elements – hereafter referred to as comprehensive elements, for they provide
students with a comprehensive overview – can be integrated readily into
existing curricula. Those who design general education curricula often
complain of a tension between a desire to cover many topics, and a wish to
cover certain topics in depth. The comprehensive material contained in the
classifications allows curricula to achieve maximum breadth – students are

11
One oft-stated goal of General Education is for students to gain an appreciation of different
disciplinary perspectives. As noted above, it is often difficult to define what this means.
Most of what is meant can be captured with respect to differences in theories and methods
used, phenomena studied, and ethical approaches applied. Note that this approach would
serve to highlight the advantages and disadvantages inherent in disciplinary specialization.
8. Concluding Remarks 261

exposed to the full set of theory types, methods, phenomena, and so on – at a

minimal cost in instructional time. Students will still have scope to pursue
specializations; the comprehensive material will allow them to better place
any area of specialization in broader perspective, and draw connections with
other areas of study. General education programs can themselves differ in
the emphases they place on particular topics; students transferring across
institutions (an increasingly common practice) would nevertheless benefit
from a common grounding in the comprehensive material.
Too often, general education programs are designed by human scientists
– and especially humanists12 – with limited participation from natural
scientists. The comprehensive material, by providing a common structure
into which both natural and human science understanding can be placed,
should encourage the participation of natural scientists. Students would
benefit from seeing how certain theories and methods are applied to the
study of natural science phenomena. The comprehensive material also
provides a venue for cooperation among academic and student affairs
personnel. The latter wish to enhance the study skills of students, and this
involves showing them how to organize information and evaluate arguments;
the comprehensive material is well suited to that task.
It would be a useful exercise in the evaluation of general education
programs to survey the range of theory, method, phenomena, and ethical
arguments covered. If the program is found to be narrow in at least one of
these respects, this would provide a good starting point for curricular reform.
If the existing program does cover a wide range of theories, methods,
phenomena, and ethical perspectives, the comprehensive material outlined
above can be integrated in several ways.
Some courses might be created to deal with the comprehensive material.
A course on methods that compared the pros and cons of different methods

12
Kervin (1997) bemoans several problems in Humanities education: the tendency of modern
critics to read texts only to expose “villains”, not to excite the imagination; the tendency to
argue that knowledge is inherently subjective and ambiguous; and the emphasis of many
humanists on political advocacy. The first problem reflects a failure to teach all relevant
links, and particularly to appreciate the aesthetic influence of literature. The second
reflects the absence of a Golden Mean understanding. The third involves a failure to
confront bias. Shattuck (2000) suggests that there are but four ways to organize a
Humanities core curriculum: around Great Books, historical periods, overarching themes
(such as “truth” or “the other”), or ways of knowing. The comprehensive material suggests
a superior fifth approach, which would focus on the pros and cons of textual analysis and
the applicability of other methods to the subject matter of the Humanities, the various
relationships between Humanities phenomena and diverse other phenomena, the five
ethical perspectives (with perhaps an emphasis on how each can be found across diverse
philosophical traditions), and an appreciation of the strengths and weaknesses of theories
commonly employed by Humanists. Note that this approach facilitates the reasoned
comparison of works from Western and non-Western traditions.
262 Chapter 8

could be invaluable, though note that methods are best appreciated when
applied to something. An introductory course structured around diverse
causal links among the list of phenomena in Table 2.1 (and a similar list of
natural science phenomena) could help students choose the area in which
they wished to concentrate, as well as provide a base from which to integrate
later material. An ethical survey course might be especially valuable, given
the limited philosophical content of most other courses. Such a course would
provide students with some valuable guidelines on how to live their lives,
while also preparing them to appreciate the ethical issues relevant to their
other courses. Courses could also be structured around a typology of theory,
and a typology of criticisms of scholarly practice. In all cases, though, it is
critical that the comprehensive material is not isolated in one course, but that
continued reference is made to it in later courses.
The manageable scope of the comprehensive material deserves emphasis.
While human scientists care about hundreds of phenomena, these can be
organized in a handy table. This may seem overwhelming at first, but can
easily be mastered during a college education. The list of methods used by
scholars is even less difficult: colleges can reasonably hope to acquaint
students with the major strengths and weaknesses of each. Likewise, it is
certainly quite feasible to discuss each of the five broad types of ethical
analysis. The typology of theory may be the most difficult element in this
package, but since each of the five broad dimensions of the typology yields a
handful of possibilities, it too is certainly manageable in scope.
Experts on curricular reform emphasize the importance of integration of
general education with the major. “Campus leaders must attend not merely
to the coherence of the general education program but also to the
interrelation of general education and the majors” (Gaff, 1999, 4). As noted
above, if general education courses provided an introduction to the various
sorts of comprehensive material outlined above, it would be both
straightforward and advantageous for courses in the major thereafter to
constantly reference what theory type, method, phenomena, and ethical
approach were being discussed. Such a practice would make both instructors
and students more aware of potential limitations and biases, while permitting
students to place their major in a wider perspective. Ideally, it would
encourage instructors to “teach the conflicts” (Graff, 1992), rather than
pretending that only one theory or method or ethical perspective or subset of
phenomena is appropriate to the question at hand. Students, if informed of
the theories, methods, and so on being covered in each of their courses, in
terms of a shared vocabulary, would be able to recognize conflicts even if
instructors were not explicit about these; at present students often
compartmentalize their knowledge and can thus be blissfully unaware that
8. Concluding Remarks 263

different instructors disagreed (again, see Graff, 1992).13 Thus, not only can
comprehensive material be introduced without requiring any diminution in
attention devoted to the major, but this material will in fact enhance teaching
within the major. Furthermore, the comprehensive material will even help
students to choose their major (and options).
One further advantage of the comprehensive curriculum is that it
enhances the ability of a student to choose a problem-oriented rather than
discipline-determined area of specialization. While a handful of smallish
curricular changes have been entertained above, further changes could be
imagined which would see upper level coursework explicitly focussed on
particular links among specified sets of phenomena, rather than being
constrained by disciplinary boundaries. Degree programs themselves could
more readily be organized around interdisciplinary problems. A student
interested in inner city poverty or environmental degradation will struggle at
present to stitch together the appropriate courses from several disciplines; a
question-driven curriculum would greatly facilitate such a task. Given the
great diversity of complex issues that might attract the interest of a particular
student, institutions may struggle to offer a satisfactory range of upper year
courses. Several solutions are possible: directed reading courses, seminar
courses which draw together students working in related areas (perhaps with
some sort of community outreach component), openness to courses offered
by other institutions, and so on.
New technology will encourage and support this new curricular structure.
CD-Roms and the Internet both excel in allowing students to explore links
between diverse subjects; however, students need an overarching structure in
order to avoid getting lost following hypertext links. New technology also
facilitates the constant editing of both text and graphics. A world that
recognizes that knowledge is expanding gradually in all directions should
appreciate a curriculum that gives students a framework for readily
integrating new information. Adoption of these curricular reforms would
encourage technological developments in particular directions. Scholars
could, for example, work toward an electronic archive wherein academic
literature could be summarized and regularly updated in terms of causal links
and phenomena (see chapter 7). This would make it much easier for both
students and scholars to identify relevant literature.

13
Clarke and Agne (1997, 2-3), drawing on Piaget, argue that learners need to be actively
involved in their learning. They need to develop a structure of schemas, and build new
information onto these. Learning thus involves overcoming conflicts between previously
contradictory elements. They suggest that instructors would reach many more students if
they appealed to innate learning mechanisms. Later, they stress the advantages of asking
the 5W and How questions.
264 Chapter 8

Gaff (1999, 10) argues that all programs of curricular reform should have
an assessment component. While there is no one ideal method of assessment,
experts in the area often advocate methods which can be employed in-class
during a student’s final year. If the exposure of students to comprehensive
material were taken as a primary goal of curricula, it would be relatively
straightforward to test, using a variety of assignments, student mastery of this
material during their final year. Such an assessment technique would both
measure and encourage students’ preparation for lifelong learning. Note that
institutions would want to see how well the students could draw upon – not
just memorize – the comprehensive material, and thus could jointly assess
their mastery of the comprehensive material and particular coursework.
Student familiarity with how to perform interdisciplinary research could
likewise be measured in terms of both familiarity with and application of the
twelve-step process outlined above.

7.1 The Advantages for Student Learning of Particular

Classifications

A familiarity with lists of phenomena (chapter 2) would help students

both to place the subject matter of diverse courses in context, and recognize
those phenomena which none of their courses touched upon. Students at
present find it very difficult to relate material from different courses. This
problem is most severe in institutions where general education involves
taking courses from a variety of disciplines, but occurs even in
interdisciplinary programs: students are still often left to figure out on their
own how the course in political economy is connected to the course on
cognitive science or the course on environmental studies. The list of
phenomena gives them a cognitive map on which they can locate the subject
matter of diverse courses. Even if links between particular subjects were not
covered, they can see that such links exist. A course or set of courses
structured around lists of phenomena could guide students on how to explore
such links on their own (see Szostak, 2000, 2003). The classifications of
theory types, methods, and ethical perspectives provide three further
dimensions along which the content of diverse courses can be compared and
contrasted.
Without recognition of what they have not studied, students may
artificially constrain through ignorance their lifelong learning. The problems
and opportunities that people face in life rarely come in nice disciplinary
bundles. Only an interdisciplinary education gives individuals the basis from
which they can readily add to their knowledge base in order to react to the
challenges and possibilities which life presents. Yet lifelong learning
requires two broad skills: the ability to integrate theories and/or methods
8. Concluding Remarks 265

and/or phenomena and/or ethical perspectives, and a sense of where to look

for the theories, methods, and phenomena appropriate to a particular inquiry.
Note that individuals are much less likely to look for things of which they are
unaware. Most/all interdisciplinary programs do a very good job of teaching
students how to integrate, by taking them through a series of courses where
they are exposed to diverse examples of integration (though many of these
programs are recognizing that they need to more explicitly address the
process of interdisciplinary analysis). However, unless students are also
exposed to the widest range of theory, method, phenomena, and ethical
perspectives, they may be unsure of how to integrate beyond the scope of the
particular examples they were exposed to in class. Even the best of students
will eschew some opportunities out of ignorance.14
Exposure to a list of methods and a typology of theory types not only
acquaints students with the full range of possibilities. It provides a base from
which the strengths and weaknesses of these can be explored. This gives
students some guidance as they encounter diverse theoretical arguments, and
evidence from all methods, in their lives. Acquaintance with potential biases
in each of these might be particularly valuable.15 Students should learn to
appreciate evidence from multiple methods and arguments that accord with
multiple theories.
Perry (1970) famously argued that college students generally progress
from a dualistic perspective to one of “committed relativism.” In the first
phase they think that information is either right or wrong, and thus wish to
absorb the truth from their instructors. They come to appreciate over time
that differences of opinion exist. Some (hopefully most) progress to
appreciating that there is often no evidence with which to confidently choose
one view over another. Some may even appreciate that no amount of
evidence can prove any particular argument. Acquaintance with the strengths
and weaknesses of all theory types and methods should encourage students to
come to appreciate that understanding is imperfect but not impossible.

14
Klein (1999) lists twelve broad types of interdisciplinary curricula. Five are organized
around themes or issues, and one around a grand theory or method. Four involve
pedagogies that tend to be associated with interdisciplinarity: team teaching, learning
communities, student portfolios, and residential learning. Two involve developing models
of interdisciplinarity or reflecting on integration. All of these program types are laudable,
but on their own need not provide students with much-needed guidance on how to
integrate beyond their coursework. Comprehensive material could easily be integrated into
these approaches (though less easily into those that stress one grand theory).
15
Singleton and Strait (1999, 2-3) give a couple of examples of misleading media reports –
“More pedestrians are hurt in marked crosswalks” and “X% of doctors prescribe drug Y” –
and argue that students should be guided to ask the right questions: Do most pedestrians
cross streets at crosswalks, and are these where traffic is heaviest?; How many doctors
were involved, how often do they prescribe the drug; and who asked/selected them?
266 Chapter 8

There is much to be gained by exposing students to a classification of

ethical perspectives. Just as scientific theories should be tested by diverse
methods, an ethical statement that can be justified (albeit imperfectly; no
ethical argument is impervious to attack) in terms of all five types of ethical
analysis is one that individuals should place great confidence in. Since the
five types of ethical analysis are found across the world’s philosophical
traditions, such statements may apply across diverse societies. By applying
all five types of ethical analysis to the phenomena of interest to human
scientists a broad “ethical core” of such statements can be identified. Some
half of the 150 phenomena-based statements analysed in Szostak (2004)
belonged in the “ethical core”; the collective tendency in western societies at
least to focus on controversial issues diverts citizens from a realization of the
extent of philosophical consensus. Yet citizens can still understand how
individuals can have valid disagreements on societal goals, and are guided to
achieve policies that validate these differences. Societies can thus avoid
ethical anomie without privileging any one ethical tradition and thus
embracing ethical orthodoxy.
One goal of a liberal education is to prepare students for democratic
citizenship. An important facet of this is learning how to evaluate public
policy proposals. Judging by the list of common policy errors provided in
Vaughan and Buss (1998), these come from three main sources: failure to
properly specify goals [which in large part reflects inadequate philosophical
analysis of goals], reliance on faulty scientific analysis of means [which in
turn largely reflects reliance on a limited set of theories and methods], and,
perhaps most commonly, failure to appreciate potential side effects of a
proposed policy [which in turn reflects a failure to contemplate a variety of
causal linkages]. Exposure to the five types of ethical perspective would
encourage students to appreciate the diverse philosophical bases on which
opposing policy views (and political ideologies; see Szostak, 2002b) may
rest, guide them to respect the views of others, and also guide them to seek
consensus where possible. Policies that accord with all five types of ethical
analysis should be preferred. Exposure to a comprehensive list of
phenomena would guide students to appreciate the complexity of human
society, and thus the diverse effects that any policy initiative may have.
They should resist being seduced by oversimplification, while not
abandoning hope of they and others being able to evaluate alternatives. All
too often, expert policy advice ignores those links with which the expert is
unfamiliar; citizens should be aware of this tendency. Exposure to a diversity
of theory and methods should discourage students from taking arguments on
authority, while equipping them to analyze these. They should not gravitate
to the opposite extreme of thinking every argument equally valid. They
should also come to appreciate that nobody is free of subconscious biases;
8. Concluding Remarks 267

thus individuals should not ignore but rather seek to analyze their intuition.
Finally, recognition of the true complexity of human societies should
encourage everyone to seek out the views of people with diverse
perspectives.16

7.2 Responding to Recent Critiques of Liberal Education

Integrating comprehensive material into curricula allows powerful

responses to many modern critiques of the Liberal Arts.17 Gaff (1991) has
noted an unfortunate ideological element in curricular debates. Those on the
right bemoan the lack of a core of knowledge that colleges try to
communicate. To these it can be said that there is a core: a comprehensive
overview of theory, methods, phenomena, and philosophical perspectives
that should be provided to all students. This core, unlike alternative cores of
knowledge, such as a focus on Western Civilization, does not privilege any
one societal perspective. Gaff (1991) notes that proponents of core programs
are often driven by a fear that college curricula are no longer unified and
coherent, but he fears that there is no obvious way to achieve the goal of
coherence without losing the benefits of specialization. As noted above, the
comprehensive core requires little curricular space, serves to place
specializations in context, helps students to choose their area of
specialization, and shows students how to learn outside of that area.
Those on the left urge a greater focus on issues of diversity. Yet scholars
of diversity increasingly appreciate that one or two isolated required courses
do not teach students that diversity is important to their individual concerns.
The list of phenomena in Table 2.1 includes a category of social divisions by
race, gender, class, and family type, and thus a comprehensive curriculum
must inevitably acquaint students with the fact that such divisions are
causally related to all other phenomena. Cornwell and Stoddard (1999) stress
the value of understanding such linkages. They also stress the value of
recognizing that everyone is involved in cross-cultural contact; diversity can
thus be embraced without rejecting any particular civilization. The goal of
encouraging students to respect diversity will be further enhanced by
exposure to the full range of theory, method, and ethical perspectives: this is
a powerful antidote to narrow-mindedness of all sorts. The ethical survey in

16
Fish (1998) argued that reform of the academy can and should directly inform attempts to
reform the wider world. A curriculum that embraces complexity through
comprehensiveness enhances both societal ability and will to cope with the complexity of
our times.
17
As with the discovery of radio, a research quest not designed (primarily) to respond to such
criticisms may yield answers that direct attempts to respond might not.
268 Chapter 8

particular will develop an appreciation of the validity of alternative

viewpoints. Recall that each of the five types of ethical analysis can be found
across a wide range of philosophical traditions.
Several other critiques can be briefly touched upon (see Gaff, Ratcliff, et
al., eds., 1997). Some argue that students are poorly prepared for a complex
world and global economy; the classifications give them a mechanism for
coping with complexity, and highlight various ways in which societies
influence each other. Explicitly guiding students in public policy analysis
would provide further benefits. Some wonder how the Liberal Arts can cope
with the modern explosion of knowledge: the comprehensive material will
remain manageable in scope while opportunities for specialization are
expanded. Moreover, while there are limits to how much a student can learn
in college, there are no necessary limits to their understanding of where to
look for relevant information; the comprehensive curriculum teaches them
where to look and is thus admirably suited to a time of information
explosion. Concerns that universities do not encourage economic growth to
the degree possible should be alleviated by a curriculum that shows how the
economic sphere is connected to other phenomena. Critics worry that
universities serve faculty rather than society, that faculty do not work
effectively, and that much university research has little value; the
comprehensive approach breaks down the barriers between disciplines and
encourages a problem-oriented approach to both teaching and research.
Critics worry about outdated teaching methods; the suggested curriculum
allows students to link diverse course material and encourages students to
explore issues on their own (including via netsurfing and use of CD-Roms).
Such innovations might even (marginally) decrease the cost of education.
Those who worry about ethical decline should take solace in a philosophical
survey that steers a course between anomie and orthodoxy.
Defenders of the Liberal Arts often point to the skills inherent in a Liberal
Arts education. While this alone will not satisfy its critics, it is nevertheless a
worthy point. Klein (1996, 224) provides a list of desirable skills which
interdisciplinary programs can and should impart: students should know how
to structure a workable but flexible analytical framework, know how to
recognize and overcome ignorance in an area, know how to analyse
relationships and assess their importance, know how to balance depth and
breadth of knowledge, know how to identify and integrate salient concepts,
and know how to clarify and present material. The comprehensive
curriculum would serve all of these goals.
REFERENCES

Abbott, A., 2001, Chaos of Disciplines, University of Chicago Press, Chicago.

Abell, P., 2000, Sociological theory and rational choice theory, in: The Blackwell Companion
to Social Theory, B.S. Turner, ed., Blackwell, Oxford.
Almond, B., 1998, Exploring Ethics: a traveller’s tale, Blackwell, Oxford.
Amabile, T., 1996, Creativity in Context, Westview, Boulder CO.
Atkinson, P., and Hammersley, M., 1998, Ethnography and participant observation, in:
Strategies of Qualitative Inquiry, N.K. Denzin and Y.S. Lincoln, eds., Sage, Thousand
Oaks CA.
Baeza-Yates, R.,, and Ribeiro-Neto, B., 1999, Modern Information Retrueval. Addison-
Wesley, Harlow UK.
Bailey, K.D., 1994, Typologies and Taxonomies: an introduction to classification techniques,
Sage, Thousand Oaks CA.
Baird, D., 2000, Analytical instrumentation and instrumentational objectivity, in: Of Minds
and Molecules: new philosophical perspectives on chemistry, N. Bhushan and S.
Rosenfeld, eds., Oxford University Press, Oxford.
Bal, M., 2002, Travelling Concepts in the Humanities: a rough guide. University of Toronto
Press, Toronto.
Barber, B., 1993, Constructing the Social System, Transactions Publishers, New Brunswick
NJ.
Barnes, B., 1995, The Elements of Social Theory, Princeton University Press, Princeton NJ.
Baron, M.W., Pettit, P., and Slote, M., eds., 1997, Three Methods of Ethics, Blackwell,
Oxford.
Bates, M.J., 1996, Learning about the information seeking of interdisciplinary scholars and
students, Library Trends 45(2): 155-164.
Bechtel, W., 1986, Science and Philosophy: integrating scientific disciplines, Martinus
Nijhoff, Dordrecht.
Bechtel, W., 2000, From imagining to believing: epistemic issues in generating biological
data, in: Biology and Epistemology, R. Creath and J. Maienschein, eds., Cambridge
University Press, Cambridge, 138-66.
Bell, J., 1999, Doing Your Research Project,. 3rd. ed., Open University Press, Buckingham.
Beveridge, W.I.B., 1957, The Art of Scientific Investigation, Norton, New York.
Bhushan, N., and Rosenfeld, S., eds., 2000, Of Minds and Molecules: new philosophical
perspectives on chemistry, Oxford University Press, New York.
270 References

Black, J., 1997, Maps and Politics, Reaktion Books, London.

Blalock, H., 1969, Theory Construction: from verbal to mathematical models, Prentice-Hall,
Englewood Cliffs NJ.
Bouma, G.D., and Atkinson. G.B.J., 1995, A Handbook of Social Science Research, Oxford
University Press, Oxford.
Bowker, G.C., and Star, S.L., 1999, Sorting Things Out: classification and its consequences,
MIT Press, Cambridge.
Boylan, T.A., and O’Gorman, P.F., 1995, Beyond Rhetoric and Realism in Economics,
Routledge, London.
Boyne, R., 2000, Structuralism, in: The Blackwell Companion to Social Theory, B.S. Turner,
ed., Blackwell, Oxford.
Brewer, J., and Hunter, A., 1989, Multimethod Research: a synthesis of styles, Sage, Newbury
Park CA.
Bromne, R., 2000, The psychology of cognitive interdisciplinarity, in: Practising
Interdisciplinarity, P. Weingart and N. Stehr, eds.,University of Toronto Press, Toronto.
Brown, R.H., ed., 1992, Writing the Social Text: poetics and politics in social science
discourse, Aldine de Gruyter, New York.
Bryant, R., 2000, Discovery and Decision: exploring the metaphysics and epistemology of
scientific classification, Fairleigh Dickinson University Press, Madison.
Buckingham Shum, S.J., 2003, Visualizing internetworked argumentation, in: Visualizing
Argumentation: software tools for collaborative and educational sense-making, P.
Kirschner, S.J. Buckingham Shum, and C. Carr, eds., Springer, London.
Bulick, S., 1982, Structure and Subject Interaction, Marcel Dekker, New York.
Bunge, M., 1998, Social Science Under Debate: a philosophical perspective, University of
Toronto Press, Toronto.
Caws, P., 1993, Yorick’s World: science and the knowing subject, University of California
Press, Berkeley.
Clarke, J., and Agne, R.M., 1997, Integrated High School Teaching: strategies for integrated
learning, Allyn and Bacon, Boston.
Cohen, I.J., 2000, Theories of action and praxis, in: The Blackwell Companion to Social
Theory, B.S. Turner, ed., Blackwell, Oxford.
Cole, S., 1992, Making Science: between nature and society, Harvard University Press,
Cambridge MA.
Collins, H., and Pinch, T., 1993, The Golem: what everyone should know about science,
Cambridge University Press, Cambridge.
Collins, R., 1998, The Sociology of Philosophies: a global theory of intellectual change,
Belknap (Harvard University) Press, Cambridge MA.
Cornwell, G., and Stoddard, E., 1999, Globalizing Knowledge: connecting international and
intercultural studies, Association of American Colleges and Universities, Washington.
Creath, R., and Maienschein, J., eds., 2000, Biology and Epistemology, Cambridge University
Press, Cambridge.
Cussins, C.M., 2000. Primate suspect: some varieties of science studies, in: Primate
Encounters: models of science, gender, and society, S.C. Strum and L.M. Fedigan, eds.,
University of Chicago Press, Chicago.
Dahlberg, I., 1994, Domain interaction: theory and practice, in: Knowledge Organization and
Qualitative Management, H. Albrechtson and S. Oernager, eds., INDEKS Verlag,
Frankfurt.
Dale, A., and Davies, R.B., 1994, Introduction, in: Analyzing Social and Political Change, A.
Dale and R.B. Davies, eds., Sage, London.
References 271

Denzin, N.K., and Lincoln, Y.S., 1998, Introduction to this volume, in: Strategies of
Qualitative Inquiry, N.K. Denzin and Y.S. Lincoln, eds., Sage, Thousand Oaks CA.
Dogan, M., 1996, The hybridization of social science knowledge, Library Trends 45(2): 296-
314.
Dogan, M., and Pahre, R., 1990, Creative Marginality: innovation at the intersections of
social science, Westview Press, Boulder CO.
Dupré, J., 2001, Human Nature and the Limits of Science, Clarendon, Oxford.
Elliott, A., 2000, Psychoanalysis and social theory, in: The Blackwell Companion to Social
Theory, B.S. Turner, ed., Blackwell, Oxford.
Ellis, L., 1994, Research Methods in the Social Sciences, Brown and Benchmark, Madison.
Elster, J., 1989, Nuts and Bolts For the Social Sciences, Cambridge University Press,
Cambridge.
Fabian, J., 1975, Taxonomy and Ideology: on the boundaries of concept classification, Peter
de Redder Press, Lisse.
Feuer, L., 1995, Varieties of Scientific Experience. Transactions Publishers, New Brunswick
NJ.
Fiscella, J., 1996, Bibliography as an interdisciplinary service, Library Trends 45(2): 280-95.
Fish, S., 1998, Being interdisciplinary is so very hard to do, in: Interdisciplinarity: essays
from the literature, W.H. Newell, ed., College Board, New York.
Fox, G., 1997, Reason and Reality in the Methodology of Economics, Edward Elgar,
Cheltenham UK.
Fraser, C., and Gaskell, G., eds., 1990, The Social Psychological Study of Widespread Beliefs,
Clarendon, Oxford.
Freeman, D., 1994, Nature’s Imagination: the frontiers of scientific vision, Oxford University
Press, Oxford.
Freidheim, E.A., 1982, From Types to Theory: a natural method for an unnatural science,
University Press of America, Washington.
Fuller, S., 1993, Philosophy of Science and its Discontents, Guilford Press, New York.
Fuller, S., 1997, Science, University of Minnesota Press, Minneapolis.
Fuller, S., 1998, The position: interdisciplinarity as interpretation, in: Interdisciplinarity:
essays from the literature, W.H. Newell, ed., College Board, New York.
Gaff, J.G., 1991, New Life For the College Curriculum, Jossey-Bass, San Francisco.
Gaff, J.G., 1997, Tensions between tradition and innovation, in: Handbook of the
Undergraduate Curriculum, J.G. Gaff, J. Ratcliff, and associates, eds., Jossey Bass, San
Francisco.
Gaff, J.G., 1999, General Education: the changing agenda, Association of American Colleges
and Universities, Washington.
Gaff, J.G., Ratcliff, J., and Associates, eds., 1997, Handbook of the Undergraduate
Curriculum, Jossey-Bass, San Francisco.
Galison, P., 1996, Introduction: the content of disunity, in P. Galison and D. J. Stump, eds.,
The Disunity of Science: boundaries, contexts, and power. Stanford University Press,
Stanford, pp. 1-36.
Getman, J., 1992, In the Company of Scholars, University of Texas Press, Austin.
Giere, R., 1988, Explaining Science: a cognitive approach, University of Chicago Press,
Chicago.
Glassick, C.E., Huber, M.T., and Maeroff, G.I., 1997, Scholarship Assessed: evaluation of the
professoriate, Jossey-Bass, San Francisco.
Goldenberg, S., 1992, Thinking Methodologically, Harper Collins, New York.
272 References

Gower, B., 1997, Scientific Method: an historical and philosophical introduction, Routledge,
London.
Graff, G., 1992, Beyond the Culture Wars: how teaching the conflicts can revitalize American
education, Norton, New York.
Haack, S., 1998, Manifesto of a Passionate Moderate, University of Chicago Press, Chicago.
Hall, D.J., and Hall, I., 1996, Practical Social Research: Project Work in the Community,
Macmillan, Basingstoke UK.
Hall, P.A., 2003, Aligning ontology and methodology in comparative research, in:
Comparative Historical Analysis in the Social Sciences, J. Mahoney and D. Reuschmeyer,
eds., Cambridge University Press, Cambridge, pp. 373-404.
Halperin, D.F., 1997, Critical Thinking Across the Curriculum, Lawrence Erlbaum
Associates, Mahwah NJ.
Hammersley, M., 1996, The relationship between qualitative and quantitative research:
paradigmatic loyalty versus methodological eclecticism, in: Handbook of Qualitative
Research Methods, J.T.E. Richardson, ed., British Psychological Association, Leicester.
Hammersley, M., 2000, Taking Sides in Social Research, Routledge, London.
Hammersley, M., and Gomm, R., 2000, Biases in social research, in: Taking Sides in Social
Research, by M. Hammersley, Routledge, London.
Hartman, J.J., and Hedblom, J.H., 1988, Methods for the Social Sciences: a handbook for
students and non-specialists, 3rd ed., Greenwood Press, Westport CT.
Hawthorn, G., 1991, Plausible Worlds, Cambridge University Press, Cambridge.
Hedstrom, P., and Swedberg, R., eds., 1998, Social Mechanisms: An Analytical Approach to
Social Theory, Cambridge University Press, Cambridge.
Herda, E.A., 1999, Research Conversations and Narrative: a critical hermeneutics orientation
in participatory inquiry, Praeger, Westport CT.
Hintikka, J., 2000, Introduction: epistemology, in: S. Rosen, ed., The Examined Life: readings
from Western philosophy from Plato to Kant, Random House, New York.
Hjorland, B., 1994, Subject representation and information seeking: survey of a doctoral
thesis, Knowledge Organization 21(2): 94-8.
Holton, G.J., 1993, Science and anti-science, Harvard University Press, Cambridge MA.
Horn, R.E., 2003, Infrastructure for navigating interdisciplinary debates, in: Visualizing
Argumentation: software tools for collaborative and educational sense-making, P.
Kirschner, S.J. Buckingham Shum, and C.S. Carr, eds., .Springer, London.
Hough, G.A., 1984, News Writing 3rd ed., Houghton-Mifflin, Boston.
Hubenthal, U., 1994, Interdisciplinary thought, Issues in Integrative Studies 12: 55-75.
Hull, D.L., 1989, Science as a Process, University of Chicago Press, Chicago.
Hull, D.L., 1998, Introduction to part V, in: The Philosophy of Biology, D.L. Hull and M.
Ruse, eds., Oxford University Press, Oxford.
Hull, D.L., and Ruse, M., eds., 1998, The Philosophy of Biology, Oxford University Press,
Oxford.
Hunt, M., 1997, How Science Takes Stock: the story of meta-analysis, Russell Sage
Foundation, New York.
Hunter, E.J., 1988, Classification Made Simple, Gower, Aldershot UK.
Jackendoff, R., 1992, Languages of the Mind, MIT Press, Cambridge MA.
Jones, P., 1985, Theory and Method in Sociology: a guide for the beginner, University
Tutorial Press, Slough UK.
Kantorovich, A., 1993, Scientific Discovery, SUNY Press, Albany.
Katzer, J., Cook, K.H., and Crouch, W.W., 1998, Evaluating Information: a guide for users of
social science research, McGraw-Hill, Boston.
References 273

Kent, B., 1987, Charles S. Peirce: logic and classification of the sciences, McGill-Queen’s
University Press, Montreal.
Kervin, A., 1997, Changes in the humanities, in: What’s Happened to the Humanities?, A.
Kervin, ed., Princeton University Press, Princeton.
Kincaid, H., 1996, Philosophical Foundations of the Social Sciences, Cambridge University
Press, Cambridge.
Kirschner, P., Buckingham Shum, S.J., and Carr, C.S., eds., Visualizing Argumentation:
software tools for collaborative and educational sense-making, Springer, London.
Kitcher, P., 1982, Abusing Science, MIT Press, Cambridge MA.
Klee, R., 1999, Scientific Inquiry: readings in the philosophy of science, Oxford University
Press, New York.
Klein, E., 1998, Economic Theories and Their Relational Structure, St. Martin’s, New York.
Klein, J.T., 1990, Interdisciplinarity, Wayne State University Press, Detroit.
Klein, J.T., 1996, Crossing Boundaries: knowledge, disciplinarities, and interdisciplinarities,
University Press of Virginia, Charlottesville.
Klein, J.T., 1996a, Interdisciplinary trends: the current context, Library Trends 45(2): 34-54.
Klein, J.T., 1999, Mapping Interdisciplinary Studies, Association of American Colleges and
Universities, Washington.
Klein, J.T., 2000, A conceptual vocabulary of interdisciplinary science, in: Practising
Interdisciplinarity, P. Weingart and N. Stehr, eds., University of Toronto Press, Toronto.
Klein, L., and Eason, K., 1991, Putting Social Science to Work, Cambridge University Press,
Cambridge.
Kontopolous, K., 1993, The Logics of Social Structure, Cambridge University Press,
Cambridge.
Krieger, M.H., 1997, Doing Physics: how physicists take hold of the world, University of
Indiana Press, Bloomington.
Kruglanski, A., 1989, Lay Epistemics and Human Knowledge, Plenum, New York.
Kuhn, A., 1974, The Logic of Social Systems, Jossey-Bass, San Francisco.
Kuhn, T.S., 1962, The Structure of Scientific Revolutions, University of Chicago Press,
Chicago.
Lambert, K., and Brittan, G.G. Jr., 1992, An Introduction to the Philosophy of Science 4th ed.,
Ridgeview Publishing Company, Alascedero CA.
Langley, P., Simon, H.A., Bradshaw, G.L., and Zytkow, J.M., 1987, Scientific Discovery:
computational explorations of the scientific process, MIT Press, Cambridge MA.
Langridge, D.W., 1992, Classification: its kinds, elements, systems, and applications, Bowker,
London.
Lechner, F.J., 2000, Systems theory and functionalism, in: The Blackwell Companion to
Social Theory, B.S. Turner, ed., Blackwell, Oxford.
Lenoir, T., and Gumbrecht, H.U., eds., 1996, Writing Science, Stanford University Press,
Stanford.
Little, D., 1998, Microfoundations, Method, and Causation, Transactions Publishers, New
Brunswick NJ.
Longino, H.E., 2000, Toward an epistemology for biological pluralism, in: Biology and
Epistemology, R. Creath and J. Maienschein, eds., Cambridge University Press,
Cambridge, 261-86.
Madden, E.H., ed., 1989, Theories of Scientific Method: the Renaissance through the
nineteenth century, Gordon and Breach, New York.
274 References

Magnum, D., 2000, Down the primrose path: competing epistemologies in early twentieth
century biology, in: Biology and Epistemology, R. Creath and J. Maienschein, eds.,
Cambridge University Press, Cambridge, 91-121.
Mahner, M., and Bunge, M., 1997, Foundations of Biophilosophy, Springer-Verlag, New
York.
Mann, T., 1993, Library Research Models: a guide to classification, cataloguing, and
computers. Oxford University Press, Oxford.
Marcella, R., and Newton, R., 1994, A New Manual of Classification, Gower, Aldershot.
May, T., 1997, Social Research: issues, methods, and process, 2nd. ed., Open University
Press, Buckingham.
Mayer, T., 1993, Truth Versus Precision, Edward Elgar, London.
McCloskey, D.N., 1990, If You’re So Smart... University of Chicago Press, Chicago.
McCloskey, D.N., and Ziliak, S.T., 1996, The standard error of regressions, Journal of
Economic Literature 34(1): 97-114.
McKeon, R., 1994, On Knowing – The Natural Sciences, University of Chicago Press,
Chicago.
McKim, V.R., 1997, Introduction, in: Causality in Crisis: statistical methods and the search
for causal knowledge in the social sciences, V.R. McKim and S.P. Turner, eds., University
of Notre Dame Press, Notre Dame IN.
McNeill, P., 1990, Research Methods, 2nd ed., Routledge, London.
Meltzoff, J., 1998, Critical Thinking About Research, American Psychological Association,
Washington.
Merton, R.K., 1973, The Sociology of Science: theoretical and empirical investigations,
University of Chicago Press, Chicago.
Messer-Davidow, E., Shumway, D.R., and Sylvan, D.J., 1993, Knowledges: historical and
critical studies in interdisciplinarity, University of Virginia Press, Charlottesville.
Mokyr, J., 2002, The gifts of Athena : historical origins of the knowledge economy,
Princeton University Press, Princeton.
Montgomery, S.L., 1996, The Scientific Voice, The Guilford Press, New York.
Moran, J., 2002, Interdisciplinarity, Routledge, New York.
Morton, A., 1997, A Guide Through the Theory of Knowledge, Blackwell, Oxford.
Moser, P.K., Mulder, D.H., and Trout, J.D., 1998, The Theory of Knowledge: a thematic
introduction, Oxford University Press, New York.
Motterlini, M., 1999, For and Against Method, University of Chicago Press, Chicago.
Mowery, D.C., and Rosenberg, N., 1989, Technology and the Pursuit of Economic Growth,
Cambridge University Press, Cambridge.
Mukherjee, R., 1983, Classification in Social Research, SUNY Press, Albany.
Neurath, O., Carnap, R., and Morris, C.W., eds., 1970, Foundations of the Unity of Science,
University of Chicago Press, Chicago.
Newell, W.H., 1998, The case for interdisciplinary study: response to professor Benson’s five
arguments, in W.H. Newell, ed., Interdisciplinarity: essays from the literature, College
Board, New York.
Newell, W.H., 2001, A theory of interdisciplinary studies, Issues in Integrative Studies 19:1-
26.
Nicolescu, B., 2002, Manifesto of Transdisciplinarity [trans Karen-Claire Voss], SUNY Press,
Albany.
Norris, C., 2000, Minding the Gap: epistemology and the philosophy of science in the two
traditions, University of Massachusetts Press, Amherst.
References 275

Oakley, A., 2000, Experiments in Knowing: gender and method in social science, The New
Press, New York.
Pahre, R., 1996, Knowledge communication in the social sciences, Library Trends 45(2): 204-
225.
Palmer, C.L., 1996, Information work at the boundaries of science: linking library services to
research practices, Library Trends 45(2): 165-191.
Palmer, C.L., 2001, Work at the Boundaries of Science: information and the interdisciplinary
research process, Kluwer, Dordrecht.
Palys, T., 1997, Research Decisions: quantitative and qualitative perspectives, Harcourt
Brace, Toronto.
Perry, W.G. Jr., 1970, Forms of Ethical and Intellectual Development in the College Years: a
scheme, Holt, Rinehart, and Winston., New York.
Pervin, L.A., ed., 1990, Handbook of Personality Theory and Research, Guilford Press, New
York.
Plummer, K., 2000, Symbolic interactionism in the twentieth century, in: The Blackwell
Companion to Social Theory, B.S. Turner, ed., Blackwell, Oxford.
Pole, C.J., and Burgess, R.G., eds., 2000, Cross-Cultural Case Study, JAI Press, Amsterdam.
Polkinghorne, J., 1996, Beyond Science: the wider human context, Cambridge University
Press, Cambridge.
Potter, G., 2000, The Philosophy of Social Science, Prentice-Hall, Harlow UK.
Pounds, N.J.G., 2001, What economic history means to me, in: Living Economic and Social
History, P. Hudson ed., Economic History Society, Glasgow, 288-92.
Prus, R., 1996, Symbolic Interaction and Ethnographic Research, SUNY Press, Albany.
Rachels, J., 1993, Moral philosophy as a subversive activity, in: Applied Ethics, E.R. Winkler
and J.R. Coombs, eds., Blackwell, Oxford.
Ragin, C.C., 2000, Fuzzy Set Social Science, Chicago: University of Chicago Press.
Reason, P., 1998, Three approaches to participative inquiry, in: Strategies of Qualitative
Inquiry, N.K. Denzin and Y.S. Lincoln, eds., Sage, Thousand Oaks CA.
Root, M., 1993, Philosophy of Social Science, Blackwell, Oxford.
Rosser, J.B., 1999, On the complexities of complex economic dynamics, Journal of Economic
Perspectives 13(4): 169-92.
Rota, G., and Crants, J.T., 2000, Introduction: ten philosophical and contradictory predictions,
in: The Examined Life: readings from Western philosophy from Plato to Kant, S. Rosen,
ed., Random House, New York.
Rothbart, D., 2000, Substance and function in chemical research, in: Of minds and molecules:
new philosophical perspectives on chemistry, N. Bhushan and S. Rosenfeld, eds., Oxford
University Press, Oxford.
Rowley, J.E., 1992, Organizing Knowledge 2nd ed., Ashgate, Aldershot.
Rule, J.B., 1997, Theory and Progress in Social Science, Cambridge University Press,
Cambridge.
Salter, L., and Hearn, A., 1996, Outside the Lines: issues in interdisciplinary research,
McGill-Queen’s University Press, Montreal.
Sandler, T., 2001, Economic Concepts for the Social Sciences, Cambridge University Press,
Cambridge.
Sartori, G., ed., 1984, Social Science Concepts: a systematic analysis, Sage, Beverly Hills.
Schaffner, K.F., 2000, Afterword, in: Biology and Epistemology, R. Creath and J.
Maienschein, eds., Cambridge University Press, Cambridge, 287-95.
Scriven, M., 1988, Explanations, predictions, and laws, in: Theories of Explanation, J.C. Pitt,
ed., Oxford University Press, Oxford.
276 References

Seale, C., 1999, The Quality of Qualitative Research, Sage, London.

Shattuck, R., 2000, Conclusion: what now shall we teach?, in: Alive at the Core, M. Nelson,
ed., Jossey-Bass, San Francisco.
Shipman, M., 1988, The Limitations of Social Research 3rd ed., Longman, London.
Silverman, D., 2000, Doing Qualitative Research: a practical handbook, Sage, London.
Simonton, D., 1988, Scientific Genius: a psychology of science, Cambridge University Press,
Cambridge.
Sin, S., 1992. Deconstruction, in: A Companion to Aesthetics, D.E. Cooper, ed., Blackwell,
Oxford.
Singleton, R.A. Jr., and Strait, B.C., 1999, Approaches to Social Research 3rd ed., Oxford
University Press, New York.
Sklar, L., 1992, Philosophy of Physics, Westview, Boulder CO.
Smiraglia, R.P., 2001, The Nature of a Work: implications for the organization of knowledge,
Scarecrow Press, Lanham MD.
Stake, R.E., 1998, Case studies, in: Strategies of Qualitative Inquiry, N.K. Denzin and Y.S.
Lincoln, eds., Sage, Thousand Oaks CA.
Stanley, T.D., 2001, Wheat from chaff: meta-analysis as quantitative literature review,
Journal of Economic Perspectives 15(3): 131-50.
Stocklmayer, S.M., Gore, M.M., and Bryant, C., 2001, Science Communication in Theory and
Practice, Kluwer, Dordrecht.
Stoner, N.W., 1973, Introduction, in: [R.K. Merton] The Sociology of Science: theoretical and
empirical investigations, University of Chicago Press, Chicago.
Storr, A., 1996, Feet of Clay: saints, sinners and madmen: a study of gurus, Free Press, New
York.
Strauss, A., and Corbin, J., 1998, Grounded theory methodology: an overview, in: Strategies
of Qualitative Inquiry, N.K. Denzin and Y.S. Lincoln, eds., Sage, Thousand Oaks CA.
Stump, D.J., 1996, Afterword: new directions in the philosophy of science studies, in: Writing
Science, T. Lenoir and H.U. Gumbrecht, eds., Stanford University Press, Stanford.
Svenonius, E., 2000, The Intellectual Foundations of Information Organization, MIT Press,
Cambridge MA.
Swift, D.F., Winn, V., and Bramer, D., 1978, ’Aboutness’ as a strategy for retrieval in the
social sciences. Aslib Proceedings 30(5): 182-7.
Szostak, R., 1991, The Role of Transportation in the Industrial Revolution: a comparison of
England and France, McGill-Queen’s University Press, Montreal.
Szostak, R., 1999, Econ-Art: divorcing art from science in modern economics, Pluto, London.
Szostak, R., 2000, Toward a unified human science, Issues in Integrative Studies 18: 115-57.
Szostak, R., 2001 Putting social structure in its place, schematically, Issues in Integrative
Studies 19: 171-220.
Szostak, R., 2002, How to do interdisciplinarity: integrating the debate, Issues in Integrative
Studies 20: 103-22.
Szostak, R., 2002a, Intuition and interdisciplinarity: a reply to Mackey, Issues in Integrative
Studies 20: 131-7.
Szostak, R., 2002b, Politics and the five types of ethical analysis, International Journal of
Politics and Ethics 2(4): 275-90.
Szostak, R., 2003, A Schema For Unifying Human Science: interdisciplinary perspectives on
culture, Susquehanna University Press, Selinsgrove PA.
Szostak, R., 2003a, ’Comprehensive’ curricular reform: providing students with an overview
of the scholarly enterprise, Journal of General Education 52(1): 27-49.
Szostak, R., 2004, Unifying Human Ethics, Ms. under review.
References 277

Szostak, R., 2005, Evaluating the historiography of the Great Depression: explanation or
single-theory driven, Journal of Economic Methodology, Forthcoming.
Tashakkori, A., and Teddlie, C., 1998, Mixed Methodology: combining quantitative and
qualitative approaches, Sage, Thousand Oaks CA.
Taylor, C.A., 1996, Defining Science, University of Wisconsin Press, Madison.
Tessler, M., ed., with Nachtwey, J., and Banda, A., 1999, Area Studies and Social Science:
strategies for understanding Middle East politics, Indiana University Press, Bloomington.
Turner, B.S., 2000, Preface, Introduction, in: The Blackwell Companion to Social Theory, B.S.
Turner, ed., Blackwell, Oxford.
Turner, S.P., 1996, Introduction: social theory and sociology, in: Social Theory and Sociology,
S.P. Turner, ed., Blackwell, Cambridge MA.
Turner, S.P., 2002, Brains/Practices/Relativism: social theory after cognitive science,
University of Chicago Press, Chicago.
Urry, J., 2000, Sociology of time and space, in: The Blackwell Companion to Social Theory,
B.S. Turner, ed., Blackwell, Oxford.
Vaitkus, S., 2000, Phenomenology and sociology, in: The Blackwell Companion to Social
Theory, B.S. Turner, ed., Blackwell, Oxford.
Valiela, I., 2001, Doing Science: design, analysis, and communication of scientific research,
Oxford University Press, New York.
Vaughan, R.J., and Buss, T., 1998, Communicating Social Science Research to Policy-makers,
Sage, Thousand Oaks CA.
Velkley, R., 2000, Introduction: metaphysics, in: The Examined Life: readings from Western
philosophy from Plato to Kant, S. Rosen, ed., Random House, New York.
Vicedo, M., 2000, Experimentation in early genetics: the implications of the historical
character of science for scientific realism, in: Biology and Epistemology, R. Creath and J.
Maienschein, eds., Cambridge University Press, Cambridge, 215-43.
Ward, S.C., 1996, Reconfiguring Truth, Rowman and Littlefield, Lanham MD.
Weinberg, B.H., 1988, Why indexing fails the researcher, The Indexer 16(1): 3-6.
Weingart, P., 2000, Interdisciplinarity: the paradoxical discourse, in: Practising
interdisciplinarity, P. Weingart and N. Stehr, eds., University of Toronto Press, Toronto.
Wellman, B., 1998, On Norman H. Die, Dale H. Bent, and C. Hadlai Hull, SPSS, doing it
ourselves, in: Required Reading: Sociology’s most influential books, D. Clawson, ed.,
University of Massachusetts Press, Amherst.
White, H.D., 1996, Library retrieval for interdisciplinary synthesis, Library Trends 45(2):
239-64.
Whitley, R., 2000, The Intellectual and Social Organization of the Sciences, Oxford
University Press, Oxford.
Williams, R., 1985, Keywords: a vocabulary of culture and society 2nd ed., Oxford University
Press, Oxford.
Willinsky, J., 1999, Technologies of Knowledge: a proposal for the human sciences, Beacon,
Boston.
Winkler, E.R., and Coombs, J.R., 1993 From Kantianism to contextualism, in: Applied Ethics,
E.R. Winkler and J.R. Coombs, eds., Blackwell, Oxford.
Wolfe, A., 1999 Moral inquiry in social science, in: The Nature of Moral Inquiry in the Social
Sciences, C. Cochran, D. Hollenbach, A.Wolfe, and R. Wuthnow, Erasmus Institute,
Rotterdam.
Woolgar, S., 1988, Science: the very idea, Ellis Horwood, Chichester UK.
Ziman, J., 2000, Real Science: What it is and what it means, Cambridge University Press,
New York.
INDEX

Abbott, A., 110, 254 Boyne, R., 87, 89

Abell, P., 92-3 Brady, M., 4
action research, 142, 180 Brewer, J., 151
action theory, 54, 84-7, 94 Bromne, R., 21-2, 163
Almond, B., 196 Brown, R.H., 168
Amabile, T., 163 Bryant, R., 13
anthropology, 147 Buckingham Shum, S.J., 237
area studies, 141 Bulick, S., 221
art, 5, 24-5, 28, 31, 42-3, 58, 165, 175-6, Bunge, M., 17, 33, 36, 60, 65, 69-70, 74,
214-5, 221, 253 77, 108, 161, 180, 252; See also
Aristotle, 3, 9, 16, 31, 72, 99 Mahner
Atkinson, P., 119, 129, 140, 145
Carnap, R. 15, 173
Baeza-Yates, R., 238 case study, 92, 102, 113, 132, 140-1, 150,
Bailey, K.D., 9-14 152, 200-1; See also various methods
Baird, D., 40 causal argument, criteria for establishing,
Bal, M., 42-5 105-6
Barber, B., 31 causal links, 30-49, 107, 115, 128, 134,
Barnes, B., 57 161, 176, 189, 202-5, 235, 243-4, 262,
Baron, M.W., 194 266; classifying theory with respect to,
Bates, M.J., 219, 225 53; defined, 31-2; generalizability and,
Bechtel, W., 15, 21, 40 68-9, 83, 201; mapping, 32, 34, 53, 68,
Bell, J., 119, 123, 142 143-4, 206, 228, 237, 243, 257;
Beveridge, W.I.B., 163-4 notation for, 222-4, 228, 232;
Bhushan, N., 26, 38, 40, 45, 143 reductionism avoided in, 38, 244;
biases and errors, defined, 158-60 influence of individuals and societal
biology, 2, 12-3, 26, 38, 54, 59, 69, 82-3, aggregates reflected, 36, 56-7, 59-61,
87, 95, 110-1, 117-8, 121, 140, 147-8, 67, 85, 96; role in definition of
197, 212, 220, 251 science, 170, 172; role in definition of
Black, J., 144 theory, 74-7; systems of, 33, 39, 72-3,
Blalock, H., 33 87-9, 149-50, 199; ubiquity of, 33, 56-
Bouma, G.D., 119, 140, 145 7, 59, 61, 111-2, 173, 199-200, 210,
Bowker, G.C., 11, 13, 118, 228 241; See also phenomena
Boylan, T.A., 6, 207
280 Index

causation, types of, 31-2, 72, 80-2, 244, senses and, 45, 47; operationalization
246 and, 40, 48-50, 52, 63, 75-6, 78-80,
Caws, P., 6 115, 119, 131, 206-7, 244, 246, 248;
chaos theory; See complexity quantification of, 47-9
chemistry, 1-2, 5, 9, 12-3, 26, 38, 40, 69, decision-making, five types of, 66-7, 83-
73, 118, 120-1, 143, 149, 220 5, 96, 107, 181, 186, 193-8, 214, 231,
cholera, 132, 148 245, 248, 253, 266
Clarke, J., 263 deduction; See induction
classification; as a method, 102, 117-8, definitions, 7, 30-2, 34, 38, 41, 43-4, 50,
138, 149, 228; desirable qualities, 2-3, 77, 87, 113, 134, 137, 170-2, 188, 222,
5, 15-6; exhaustiveness, 2, 5, 7, 24, 31, 243-5, 256-7
41, 54-5, 97, 102, 118, 144-6, 158-61, Denzin, N.K., 110
192-4, 196, 232-3, 241-2, 253; Dewey, J., 1, 85-6
guidelines for, 9-14, 107, 117-8, 161; disciplinary worldview, 41, 151, 153-4,
importance of, 1-2, 8, 14-6, 18, 23, 31- 183, 200, 204, 243, 256, 260
6, 50, 95-8, 144-54, ch. 7, 241-68; disciplines, 26, 30, 63, 96-7, 100-1, 173,
reification and, 11-3, 48, 50, 118; 177, 236, 246, 251-2, 254, 256;
summary of efforts at, 239-40; characteristics of, 8, 41, 177, 185, 187;
unidimensional versus multidimen- classifying, 14-6, 23, 33, 38; problems
sional, 2, 10, 37 inherent in, 19-21, 33, 59, 104, 147-8,
Cohen, I.J., 84-7 168-9, 177, 179, 181-4, 187, 189, 192,
Cole, S., 156-7, 167, 182, 184, 188, 190- 201, 205, 208-12, 214, 216, 221-2,
1, 212 226, 263. See also human science;
Collins, H., 121, 156, 167, 191, 235 humanities; interdisciplinarity; natural
Collins, R., 54, 155, 183-4, 210, 220, 232 science; social science; various discip-
common ground theory, 22 lines
community studies, 141 diversity, concerns with, 35-6, 111, 137,
comparative method, 141 166, 168, 198, 253, 267; See also
complexity theory, 72, 83, 94 postmodernism
Comte, A., 42, 76 Dogan, M., 21, 26, 30, 41-2, 80, 96, 180,
concepts, scientific, 7, 21, 41-5, 76-7, 184, 208-10, 220, 227, 232, 254
113, 157, 172, 241, 250 Dupré, J., 3, 32, 39, 67, 96, 174
Cornwell, G., 36, 267 Durkheim, E., 54
Courage, 16, 164, 252
Creath, R., 147 economics, 32-4, 39-40, 42, 49, 55-6, 63-
culture, 24-6, 28, 30-1, 33-4, 36-8, 43, 46, 4, 69-70, 96, 115, 121, 130-2, 147,
54, 58, 66, 76, 84-7, 93, 95, 105, 112, 151, 177, 179-80, 184, 190, 193, 200,
166, 172, 175, 267; influences on 205, 214, 229
science, 165-9, 182, 190 economy, 24, 28, 30-1, 38, 58, 117-8,
Cussins, C.M., 157-8, 160, 166, 187, 212, 144, 165, 268; influences on science,
250 160, 165
Einstein, A., 22, 163, 165
Dahlberg, I., 44, 229 Elliott, A., 90
Dale, A., 133 Ellis, L., 103, 145
data, 4, 6, 38-41, 45-50, 63, 133-4, 152, Elster, J., 31, 81
157, 183, 206-7, 210, 223, 244-6; empiricism versus rationalism, 79, 107
defined, 45-7; desirability of using engineering; See technology
different types of, 19, 46, 50, 106, environment, non-human, 25, 29, 38, 46,
108, 200, 207, 244; empirical 58, 165, 213
regularities in, 47, 52, 78, 172-3; five errors and biases, defined, 158-60
Index 281

ethics, 98, 127, 129, 174-7, 181, 183, 186, Getman, J., 164
194-8, 213-4, 231, 245, 248, 262, 266- Giddens, A., 85-6, 92
8 Giere, R., 161, 163, 166, 211
ethnography; See observation Glassick, C.E., 163, 178, 180, 183, 218,
evaluation research, 102, 141-2 258-9
evolution, 44, 59, 81-3, 87, 89, 94, 117-8, golden mean, 8; applied, 38, 40, 59-60,
121, 156, 162, 165, 183, 186, 188, 62, 71, 79, 107, 109, 137, 144, 156-7,
198, 210 159, 161, 168, 259, 261; described,
experience, 102, 107, 124-5, 138 16-9; summary of arguments, 247-8
experiments, 101, 104, 112-3, 116, 118- Goldenberg, S., 4, 52, 105, 113, 115-6,
21, 130, 132, 138, 142, 148-9, 152, 122, 127, 129, 137, 143, 203
154, 159-60, 201-4, 215, 235-6 Gomm, R., 112, 134, 158-60, 196
Gower, B., 17, 40, 106, 121, 144, 172,
Fabian, J., 13 187
feminist theory and method, 51, 54, 84, Graff, G., 35, 262-3
90, 102, 111, 120, 122, 166 grounded theory, 91, 206-7
Feuer, L., 161, 190
Feyerabend, P., 170, 202, 212 Haack, S., 183
field theory, 58 Habermas, G., 88, 90, 92
financing of research, 155, 166, 180-2, Hall, D.J., 7, 102, 109, 142, 151
185-6, 197, 215-6, 234, 247 Hall, P.A., 200
Fiscella, J., 225 Halperin, D.F., 41, 106, 162, 168, 254
Fish, S., 21, 267 Hammersley, M., 112, 134, 152, 158-60,
5W questions; advantages of, 3-4, 10-1, 196
14, 37, 102-3, 109, 144-51, 192-3, Hartman, J.J., 103, 145, 147
195, 197, 205-6, 220, 226, 232-3, 241- Hawthorn, G., 71
5, 253; and hypotheses, 78-9; applied, Hawthorne effect, 120
5-7, 37-40, 48-50, 55-75, 77, 79-82, health, 25, 29, 58, 165
98, 104-9, 116-44, 161-92, 195, 220; Hedstrom, P., 126
described, 3-5 Herda, E.A., 129, 180
Fleming, A., 120 hermeneutics, 81, 102, 142-3
Foucault, M., 165-6, 170 Hintikka, J., 99
Fox, G., 176 history, 63
Fraser, C., 148 Hjorland, B., 222
Free will, 85-6, 88, 90, 112 Holton, G.J., 156
Freeman, D., 156 honesty, 176, 181, 196
Freidheim, E.A., 54-5, 57, 59, 69, 72, 74, Horn, R.E., 237
76, 95, 104, 200 Hough, G.A., 5
Fuller, S., 1, 70, 100, 121-2, 148, 151, Hubenthal, U., 21
157, 162, 166, 173, 182-3, 186, 208 Hull, D.L., 26, 31, 87, 95, 177, 183, 216
functionalism, 54, 81-2, 84, 87-90, 93-4 human science, 14, 33, 59, 73, 80, 83,
134, 140, 142, 146, 158, 167, 172,
Gaff, J.G., 36, 262, 264, 267-8 197-8, 205, 236; defined 6; See also
Galison, P., 1, 20, 22, 167, 173 humanities; natural science; social
game theory, 92-3, 126 science; various disciplines
genetic predisposition, 24, 27, 30, 33, 56, humanities, 3, 42-3, 51, 101-2, 175, 214,
58, 60, 66, 86, 93, 165, 213; See also 250, 261
perception and cognition Hume, D., 4
geography, 147 humility, 156, 163-4, 186
geology, 26, 140, 201, 207 Hunt, M., 153
282 Index

Hunter, B., 51 Kincaid, H., 18, 33, 59, 65, 70, 81, 155,
Hunter, E.J., 221, 224, 232 171, 176, 188, 207, 252
hypotheses, defined, 78-80 Kirschner, P., 237
Kitcher, P., 18, 78, 156, 174
idiographic; See nomothetic Klee, R., 48, 63, 79, 202
incentives facing scientists, 177-81, 189, Klein, E., 55
192-3, 197, 214, 218, 234-5, 258-9 Klein, J.T., 14, 20-1, 35, 41, 204, 255,
induction; problem of, 5, 152, 160; versus 265, 268
deduction, 2-3, 5, 10, 15-6, 25, 41, 44, Klein, L., 180
49, 62, 66, 81, 95, 101, 106-7, 110, Kontopolous, K., 72
115, 119-20, 124-5, 127-8, 132, 140-1, Krieger, M.H., 23, 46, 57-8, 75, 104, 120,
143, 161, 192, 194, 201-3, 224, 255 190
institutional structure of science, 160, Kruglanski, A., 162
163-4, 174, 177-86, 189-93, 245, 252, Kuhn, A., 30
258-9 Kuhn, T.S., 106, 155, 187, 208, 210
institutions, 28, 30, 57-8, 71, 74, 87, 89-
90, 214 Lakatos, I., 52, 106, 174, 212
interdisciplinarity, 9, 19-22, 38, 80, 83, Lambert, K., 99, 170
90, 92, 96, 130, 134, 184, 189, 191, Langley, P., 49, 52, 111, 146, 171, 189
204, 214, 221-3, 225, 229-30, 234, Langridge, D.W., 13, 224
238, 244, 263, 268; benefits of, 14, 20- Lechner, F.J., 87-9
1, 107, 182-3, 194, 200-1, 203, 209, Lenoir, T., 173
214, 216, 242-3, 254-5, 258; boundary life history, 123
work, 30, 173; complement-ary to Little, D., 60, 70-1, 141
specialized research, 21-2, 30, 193, Longino, H.E., 95, 120, 251
210, 242-3, 251, 254-5; defined, 20, longitudinal analysis, 133, 135-6
155; guidelines for, 19-21, 45, 100,
109, 140, 146, 153, 179, 183, 210-1, Madden, E.H., 160
214, 255-7, 260, 264; teaching, 34-5, Magnum, D., 148
101, 268; value of classifications for, Mahner, M., 26, 38, 40, 45-6, 79, 98, 170,
20-2, 30, 34-5, 45, 92, 169, 188-9, 174
219, 255-7. See also disciplines Mann, T., 225
interpretivism, 6, 32, 37, 62-5, 78, 104-6, mapmaking, 102, 143-4, 149, 237; See
110-3, 149, 153, 175, 198, 201, 248 also causal links, mapping
interviews, 101, 104, 113, 116, 119, 122- Marcella, R.,217, 219-24, 228-9, 232, 238
3, 127, 129, 135-6, 138, 140-2, 146, Marxism, 34, 52, 54-5, 76-7, 84
148, 152, 159, 179 May, T., 122-3
intuition, 66, 84, 90, 92, 102, 124-5, 138, Mayer, T., 175
149, 152, 163-4, 172-3, 175, 181, 186, McCloskey, D.N., 131, 155, 167
194, 197, 249, 267 McKeon, R., 3, 30
McKim, V.R., 39, 131
Jackendoff, R., 45 McNeill, P., 104-5, 120, 123, 127, 133,
Jones, P., 60, 84, 201 141, 201, 204
journalism, 4-5, 102 Mead, M., 114, 116, 128
mechanics, 22, 57, 64, 208
Kant, I., 13, 99, 164, 194, 197 medicine, 152-3, 166, 181, 215, 234
Kantorovich, A., 74, 162, 183, 188-9 Meltzoff, J., 41, 49, 116, 119, 132, 235
Katzer, J., 157 Merton, R.K., 52, 157, 182
Kent, B., 1, 15, 169 Messer-Davidow, E., 14
Kervin, A., 175, 261 meta-analysis, 153-4, 229, 234-6, 242
Index 283

method, 4, 6, 8, 11, 19-22, 41, 43, ch. 4, nomothetic versus idiographic analysis, 1,
158-9, 161, 166-9, 173, 178-9, 183-5, 68-71, 77, 82-3, 85, 96, 98, 105, 108,
189, 213-6, 218-9, 225, 227-8, 235, 110, 116, 118, 120, 123, 126, 128,
261-2, 265; bias toward supporting 133, 136, 138-9, 141, 175, 200, 248,
particular theory types, 104, 109, 150- 250, 257
1, 189, 200, 242-3, 246; defined, 100, Norris, C., 19
144-5, 172, 245; desirability of
flexibility in, 19, 34, 36, 50, 90-2, 100, Oakley, A., 38, 103, 110, 113, 119-20,
103-4, 108-9, 115-7, 132, 144-54, 122, 148, 151-2, 170, 235-6
177-9, 181, 185, 188-9, 201-6, 208, observation (participant), 91, 104-5, 107,
210, 214, 236, 241-2, 245-70, 250, 113-4, 120, 127-9, 138-42, 146, 148-9,
252; evaluating, 113-6; existence of 152, 159, 179, 213
more than one, 100, 121, 144, 147, openmindedness, 18, 164, 171, 181, 187,
153, 171-2; range of applicability, 208-9, 211-2
103, 144, 212, 218, 228; reliability, organizations, 28, 57
50, 113-9, 121, 123, 125, 129, 134,
140, 142, 146, 152-3; role in Pahre, R., 48, 56, 96, 183; See also
definition of science, 170-2, 175; Dogan
validity, 50, 113, 115-8, 120, 123, 125, Palmer, C.L., 21, 84, 217-8, 222, 225,
129-30, 134, 140, 142, 146, 151-3, 234, 238
168; See also theory; various Palys, T., 49-50, 105-6, 108, 110-2, 122,
methods 128-9, 134, 146, 168
Mill, J., 105-6 paradigms, 8, 96, 152, 157, 187-90, 192,
models (mathematical), 102, 107, 125-6, 208-12, 216, 247, 252, 254
138, 149, 172, 200 Parsons, T., 57, 84, 86-90, 93
Mokyr, J., 47 Pasteur, L., 164, 202
Montgomery, S.L., 168 patriarchy, 42
Moran, J., 20-1, 51-2, 140 Peirce, C.S., 1, 15, 148, 169, 178
Morton, A., 211 periodic table, 2, 5, 9, 13-4, 26, 87, 118
Moser, P.K., 18, 212 Perry, W.G. Jr., 265
Motterlini, M., 52, 63, 156, 174, 212 Pervin, L.A., 148
Mowery, D.C., 180 perception and cognition, limits of, 45,
Mukherjee, R., 1 160-2, 170, 192, 207, 213, 230-1, 246-
7
narrative analysis, 123 personality, 1, 5, 24, 26-7, 30, 33, 38, 46,
natural science, 6, 56-7, 59, 82, 94, 140, 56, 58, 60, 64, 77, 86, 149; influences
184, 205, 215; compared with human on science, 163-6, 192
science, 1, 3, 59-60, 62-4, 69-71, 78, phenomena, 4-8, 14, ch. 2, 143, 169, 177,
102, 147-50, 154, 165-6, 186, 198- 196, 203-7, 213-6, 222-3, 232, 241-6,
202, 213-6, 221, 244-6, 248, 261; See 255, 264-6; classifying theory with
also biology; chemistry; geology; respect to, 53, 58-9; defined, 30-1, 34,
physics 38-9; induction and, 106-7, 120, 124-
necessity, 81, 88, 100, 132 6, 132-3; realizations of, 31-2, 37, 40,
network analysis, 135, 143 42, 56, 205-6, 215; role in definition
Neurath, O., 15, 173 of science, 170-2, 175; second and
neuroscience, 65, 95 third level defined, 25; triangulation
Newell, W.H., 4, 21, 83, 134, 255, 257 and, 152-3; unpacked, 24-7, 29, 31;
Newton, I., 11, 22, 57, 64, 96, 164-5, 188, versus events, 31; See also causal
208 links; various phenomena
Nicolescu, B., 31 phenomenology, 54, 94-5, 213
284 Index

philosophy, 4, 13-4, 16, 30, 39, 51, 53, Ragin, C.C., 10, 40, 73, 76, 81, 104, 132-
67, 152, 172, 174-5, 214, 231, 247-8; 3, 140-1, 152, 169, 205
of science, 1, 3, 6-7, 16-9, 26, 63, 68, Ranganathan, S., 219-20, 228-9
77, 79, 81, 95, 99-100, 107, 109, 121, rational choice theory, 54, 60, 70, 92-4,
149, 155, 157-8, 160-1, 171-2, 176, 151, 200, 229
192, 197, 206-8, 230; realist, 5, 32, 37- reality, existence of, 37-8, 111-3, 115,
8, 62, 74, 112 153, 247
physical traces, 102, 129-30, 139-40, 149 Reason, P., 142
physics, 1-2, 5, 17, 19, 26, 38, 40, 47, 57- refereeing, 178, 183-4
8, 72, 75, 104-5, 121, 149, 190-1, 196, relationships, 27, 57-8, 67, 84, 86-7, 89-
201, 208 91, 93, 95, 104, 118, 122, 135, 137,
Plummer, K., 84, 91-2 150, 166-7
Polanyi, M., 100 relativity, 22, 64, 188, 208
Pole, C.J., 112 religion, 28; and science, 156, 159, 165,
politics, 24, 28, 30-1, 33, 58, 86; influ- 198
ences on science, 118, 159-60, 165-9, representativeness; See sampling
186, 213, 246 rhetoric, 167-8, 192, 215, 245, 247, 253
political science, 33-4, 200, 205, 229 Root, M., 12, 198
Polkinghorne, J., 19, 79, 99, 100, 119, Rosser, J.B., 83
191, 196 Rota, G., 177
Popper, K., 52 Rothbart, D., 40, 120
population, 25, 29, 58 Rowley, J.E., 223-4, 233
positivism, 62-5, 77-8, 104-6, 110, 113, Rule, J.B., 2, 96-7, 202, 208
149, 170, 173, 201, 248
postmodernism, 34, 37-8, 42, 54, 62, 78, Salter, L., 20, 22, 41, 256
84, 97, 170, 241, 250 sampling, 104, 116-9, 123, 140-1, 159
Potter, G., 32, 62, 65, 71, 173, 203 Sandler, T.,42
Pounds, N.J.G., 143 Sartori, G., 42
practice, scientific, 4, 7-8, 11, 18-9, 21-2, Schaffner, K.F., 69, 147
32-3, 51-2, 63-4, 79, 100, 106-8, 111, schemas, 27, 66, 86, 93, 149, 164, 168,
115, 122, 142, ch.5, 213-6, 245-7, 211, 263
251-2; See also triangulation; various science; as phenomenon, 25, 29, 58;
biases and errors boundedness of, 97, 241; cumulative
protocol analysis, 129 development of, 4, 17-9, 21, 52, 80,
Prus, R., 36, 38, 78 96-8, 101, 106, 156, 158, 186-9, 191,
psychoanalysis, 54, 90, 94, 213 202, 218, 241, 244, 248-9; defined, 3,
psychology, 32, 40, 56-7, 65, 90, 112, 7, 39, 74, 77, 161, 169-76, 243, 249;
121, 148, 207, 252 existence of, 7, 16, 77-8, 160, 170-1,
public policy, 32-4, 55, 142, 156-7, 164, 173-5, 248; improving, 7-9, 15, 18,
166, 180, 186, 192, 201, 215, 235, 32-4, 39, 42, 50-1, 146, 160, 162, 168,
237, 252, 260, 266-8 171, 174, 177-8, 180, 183, 193, 197,
209, 218-9, 241-5, 249-59; normal
qualitative versus quantitative analysis, 1, versus revolutionary, 208-10; unity of,
40, 47, 63, 102-3, 105-6, 109-15, 120, 3, 14-6, 20-1, 23, 173,205, 241;
128-9, 131-2, 134, 140, 152, 154, 178, unpacked, 5-7, 41, 78, 155, 161, 171
200-1, 248 scientific communication, 2, 4, 21-2, 34,
Quine, W., 6, 79 42-5, 48, 97, 140, 183-4, 186, 188,
210, 212, 222, 227, 247; clarity and,
Rachels, J., 211 80, 129, 154, 171, 185, 241, 250, 253-
Index 285

4, 256-7; to public, 156-7, 186, 211, 197, 230; See also philosophy of
214, 234-6, 242, 249-50 science
scientific documents, classification of, 8, Stump, D.J., 32, 100, 190
13, 44, 53, 62, ch. 7; faceted versus sufficiency, 73-4, 76, 81, 88, 132
enumerative, 219-20, 223, 229, 232; survey articles, 150, 229, 234-5, 237, 242,
thesauri and, 222, 226-7, 233; value 263
of, 8, 34, 92, 98, 184, 205, ch. 7, 242- survey method, 101, 104, 112, 116, 129-
3, 255 30, 135-6, 139, 141-3, 149, 152, 159
scientific values, 160, 164-5, 168, 170-1, Svenonius, E., 217, 238
182, 196-8 Swift, D.F., 218
Scriven, M., 6 symbolic interactionism, 54, 85, 91-2, 94
Seale, C., 18, 38, 62, 110, 112-4, 116, systems theory, 54, 87-90, 94
152-3, 170, 181, 206, 209 Szostak, R., 5, 7, 12, 16, 22, 25-6, 32-3,
selfish motives of scientists, 161-3, 177, 36, 41, 46, 59, 65, 72-3, 75-8, 80-1,
196-7 83, 88, 93, 144, 151, 164, 166, 168,
sexual orientation, 12, 27, 51, 166 174-5, 177, 179, 184, 187-8, 190, 193-
Shattuck, R., 261 4, 196, 201-2, 211, 214, 222, 231,
Shipman, M., 150, 157, 253 254-5, 259, 266
Silverman, D., 114, 129, 131, 209
Simonton, D., 188, 207 Tashakkori, A., 46-7, 50, 109-13, 148,
simulation, 126 151-2
Sin, S., 140 Taylor, C.A., 169
Singleton, R.A. Jr., 23, 32, 80, 99, 102-3, teaching, 9, 32-6, 97, 99, 101-2, 146-7,
106, 115, 119, 130, 142, 145, 150-2, 178, 181, 183, 234, 236-7, 243, 252,
265 258-68; lifelong learning, 35-6, 243,
Sklar, L., 47, 72 260, 264-5
Smiraglia, R.P., 229 technology, 25, 29, 58, 144, 174, 180,
social influences on science, 118, 165-6, 192, 263
168, 213, 246 Tessler, M., 141
social science, 40, 42, 56, 60, 97, 130, textual analysis, 102, 113, 129, 136-43,
151, 180; See also human science; 148, 175, 214-5
various disciplines theory, 4-6, 11, 19-22, 33, 37-41, 43, 45,
social structure, 24, 26, 29, 33, 36-7, 57- 47, ch. 3, 158, 163, 166-9, 173, 183-5,
8, 90, 267 187, 189-90, 195, 213-6, 218-9, 225-8,
sociology, 59-60, 80, 101, 105, 156, 162, 235, 262, 265; cluster, 72, 75-6, 89,
214, 229-30 98, 202, 244; defined, 51, 74-8, 98,
stability versus change, 72, 88-9, 128 244; desirability of flexibility in, 19,
statistical analysis, 102, 112, 130-4, 139- 34, 36, 58-60, 65, 67, 70-2, 78, 95-7,
43, 146, 148, 150-4, 179, 200-1, 203, 177-9, 181, 189, 204-8, 210, 214, 246-
213, 227, 229 7, 250; evaluating, 78-80, 97, 244;
Stake, R.E., 141, 153 generalizability, 40-1, 50, 68-71, 79,
Stanley, T.D., 154 91; grand, 34-6, 52-4, 75, 160, 243,
Stocklmayer, S.M., 186 250-1; interaction with method, 8, 64-
Stoner, N.W., 157 5, 78, 80, 100, 102-3, 106, 108, 115,
Storr, A., 165 117, 141, 148, 150, 178-9, 192, 199-
Strauss, A., 91 203, 208-12, 246; range of
structuralism, 54, 81-2, 84, 87, 89 applicability, 69, 75, 78-80, 85-6, 96,
study of science, 1-3, 8, 19, 32, 148, 156- 98, 100, 106-8, 114, 202, 205-6, 209,
8, 161-2, 167, 173, 181, 188, 191-3, 212, 218, 226, 242-4, 257; role in
definition of science, 170-2, 175;
286 Index

scientific versus philosophical, 51-2; Weber, M., 84, 86, 196

triangulation and, 152-3; validity, 63, Weinberg, B.H., 217-8, 225
96, 106; See also method; various Weingart, P., 20-1, 38
theories Wellman, B., 179
Tinbergen, J., 96 White, H.D., 225
triangulation, 151-3, 178, 209, 242 Whitley, R., 1, 3, 8, 14, 23, 185-6, 190,
Turner, B.S., 54, 65, 84, 95, 98 216
Turner, S.P., 52, 95, 132, 210 Williams, R., 43
Willinsky, J., 211, 234-6
Urry, J., 71 Winch, P., 65
Winkler, E.R., 211
Vaitkus, S., 95 Wittgenstein, L., 7, 41, 100, 144, 172
Valiela, I., 47, 50, 118, 121, 132-3, 156-7, Wolfe, A., 197
162, 177, 180, 185, 207, 253 Woolgar, S., 38
Vaughan, R.J., 266
Velkley, R., 32 Ziman, J., 2, 7, 13, 18, 74, 78, 158, 161,
Vicedo, M., 46 167, 169, 175-7, 181-2, 189, 205-6,
211, 217
Ward, S.C., 156
 Information Knowledge and Science Management
1. C. W. Choo, B. Detlor and D. Turnbull: Web Work. Information Seeking and Knowl-
edge Work on the World Wide Web. 2000 ISBN 0-7923-6460-0
2. C. A. Bean and R. Green (eds.): Relationships in the Organization of Knowledge.
2001 ISBN 0-7923-6813-4
3. R. Green, C.A. Bean and S.H. Myaeng (eds.): The Semantics of Relationships. An
Interdisciplinary Perspective. 2002 ISBN 1-4020-0568-7
4. M. Huysman and D. de Wit: Knowledge Sharing in Practice. 2002
ISBN 1-4020-0584-9
5. B. Detlor: Towards Knowledge Portals. From Human Issues to Intelligent Agents.
2004 ISBN 1-4020-2053-8
6. A. Spink and B.J. Jansen: Web Search: Public Searching of the Web. 2004
ISBN 1-4020-2268-9
7. R. Szostak: Classifying Science. Phenomena, Data, Theories, Method, Practice. 2004
ISBN 1-4020-3094-0

springeronline.com