Aharon Kantorovich-Scientific Discovery PDF

cover
cover
title:
author:
publisher:
isbn10 | asin:
print isbn13:
ebook isbn13:
language:
subject
publication date:
lcc:
ddc:
subject:
next page >
Scientific Discovery : Logic and Tinkering SUNY Series in

Philosophy and Biology
Kantorovich, Aharon.
State University of New York Press
0791414787
9780791414781
9780585078342
English
Science--Methodology, Science--Philosophy, Creative
ability in science, Serendipity in science.
1993
Q175.K19 1993eb
501
Science--Methodology, Science--Philosophy, Creative
ability in science, Serendipity in science.
cover
If you like this book, buy it!
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/cover.html[15/01/2011 19:26:39]
next page >
page_i
< previous page
page_i
next page >

Page i
Scientific Discovery
< previous page
page_i
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_i.html[15/01/2011 19:25:48]
next page >
page_ii
< previous page
page_ii
next page >

Page ii
SUNY Series in Philosophy and Biology

David Edward Shaner, editor
< previous page
page_ii
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_ii.html[15/01/2011 19:25:48]
next page >
page_iii
< previous page
page_iii
next page >

Page iii
Scientific Discovery
Logic and Tinkering
Aharon Kantorovich
STATE UNIVERSITY OF NEW YORK PRESS
< previous page
page_iii
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_iii.html[15/01/2011 19:25:48]
next page >
page_iv
< previous page
page_iv
next page >

Page iv
Published by
State University of New York Press, Albany
1993 State University of New York
All rights reserved
Printed in the United States of America
No part of this book may be used or reproduced
in any manner whatsoever without written permission
except in the case of brief quotations embodied in
critical articles and reviews.
For information, address State University of New York
Press, State University Plaza, Albany, NY 12246
Production by Marilyn P. Semerad
Marketing by Dana E. Yanulavich
Library of Congress Cataloging-in-Publication Data
Kantorovich, Aharon. (Date)
Scientific discovery: logic and tinkering / Aharon Kantorovich.
p. cm. (SUNY series in philosophy and biology)
Includes bibliographical references and index.
ISBN 0791414779. ISBN 0791414787 (pbk.)
1. ScienceMethodology. 2. SciencePhilosophy. 3. Creative
ability in science. 4. Serendipity in science. I. Series.
Q175.K19 1993
501dc20
921766
CIP
10 9 8 7 6 5 4 3 2 1
< previous page
page_iv
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_iv.html[15/01/2011 19:25:49]
next page >
page_v
< previous page
page_v
next page >

Page v
in memory of my father
< previous page
page_v
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_v.html[15/01/2011 19:25:49]
next page >
page_vii
< previous page
next page >
page_vii
Page vii
CONTENTS
Acknowledgments
Introduction
xi
1
Part I In Search for Logic of Discovery

Chapter 1: Exposing and Generating
11
11
1.1 What is a Discovery?

16
1.2 The Products of Scientific Discovery
16
1.2.1 What Do Scientists Discover When They Look at the World?
17
1.2.2 Objects and Events Contaminated by the Scientist's Intervention
20
1.2.3 The Plasticity of Theories
27
1.2.4 Explanations, Problems and Solutions
29
1.3 The Kinds of Discovery Processes
29
1.3.1 Exposure
32
1.3.2 Generation
34
1.3.3 Poincar: The Poverty of Creation
35
1.3.4 Eureka Events and Unintentionality
36
1.4 The Creative Element in Discovery and the Issue of Realism
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_vii.html[15/01/2011 19:25:49]
page_vii
36
1.4.1 Discovery, Invention and Creativity
39
1.4.2 The Case of Particle Physics: An Active Look at Matter
44
1.4.3 Epistemological Realism: Construction, Transaction and Representation
Chapter 2: The Scope of Method
49
49
2.1 The Nature and Function of Method

49
2.1.1 Who Needs a Method?
52
2.1.2 What is a Method of Discovery Supposed to Do?
58
2.1.3 The Origin of Method
< previous page
page_vii
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_vii.html[15/01/2011 19:25:49]
next page >
page_viii
< previous page
next page >
page_viii
Page viii
60
2.2 Inferring and Reconstructing
61
2.2.1 Reasoning vs. Creativity
62
2.2.2 Discovery as Inference or Reasoning
68
2.2.3 The Quest for Certainty or: How Ampliative Inference Can Be
Converted into Deductive Inference
74
2.2.4 The Hierarchy of Material Logics
77
2.2.5 The Discovery Machine
80
2.2.6 Theory-Construction and Research Programs
85
2.2.7 The Calculus of Plausibility: Logic of Pursuit
92
2.2.8 Discovery as a Skill: The Invisible Logic
Chapter 3: Why did Taditional Philosophy of Science Ignore Discovery
97
97
3.1 The Distinction between the Context of Discovery and the Context of
Justification
97
3.1.1 John Herschel's Distinction: Consequentialism
99
3.1.2 Reichenbach's D-J Thesis: Generationism
101
3.2 Objections to the Distinction
101
3.2.1 Justification and Discovery are Inseparable
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_viii.html[15/01/2011 19:25:50]
page_viii
102
3.2.2 Justification is Not Aprioristic
104
3.2.3 Information about Generation is Necessary for Evaluation: Predictability
and Novelty
106
3.2.4 The Context of Generation Has an Epistemic Dimension
Part II Discovery Naturalized
The Prepared Mind: Cultivating the Unintentional
113
Chapter 4: Philosophy of Science: From Justification to Explanation
117
118
4.1 Normative Philosophy of Science: Justification Relativized

118
4.1.1 Instrumental Rationality: Science as a Goal-Directed Activity
119
4.1.2 The Dilemma of the Normative Methodologist and Goodman's Solution:
Rationality without Goals
122
4.1.3 From Justification to Explication
123
4.1.4 From Explication to Explanation: Paradigms of Rationality
127
4.2 From Description to Explanation
130
4.3 Explanatory Philosophy of Science
134
4.4 Normative Naturalism: Shallow vs. Deep Theories of Scientific Rationality
< previous page
page_viii
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_viii.html[15/01/2011 19:25:50]
next page >
page_ix
< previous page
next page >
page_ix
Page ix
134
4.4.1 Phenomenological Theories of Rationality
138
4.4.2 Explanatory Theories of Rationality: How the Is-Ought Fallacy is
Avoided
139
4.4.3 Ideal Theories of Rationality and the Competence-Performance
Distinction
141
4.4.4 The Therapist Model of Rationality and its Implications for Involuntary
Processes of Discovery
Chapter 5: An Evolutionary Theory of Discovery: In Search for the Unexpected
145
145
5.1 Evolutionary Epistemology: Taking Natural Selection Seriously

148
5.2 Blind Variation: The Principle of Serendipity
148
5.2.1 Are Scientific Discoveries Analogous to Blind Mutations?
153
5.2.2 The Evolutionary-Epistemic Significance of Serendipitous Discovery
157
5.3 Some Implications of the Principle of Serendipity
157
5.3.1 Predictability and Epistemic Profit
160
5.3.2 The Mathematical and the Social Media
160
5.4 Two Landmarks of Serendipity in Physics
160
5.4.1 Kepler: The Conscious Sleepwalker
163
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_ix.html[15/01/2011 19:25:50]
page_ix
5.4.2 Planck: The Reluctant Revolutionist

165
5.5 Serendipitous Discovery of Natural Phenomena
168
5.6 Cultivating Serendipity
Chapter 6: Intrapsychic Processes of Creation
173
174
6.1 A Psychological Theory of the Creative Process

174
6.1.1 The Chance-Configuration Model
176
6.1.2 Phenomena Explained by the Theory and Evidence for Its Support
178
6.1.3 An Associative Mechanism of Generating Chance Permutations
180
6.2 Implications of the Theory
180
6.2.1 Explaining Serendipity
182
6.2.2 Individual vs. Collective Creativity
184
6.2.3 Cultivating the Creative Potential
185
6.2.4 Multiple Discovery
Chapter 7: A Socio-Evolutionary Theory of Science
189
189
7.1 Epistemic Cooperation and the Social Dimension of Discovery
< previous page
page_ix
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_ix.html[15/01/2011 19:25:50]
next page >
page_x
< previous page
next page >
page_x
Page x
197
7.2 The Social Dimension of Blind Variation, Selection and Dissemination
199
7.3 Has Science Liberated Humankind from the Tyranny of the Genes?
199
7.3.1 Genetically Controlled Human Understanding
203
7.3.2 Transcending Our Natural Habitat
7.3.3 Two Patterns of Human Evolution:
206
(a) The Principle of Growth by Expansion
209
(b) The Coevolution of Human Action and Human Understanding
212
7.3.4 The Epistemological Significance of Cooperation in Science: The
Evolutionary Perspective
215
7.4 The Tension between Change and Stability
218
7.5 Implications for Discovery
218
7.5.1 The D-J Distinction Revisited
219
7.5.2 Cultivation: Preparing the Collective Mind
220
7.5.3 Strategies of Discovery
Chapter 8: Tinkering and Opportunism: The Logic of Creation
223
223
8.1 Evolutionary Tinkering in Science

227
8.2 Tool-Oriented Scientists: Intellectual Migration
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_x.html[15/01/2011 19:25:51]
page_x
229
8.3 Tinkering in Particle Physics
230
8.3.1 Symmetries without Dynamics
233
8.3.2 The Resources of Quantum Field Theory
235
8.3.3 Playing with Quarks
235
8.3.4 Tool-Oriented Particle Physicists
Chapter 9: Completing the Picture: Is There a Role for The Genotype-Phenotype
Process?
243
244
9.1 Non-Creative Discovery: The Genotype-Phenotype Logic of Growth

249
9.2 The Selection Cycle in Science
Conclusion
253
Epilogue: Implications for Science Education
255
Notes
259
Bibliography
261
Index
271
< previous page
page_x
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_x.html[15/01/2011 19:25:51]
next page >
page_xi
< previous page
page_xi
next page >

Page xi
ACKNOWLEDGMENTS
The seeds of this book were sown when I taught philosophy of science in the Institute for History and Philosophy
of Science at Tel Aviv University, which was my first philosophical home. I would like to thank Joseph Horovitz,
the first coordinator of the Institute, for his moral support over the years.
Later I returned to Tel Aviv University as the guest of Yuval Ne'eman, and we had a very fruitful collaboration. I
would like to thank him for his agreement to include in the book material from our paper on serendipity and for
providing partial support through the Wolfson Chair Extraordinary of Theoretical Physics.
At the preliminary stages of writing the book, during autumn 1989, I had a very stimulating break when I stayed at
the Center for Philosophy of Science at the University of Pittsburgh as a Visiting Fellow. I would like to thank
Jerry Massey, the director of the Center, for inviting me to the Center and for his generous hospitality. The
intellectual and social environment in the Center provided an atmosphere conducive to the development of my
work. In particular, I enjoyed the discussions I had with Sam Richmond.
I am grateful to Paul Levinson, Kai Hahlweg and Cliff Hooker for inviting me to the conferences on evolutionary
epistemology they organized, which provided stimuli for my work.
I also thank Samuel Goldsmith and David Naor for their interest and Nomy Arpali for her assistance.
Finally, I am indebted to Donald Campbell for his encouragement and to the three anonymous readers of SUNY
Press for providing some helpful suggestions.
Material from the following papers is included in this book:
1.Philosophy of science: from justification to explanation, published in the British Journal for the
Philosophy of Science (1988).
2.Serendipity as a source of evolutionary progress in science, which I wrote with Yuval Ne'eman and
which was published in Studies in History and Philosophy of Science (1989), reprinted with
permission of Pergamon Press Ltd.
< previous page
page_xi
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_xi.html[15/01/2011 19:25:51]
next page >
page_xii
< previous page
page_xii
next page >

Page xii
3.Naturalistic philosophy of science: a socio-evolutionary view, published in Journal of Social and

Biological Structures (1990).
4.A Theory of the creative process in science, published in Journal of Social and Evolutionary Systems
(1992).
5.A genotype-phenotype model for the growth of theories and the selection cycle in science, published
in Hahlweg and Hooker (eds.), Issues in evolutionary epistemology. SUNY Press (1989).
< previous page
page_xii
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_xii.html[15/01/2011 19:25:51]
next page >
page_1
< previous page
page_1
next page >

Page 1
Introduction
Unless you expect the unexpected you will never find truth, for it is hard to discover and hard to attain.
Heraclitus, Fragment 19 (Wheelwright 1966, 70)
Discovery and the Philosophy of Science
Scientific discovery seems to be the most impressive and mysterious feature exhibited by modern natural science.
From the material or practical point of view, it enables humankind to control, reproduce and predict natural
phenomena, to create novel phenomena and technologies and to expand its living space. From the epistemological
point of view, it immensely increases humankind's capacity for knowing and understanding itself and the natural
world. Yet, the philosophers of science, who attempt to understand science, have rediscovered this phenomenon
only recently after many years of neglect. Logical empiricism, the dominant school of twentieth-century
philosophy of science, regarded the study of the process of discovery as an empirical inquiry to be dealt with by
scientific disciplines such as psychology or sociology. The only respectable engagement of the philosopher of
science was considered to be the logical analysis of the products of scientific discovery. However, formal logic
proved not to be a very effective tool for understanding the phenomenon of science, and the discipline of
philosophy of science became almost irrelevant for the understanding of the peculiarities and the significance of
science.
Only in the last two decades, with the decline of logical empiricism and the emergence of new approaches to the
study of science, has the study of scientific discovery regained its respectability. It is widely believed now that
there is no universal logic of discovery. However, philosophers of science who deal with discovery moved from
one extreme to another: from the search for universal method to ''particularism." Some of them are now engaged
with particularities; they tell us about methods of discovery which they extract from specific domains and
particular contexts. This historicist-particularist movement mainly concentrates on case studies of discovery,
without trying to rise above this level of study and develop a deeper, or more general, theory of science.
< previous page
page_1
file:///C|/Users/Emiliano/Desktop/Scientific%20Discovery/files/page_1.html[15/01/2011 19:20:51]
next page >
page_2
< previous page
page_2
next page >

Page 2
According to this approach, science seems to be an ad hoc federation of loosely connected research areas. The
question arises as to what is the value of this occupation for understanding the phenomenon of scientific discovery.
Another trend in the study of discovery has developed in recent years. With the explosion in the field of computer
science and artificial intelligence, a new movement has emerged"mechanized discovery." However, this
technology-driven approach does not yield universal methods of discovery either. For the time being, some of the
proponents of this enterprise, the cognitive scientists, have succeeded in mechanizing only marginal stages of
discovering limited kinds of regularities. Moreover, they claim to have some success in concept formation. Others,
who are engaged in more practical directions, have made some contributions to areas such as medical diagnostics
and drug research. For example, we can find in a recent publication a description of such a product: "a drug
discovery software system that enables medicinal chemists to design realistic new molecules interactively;
construct, test, and refine hypotheses that explain and predict their bioactivity ..." (Science, 28 Feb. 1992, Products
and Materials, 1153). This kind of tool may serve as a useful technological aid for conducting research in a
specific area, which may have implications for a heuristic-guided discovery in general.
In this book I have attempted to proceed in a different direction. Instead of looking for a universal logic of
discovery, I treat discovery as a universal phenomenon. Instead of mechanizing discovery, I attempt to naturalize
it. I do this within the naturalistic approach which views the philosophy of science as a "science of science." This
approach will allow us to treat those facets of discovery which have escaped the dissecting tools of the logician.
Yet, since the naturalistic tools have not yet been crystallized, I have tried to develop some of them in this book.
This situation is symptomatic of the present state of the field. Indeed, the philosopher of science is very frequently
engaged in the meta-level problems, as well as in the object level problems; he is evaluating and reconstructing his
tools of investigation, while using them.
The main message of this book is that the creative process of discovery is not a purely rational enterprise in the
traditional sense which equates rationality with logical reasoning. Yet, although it is not governed by logical rules,
it is a manifestation of universal phenomenon which may be treated as a natural phenomenon in its own right. In
studying discovery, I shall transcend the traditional boundaries of the philosophy of science and incorporate into
the theory of discovery ideas from evolutionary biology, psychology and sociology.
Mechanisms of Natural Selection in Scientific Creation
The theory of creative discovery which I propose belongs to the family of evolutionary models, appearing under
the title "evolutionary epistemology" (EE)
< previous page
page_2
next page >
page_3
< previous page
page_3
next page >

Page 3
(Campbell 1974b). 1 I suggest that creative processes in science are governed by two interrelated natural selection
mechanisms: the intrapsychic (subconscious) process of creation, such as the process of incubation, and the
interpsychic process of epistemic cooperation. These processes can be categorized as unintentional or involuntary.
The incubation process is well known, but is not well understood. We cannot control it since it is involuntary. Yet,
we can cultivate it (I will explain this notion below). This phenomenon can be explained by the model of natural
selection which is applied to mental elements created quasi-randomly in the discoverer's mind (Simonton 1988);
the discoverer hosts the process in his mind, so to speak. The second creative process takes place in a sociohistorical setting which fosters unintentionality and serendipity (Kantorovich and Ne'eman 1989); epistemic
cooperation generates "blind" variations, or unexpected discoveries. Here, too, the scientist is not fully in command
of the process. Yet, epistemic cooperation can be cultivated.
The notion of tinkering sheds further light on the above facets of scientific creation. Levi-Strauss (1962) introduced
this notion in describing savage thought and Francois Jacob (1977) borrowed it for characterizing evolutionary
progress. I suggest that an inevitable consequence of the natural-selection model is that the creative steps in the
evolution of science are the products of tinkering. And this implies that scientific creation is not method-governed
and that science has no predetermined goal. The notion of tinkering encompasses all kinds of unintentional,
serendipitous and opportunistic processes of scientific creation. In this book, I intend to elevate the phenomenon of
tinkering in science from the level of anecdotes and curiosities. I hope to show that it is part and parcel of the very
nature of scientific discovery and human creativity in general. Scientists rarely include descriptions of tinkering in
their scientific writings. In popular stories and autobiographies, scientists sometimes do include such descriptions
in order to entertain the reader. This reflects the official attitude of traditional philosophy of science towards this
phenomenon; cases of tinkering should be hidden from the public, since they contradict the ethos of science.
Science should appear as a nice and neat rational enterprise. This is one of the reasons why the nature of scientific
creativity is wrapped in a shroud of mystery.
Yet, reason is an indispensible part of the process of scientific discovery. Inference and reasoning are employed for
exposing the far-reaching consequences of new ideas and theories and for evaluating them. Furthermore, there has
been an attempt which can be traced back to Francis Bacon and John Stuart Mill, to represent all kinds of
discovery processes as rule-governed processes. Part I is devoted to the role of reason in discovery and to the
attempts of representing discovery as inference. In Part II, I develop an evolutionary theory of creative discovery. I
complement the evolutionary picture by a conjecture about the evolutionary role of non-creative processes of
discovery.
< previous page
page_3
next page >
page_4
< previous page
page_4
next page >

Page 4
Discovery as Inference
One of the main themes which is explored in Part I is the distinction between two main notions of discovery:
discovery by exposure and discovery by generation. Examples of exposure are discovery by observation or by
deductive inference; in the former we expose objects, such as a new star or particle, in the latter we expose the
information hidden in a set of statements, such as a prediction derived from a new theory. Examples of
generational discovery are discoveries made by active experimentation and theory-construction. I argue that the
distinction between these two kinds of discovery processes is sharper and more useful than the traditional
distinction between discovery and invention. The discovery-invention distinction makes sense only if we refer to
discovery by exposure. However, in science, the meaning of the word discovery has long transcended its original
etymological origin which refers to exposure. Newton's law of universal gravitation, quantum mechanics and the
theory of natural selection, for example, were not literally dis-covered or un-covered; they were generated. I
discuss the implications of the exposure-generation distinction for the issue of realism.
The central questions discussed in Part I are: to what extent is scientific discovery regulated by reason or method?
and what is the role of creativity in the various discovery processes? The various methods and rules are presented
in the order of decreasing strength. The more creative is the process or the act of discovery, the weaker is the
method for generating it.
I describe different kinds of methods which stem from the inference-view of discovery and, in particular, the
attempts to convert ampliative, or content-increasing, processes into deductive inference. This leads to the issue of
material, or content-dependent, logic of discovery and to community-specific logics and domain-specific methods.
This subject is directly related to the issue of expert systems in mechanized discovery. When generational
discovery, such as theory construction, is viewed as deductive inference, it becomes a discovery by exposure. Thus,
the inference-view is intimately related to the view which treats all kinds of discovery processes as processes of
exposure. Since inference is in principle method-governed, the search for a method of discovery stems from the
exposure-view of discovery. Yet, no discovery process in empirical science is entirely guided by mechanical
algorithm so that creative steps always play some role in the process.
Some attempts to represent generational processes of discovery as method-governed can be categorized as
"postmortem" procedures. Typical examples are Peirce's and Hanson's logic of retroduction, Musgrave's "inventive
arguments" and Simon's discovery machine, which reconstruct the discovery process from the vantage point of one
who benefits from the knowledge of the final result. They help us, therefore, only in justifying or reproducing
something which has already been generated.
< previous page
page_4
next page >
page_5
< previous page
page_5
next page >

Page 5
Another postmortem procedure is the logic of pursuit which deals with the plausibility of a hypothesis and with the
question of whether or not a hypothesis is worth pursuing. This logic of discovery is a method of initial evaluation
of the product of discovery, rather than a method of generating the product. However, evaluation is an integral part
of the process of discovery, so that the logic of pursuit should not be dismissed as a partial method of discovery.
Finally, I treat discovery as a skill. Skill is a predominantly tacit faculty. Some basic field-specific presuppositions
are internalized by the discoverer and serve as suppressed premises in the process of discovery. These premises can
be treated as "transparent" or "invisible" so that the process of discovery looks like an inference or an exposure.
Thus, the following distinction can be made between generational discovery and discovery by exposure.
Generational discovery creates new conceptual or experimental tools, which can be viewed as new communication
channels with nature. When these channels become transparent, new phenomena and new aspects of reality may be
exposed through these channels. This kind of discovery by exposure can be viewed as inference, where the
suppressed premises used by the skilled discoverer are his communication channels with reality.
One of the contributions of traditional philosophy of science to the subject of discovery is the distinction between
the context of discovery and the context of justification. This is one of the controversial theses of logical
empiricism, which was one of the major reasons for ignoring discovery. It stems from the inference view of
discovery. I discuss some main objections to the distinction thesis. In a naturalistic philosophy of science, the
distinction vanishes and the process of discovery regains its legitimacy as a proper subject.
When we view discovery as a skill, we encounter for the first time a process of discovery which is partially
involuntary. This is predominantly a discovery by exposure. Yet major kinds of generational or creative processes
discovery are involuntary processes: the incubation process, the eureka event and the cooperative-historical process
of discovery. What is common to these processes is that the discoverer is not in full command of the process.
Traditionally, method-governed discovery has been contrasted with so-called chance discovery. The notion of
"chance" or "accidental" discovery is employed whenever the process of discovery is unintentional. But this does
not mean that there is no explanation for this kind of discovery. For instance, discovery may be a subconscious
process, or a cooperative process, which can be described and at least partially explained by psychological and
sociological theories. In these cases, the discoverer does not generate the product of discovery. The discoverer can
only cultivate and expose it. Cultivation can be guided by recommen-
< previous page
page_5
next page >
page_6
< previous page
page_6
next page >

Page 6
dations for the discoverer, which are similar to the recommendations given to the farmer who grows plants or
animals, or to the educator who "cultivates" children. From the viewpoint of the discoverer's contribution to the
discovery, the act of cultivation is weaker than the act of intentional generation. In the latter case the discoverer
generates the product of discovery, whereas in the former the discoverer contributes to the process by generating
appropriate conditions for it and by nourishing it. Following the phrase "chance favors the prepared mind," coined
by Louis Pasteur, we may interpret cultivation as preparing the mind for unexpected discovery. Qua-cultivator, the
discoverer acts here as a spectator at the act of creation.
The highest degree of creativity is exhibited in unintentional or involuntary processes, which are not governed by
method. Although we may expect the highest degree of novelty to be created by processes of this kind, the
philosophy of science has been totally ignorant of them. In Part II, I will suggest how this deficiency may be
corrected.
Discovery as an Evolutionary Phenomenon
The explanatory or naturalistic approach to the philosophy of science is most appropriate for treating the
involuntary or unintentional processes of discovery. Part II is devoted to these phenomena. Instead of being
regulated by methodological rules, these processes of discovery may be governed by natural laws, or explained by
a theory of discovery. Thus, naturalism would provide us with an alternative to traditional logicism, on the one
hand, and particularism or historicism, on the other.
I start with meta-philosophy-of-science. In order to comprehend the significance of the evolutionary approach to
science in a wider perspective, I develop my version of the naturalistic or explanatory approach to the philosophy
of science. I present a general view on the nature of the philosophy of science. In this framework I classify
different approaches to the philosophy of science, such as logicism, historicism, sociologism, cognitivism or
evolutionism. I call these various approaches paradigms of rationality. The main difference between this scheme
and most contemporary historicist and naturalistic treatments of science is that the latter characterize a theory of
science as (merely) descriptive, whereas I emphasize its explanatory role. I introduce here a scheme which
provides what I would call a "deep" naturalistic theory of scientific rationality. This contrasts what I would term
"shallow" theories, which draw their rules of rationality from the practice of science without trying to find a deeper
explanation for them. The latter category includes historicist theories. (The word "shallow" is not used here in its
negative connotation; in the same sense, phenomenological laws or models in physics are shallower than the
theories which explain them). For example, I not merely pursue the
< previous page
page_6
next page >
page_7
< previous page
page_7
next page >

Page 7
claim that scientific creativity is dominated by tinkering, I furthermore attempt to provide a deep explanatory
theory of science which will explain this phenomenon. In fact, it was the theory which came first and the
phenomenon of tinkering was then predicted by the theory.
The heart of Part II is the evolutionary theory of science which I propose. In general, it might be categorized as a
version of EE, which models the development of science on natural selection. The theory views science as a
continuation of biological and cultural evolution. In applying this theory to scientific discovery, I present the view
expounded in my paper with Ne'eman which has far-reaching implications for the nature of scientific discovery. It
presents our interpretation for the notion of "blind" variation in science, modeled on the notion of quasi-random
mutation in evolutionary biology. According to our view, one of the most important kinds of creative discovery
made in science are serendipitous discoveries, which are made when scientists unintentionally solve a problem (or
explain a phenomenon), while intending to solve a different problem (or to explain a different phenomenon). We
claim that the phenomenon of serendipity is essential to scientific progress, especially to revolutionary progress.
Serendipitous discovery is a typical unintentional process. The discoverer can only cultivate the process, be
prepared for its unexpected outcome and expose it. Serendipitous discovery, like a biological mutation, can be
explained as an "error" which infiltrated a routine procedurea research program.
Serendipitous discovery is generated by psychological and social processes. I introduce Simonton's psychological
theory of creative discovery which accounts for some major kinds of serendipitous discovery. The theory is based
on the natural-selection model and describes processes which take place in the individual's mind. This theory
attributes an element of chance even to the seemingly most intentional processes of discovery, such as Einstein's
discovery of special relativity, which would otherwise undermine the blind-variation view of discovery. It explains
an important kind of serendipity-generating mechanism which is related to the creative process of incubation.
Thus, according to this theory, the most "intentional" processes are partially involuntary.
Then, I complement the evolutionary theory of science by the social dimension. I claim that the social dimension is
essential to both the phenomenon of serendipity, i.e. to "blind" variation, and to the process of selection.
Furthermore, the social dimension enables science to transcend our evolutionary heritage and to expand our
knowledge into new domains, such as the domains of cosmic and microcosmic phenomena. This approach
combines elements of social epistemology and evolutionary epistemology. In the framework of this approach, I
propose two evolutionary patterns, or models, for the growth of scientific knowledge. One pattern may be labeled
the principle of growth by expansion, which is exhibited in biological evolution when a species
< previous page
page_7
next page >
page_8
< previous page
page_8
next page >

Page 8
solves problems arising in its original niche by expanding into a new living space. This model accounts for the
manner in which science solves problems by active experimentation and theory-construction, thus expanding into
new domains of reality. The second pattern is the coevolution of sensorimotor organs and the brain which is
reflected in science by the coevolution of observational-experimental techniques and the theoretical apparatus.
As has been noted above, the notion of tinkering encapsulates the evolutionary facets of scientific creativity. In
applying this theme to science as an evolutionary phenomenon, new light is shed on some chapters from the
history of science. I offer examples which illustrate this aspect of discovery. To make the case stronger, I bring
some evidence from what is considered to be one of the most advanced natural sciences: theoretical physics. The
main piece of evidence will be drawn from the history of particle physics which demonstrates the role of tinkering
in generating novelty. In other words, I try to interpret episodes from the history of particle physics as the work of
a tinkerer.
Yet, creative discoveries do not fall from the sky. Radical innovation comes from within science; the source of
scientific novelty lies in routine research. Moreover, the process of consolidating the ideas generated in the creative
process is by itself predominantly non-creative. The evolutionary model, therefore, cannot be complete without
leaving room for routine research programs which are characterized by inference and reason. In the last chapter, I
conjecture that the genotype-phenotype structure underlying ontogeny will do the job.
The highly abstract approaches to discovery which are expounded in this book yield down-to-earth consequences.
The notion of the scientist as a tinkerer is one of them. They also have practical implications for society in two
areas. The principles of serendipity and tinkering have important implications for science policy and science
education; the former are indicated in Chapters 5 and 8 whereas the latter are outlined in the epilogue.
< previous page
page_8
next page >
page_9
< previous page
page_9
next page >

Page 9
PART I
IN SEARCH FOR LOGIC OF DISCOVERY
< previous page
page_9
next page >
page_11
< previous page
page_11
next page >

Page 11
Chapter 1
Exposing And Generating
Nature loves to hide. The hidden harmony is better than the obvious.
Heraclitus, Fragments 17 and 116 (ibid., 70, 79)
1.1 What is a Discovery?
I would like to start by trying to clarify or explicate the intuitive notion of discovery as it is used in everyday life
and in science. This may provide us with clues for understanding the epistemological and pragmatic significance of
discovery in general and of scientific discovery in particular. We may find differences among the usages of the
term in everyday discourse and in science, as it happens also with scientific terms which are borrowed from
ordinary language. This time, however, we are dealing with a concept which belongs to the metalanguage of
science rather than to science proper. 2
The focus of this book is on the process or the act of discovery. However, we cannot deal with the process of
discovery without also considering the object or product of discovery. For example, we will be engaged with the
issue of the ontological status of certain scientific discoveries; e.g., whether a certain (product of) discovery is a
real entity which exists independently of the inquiring mind or whether it is our own creation.
Discovery is a "success" word. When we say we have discovered something, it means, for example, that the
product of discovery is useful, that it solves a problem, explains some phenomena or that it is the lost object we
have been looking for. When one tells us he saw a flying saucer, we would not say that he had discovered a flying
saucer unless it was proved that what he
< previous page
page_11
next page >
page_12
< previous page
page_12
next page >

Page 12
saw was indeed an extraterrestrial vehicle. We would not say that the magnetic monopole was discovered, since its
existence has not been confirmed by experiment. Phlogiston was regarded as a discovery by Stahl and his
contemporaries. But since Lavoasier's revolution in chemistry, the history of science treated it as a false theory.
The same can be said about the status of the aether after Einstein's revolution in physics.
A contemporary scientist who does not believe in the existence of the aether or phlogiston, would not say that the
aether was discovered. But a historian of science who is aware of the fact that theories may be overthrown, that an
overthrown theory may be revived in the future and that contemporary theories are not the final truths, would
relativize the notion of discovery to a historical period and to some community.
We would be interested in the process of discovering a theory. This encompasses the stage of evaluation and
confirmation in which the scientist finds out that the theory is successful in providing predictions, explanation,
understanding or unification of phenomena. I do not use the term confirmation in a logical sense of proof. No
theory can be proved in this sense. A confirmed theory may be overthrown or refuted (and refutation in this sense
is not a logical notion either). Moreover, most scientific theories have been refuted. Both the theory of the aether
and Kepler's laws were refuted. So why the latter are considered to be a great discovery and theory of the
phlogiston is not? I will turn to this question below.
Epistemological Aspects of Discovery
From the epistemological point of view, discovery is a major vehicle for the growth of knowledge. And yet, our
knowledge grows also by other means. As individuals, perhaps the main way we learn new things or acquire new
information is by instruction or by reading. Even, and in particular, scientists learn most of what they get to know
from other scientists or from the scientific literature. Personally, a scientist may make very few scientific
discoveries, if any, during his lifetime; most scientific discoveries are products of collective efforts. Therefore, it is
sometimes difficult to judge who participated in, or contributed to, the discovery; sometimes it is perhaps a whole
community which should be credited. If the process extends over a long period of time, only the final step in the
process is regarded as a discovery. Yet the contributions of the other participants are sometimes no less important
than the contribution which constituted the breakthrough. The example of the electroweak unification in particle
physics, which is discussed in Chapter 8, illustrates this point. The graduate student 't Hooft, who made the
breakthrough which led to the theory, entered the field at a stage where almost everything was ready for the
discovery, and he solved the crucial problem which was presented to him by his supervisor, Veltman. So, should
we regard him as the discoverer of the theory, or only as the discoverer of the solution of the specific problem
which his supervisor asked him
< previous page
page_12
next page >
page_13
< previous page
page_13
next page >

Page 13
to solve? This issue will be elaborated when the social dimension of discovery will be discussed.
On the other hand, an individual may discover something which would not count as a scientific discovery since it
does not constitute a novelty with respect to the body of knowledge or to the system of beliefs shared by the
scientific community. For example, someone might ''discover" today that the earth rotates around the sun. We
should refer, therefore, to the total, or collective, knowledge of a culture or a community, such as the scientific
community. This would lead us to the following characterization of the act of scientific discovery: the acquisition
of an item of knowledge which constitutes an increment in the body of knowledge of the scientific community.
One may discover the neutrino, a black swan, the theory of general relativity, that phlogiston theory is false, that
leaves change their color, or that Mary has eaten her breakfast. The last example, however, does not seem to be a
scientifc discoverynot even an ordinary discovery. In order to exclude cases such as this, we might require that the
process of discovery will result in a new knowledge item, in the above sense, which is either (a) unexpected, (b)
has a special interest or (c) constitutes an increment of general knowledge or a change in our general beliefs, as
opposed to beliefs in particular matters. Thus, seeing a black swan may be unexpected since it may contradict a
general entrenched belief (that all swans are white). A discovery may contradict an implicit generalization of which
we become aware as a result of the discovery. Hence, discoveries may lead to "negative" as well as to "positive"
increments of knowledge; we may acquire a new generalization, or learn that some of our general beliefs are false.
All the above cases exemplify discoveries made with respect to a given body of general beliefs; a discovery is a
process or an act which adds something to our system of general beliefs, changes it or solves a problem which
arises in it.
In the above discussion, a sharp distinction was not made between knowledge and belief. A traditional
epistemological characterization of knowledge, which can be traced back to Aristotle, is expressed by the familiar
slogan: "knowledge is justified true belief." There has been a long debate centering on this definition in twentiethcentury epistemology. Although it is out of the scope of this book to dwell in depth on this issue, we must take
sides in the debate since discovery is a salient epistemological notion.
We do not have to accept the Popperian conception of knowledge in order to reject the above definition. Popper, in
defiance of the above definition, denies that our beliefs can be justified, hence all our knowledge is conjectural.
However, even if we admit some sort of justification or confirmation, we still cannot retain the qualification "true"
in the above definition without excluding practically most of science from the realm of knowledge. If we do not
want to exclude, for example, Newtonian physics from the corpus of eighteenth- or ninteenth-century scientific
knowledge, or quantum physics
< previous page
page_13
next page >
page_14
< previous page
page_14
next page >

Page 14
from the corpus of twentieth-century scientific knowledgein fact, if we want to retain the notion of scientific
knowledgewe cannot accept the traditional definition. Scientific theories can be confirmed but cannot be "proved"
beyond any doubt to be true. Newtonian theory, which was perhaps the most established scientific theory in all
times, confirmed by so many observations and experiments, turned out to be strictly false. Thus, according to the
traditional definition, it would not qualify as knowledge. We would not be able to say that Newtonian mechanics
was part of eighteenth-century scientific knowledge, for example. And there is no reason why the fate of twentiethcentury physics would be better. In science, we accept not-yet falsified conjectures as knowledge, provided they
are well confirmed, in some non-logical sense of the word. Well-confirmed theories or laws may be regarded as
"partially true" or may be treated as good approximations to the truth. Scientific knowledge consists mostly of
partial truths and good approximations to the truth. Indeed, we may find out that something we have believed in
and which has been treated as legitimate knowledge is false, or true only in a restricted domain.
No one would deny that Kepler's discovery of the laws of planetary motion was a real discovery, although the laws
turned out to be strictly false. The reason for this is that in hindsight Kepler's laws are considered to be
"approximately true." This means that the discovery of the laws constituted an indispensable step in a progressive
research program (to use Lakatos' terminology, to which I will turn in section 1.3) which led to the highly
confirmed Newton's theory from which a revised version of the laws was derived. The revised laws are considered
to be a better approximation to the truth. On the other hand, the "discovery" of the aether is not treated by
historians of science as a real discovery, since it did not yield a progressive research program. Both Kepler's laws
and the theory of the aether were strictly false, hence, it is not the notion of truth which distinguishes between
them. Rather, it is a notion such as fruitfulness, or progressiveness, which may account for the distinction. Thus,
only in hindsight can we say that the discovery of Kepler's law was a real discovery. Hence, paradoxically, the
notion of (absolute) truth does not prove to be helpful in characterizing knowledge. A discovery of a "false'' law or
theory might turn out to be an important step in the growth of scientific knowledge. Knowledge is construed here
as a dynamic entity rather than as a static entity which is either true or false. The dynamic character of scientific
knowledge and scientific discovery will be discussed in section 1.3.
According to Popper's falsificationism (Popper 1959), theories do have truth values, and the only possible
theoretical discovery is the discovery that a theory we have (irrationally) believed to be true is false. Thus, we may
say that according to Popper, scientific discovery is not a "success notion," but a "failure notion"contrary to our
intuition. Perhaps the only "positive" discovery possible, according to this view, is a discovery of a method of
testing a theory.
< previous page
page_14
next page >
page_15
< previous page
page_15
next page >

Page 15
A "discovery" might turn out to be a fake discovery. One may "discover" an oasis which turns out to be a mirage.
Since the mirage is not a useful phenomenon, nor of any interest to someone lost in the desert, we would not
qualify such an event as a discovery. To be sure, the discovery of the (general) phenomenon of the mirage is by all
means a real discovery. A similar pattern of discovery was the discovery of the American continent. This time,
however, the final result was of great interest and useful. When Columbus arrived at the American continent, he
thought he had discovered a short route to India. Only later, with the identification of the object of discovery as a
new continent, was the discovery completed. If A "discovers" X but mistakenly identifies it with Y, we cannot say
that A discovered X, even if X turns out to be useful and of interest. Thus, a discoverer should realize the
significance of his discovery, in particular, its usefulness or its special interest. As I have said, a discovery
amounts to an acquisition of a new item of knowledge, not just to an encounter with a new object or phenomenon
without identifying it or realizing its significance. Knowledge of X implies the identification of X as X. We would
not say that someone who is ignorant of zoology and who encounters a bird of a species unknown to zoologists,
without realizing its significance, discovered this bird. Yet, it may happen that the person who saw the bird started
a chain of events leading to the identification of the new bird, for example, by reporting his observation to an
ornithologist. In this case we may say that the first observer contributed to the discovery.
In the case of discovering the American continent we have a similar pattern of discovery. However, Columbus is
considered to be the discoverer of America since, as in Kepler's case, he made the initial (and decisive) step in the
process of discovery. The hard work had been already done by Columbus when the final interpretation to the
discovery was given. Another example is the discovery of penicillin by Fleming. Many researchers, including
Fleming himself, had encountered the same phenomenon of a bacterial culture contaminated by mold, without
realizing its significance and usefulness. None of them would be considered as the discoverer of penicillin,
although they encountered the same phenomenon. The discovery of X-rays was missed by several scientists. Sir
William Crookes, was experimenting with an evacuated glass tube in which negatively charged cathode was
embedded. He observed that some of the photographic plates enclosed in containers became fogged when the
Crookes tube emitted cathode-rays, but he did not attach importance to this finding. Philip Lenard, who modified
Crookes tube by inserting at the end of the tube, opposite to the cathode, a very thin alluminium window, also
missed the discovery of X-rays. In 1894 he observed that about 8 cm. outside the window there appeared a glow on
a paper coated with platinum cyanide. He interpreted this as resulting from a flow of "streaming electricity" (Kohn
1989, 16). Roentgen did not provide an explanation to the new phenomenon he had discovered, but he treated it as
a new anomalous phenome-
< previous page
page_15
next page >
page_16
< previous page
page_16
next page >

Page 16
non which requires an explanation. He initiated the process of discovery which eventually led to the understanding
of the phenomenon. Lenard, on the other hand, misinterpreted the phenomenon. Hence, he was not considered to
be the discoverer.
Ontological Aspects of Discovery
Besides having strong epistemological implications, the notion of discovery has important ontological aspects.
Indeed, epistemology cannot be divorced from ontology. Some dictionary definitions of discovery may shed light
on the ontological status attributed to the object or product of discovery. We will find in the dictionary under the
item "discover" formulations such as: "to disclose a secret," "to expose (bring to light) something hidden" or "to
uncover." Implicit in these connotations is that something which is hidden from us is discovered; its existence is
independent of the process of discovery. This view is reflected in the expression: ''science reveals the secrets of
nature." This issue will be discussed in section 1.4.
1.2 The Products of Scientific Discovery
1.2.1 What Do Scientists Discover When They Look At the World?
When ordinary mortals look at the world, they discover entities which exist in the world such at ordinary objects,
properties, events, phenomena, causes and regularities. When the scientist investigates the world, the most
important kinds of items he discovers are not entities in the world, but new concepts, ideas and scientific theories,
which belong to the realm of his cognitive representation of the world. Epistemologically, the discovery of fruitful
concepts and theories is more important than the discovery of entities existing in the material world, since theories
enable science to make further discoveries of natural objects, events and phenomena. Thus, on the one hand,
physicists discover new kinds of particles, such as, the electron or the neutrino, and on the other hand, concepts
and ideas, such as the concept of spin or the idea of the field, or theories, such as quantum mechanics, which guide
the physicists in discovering new particles and their properties.
Scientific theory is one of the distinguishing characteristics of modern science. In understanding its nature, we will
increase our understanding of modern science. Therefore, my main concern in this section will be to elaborate on
the nature of scientific theories as objects or products of scientific discovery and in particular, to distinguish them
from laws of nature. Accordingly, as we will find out in the following chapters, the process of discovering or
generating a theory is substantially different from other kinds of processes of discovery in science, in
particularfrom discovering empirical generalizations and regularities. In fact, the notion of discovery itself, which
was inher-
< previous page
page_16
next page >
page_17
< previous page
page_17
next page >

Page 17
ited by science from ordinary experience, may not be suitable for describing the way scientists arrive at a theory or
how they generate it.
Before we turn to theories, let us deal with less structured products of discovery: objects and events.
1.2.2 Objects and Events Contaminated by the Scientist's Intervention
The distinction between observational and theoretical terms or statements has a long history in twentieth-century
philosophy of science. It is now widely agreed that all descriptive statements are "theory-laden" so that there are no
purely observational statements. However, when we turn to discoveries, we may employ the following
methodological definition: every object of discovery which can be discovered by observation, i.e. by the senses or
by using observational instruments and methods, will be qualified as "observational" discovery. It is a matter of
methodological decision or convention to determine what are the observational instruments.
In making an observational discovery, we have to rely on our system of categories or conceptual system through
which we grasp the phenomena we observe. In the process of observation by the senses, i.e. in so-called direct
observation, we process the raw data of observation using the conceptual and cognitive apparatus with which we
are equipped. Our cognitive apparatus guides us in dividing our visual field into enduring objects and natural kinds.
We also rely on some tacit assumptions and on inferences we make spontaneously and unawares. For example,
when we observe an object, we assume that what we observe is more or less what is there; that the light coming
from the object is not radically distorted or that no one painted it or sculptured it in order to deceive us. All these
background presuppositions and beliefs would not make the observed object conjectural if we make the
methodological decision to treat our cognitive apparatus as reliable and our innate or spontaneous presuppositions
unquestionable. We would make this decision in order to retain the commonsensical notion of observation which in
general does not deceive us in everyday experience. In fact, in ordinary experience we do not make any
methodological decision such as this. Normally, we are not aware of our cognitive apparatus. Or, rather, we are not
focally aware of it, to use Polanyi's expression (1958, 55-6). We may express it by saying that in the process of
observation our cognitive apparatus is "invisible" or "transparent." 3
In "indirect" observation, which involves the usage of instruments, we rely not only on our innate (genetically and
culturally determined) sensory and cognitive machinery; we also rely on established theories which govern the
operation of the observational instruments and methods. In science, we rely also on established theories which
supply us with the categories and concepts by which we describe the phenomena exposed by using the
observational instruments. These theories are (provisionally) considered unquestion-
< previous page
page_17
next page >
page_18
< previous page
page_18
next page >

Page 18
able. For example, when the particle physicist observes the products of collision experiments, he describes the
phenomena using concepts, such as particle charge or spin, electrons and protons, which he treats as belonging to
his observational vocabulary. Again, we may say that after the scientist acquired the skill of using his observational
or experimental apparatus, he treats it as transparent. Or when a scientist relies on a set of theoretical
presuppositions which he regards as unquestionable, we may say that these presuppostions are transparent.
Thus, the way we describe our object of observational discovery, depends on our cognitive apparatus, conceptual
system and background knowledge. All this machinery determines what we observe. However, it does not uniquely
determine what are the objects of discovery exposed by the act of observation. Indeed, we may observe a lot of
things, none of which will be considered to be a discovery. And different observers may be making different
discoveries by watching the same thing. Let us take the simplest kind of observational discovery. When we
discover a black swan, the object of discovery seems to be an individual or an object in the world. Without
entering now into questions related to the realism-antirealism debate, let us assume that there is an objective state
of affairs which we encounter in our discovery. However, what we discover is not this objective state of affairs;
our object of discovery is not identical with the state of affairs we encounter. It rather depends on our present
beliefs, expectations and interests. If, for example, we have been believing that there are no black swans, we would
be interested in the very existence of black swans. The product of discovery may then be the statement: "there
exists a black swan." If we are interested in the fact that the black swan was discovered in a certain place and/or at
a given time, we may be interested in the event observed, which is referred to by the statement: "there is a black
swan at time t in place x." If so, then what is the object of discovery? Is it the particular swan observed swimming
in an Australian lake? Is it the fact that it was black? Is it the fact that it was swimming? Is it the fact that it was
swimming in a certain direction, at a certain speed, at a particular hour in the day, in particular weather conditions,
at a particular distance from the shore, etc? Thus, given the same state of affairs, many discoveries may be madein
fact, an infinite number of different discoveries. The object of discovery depends on both the state of affairs in the
world and on the discovereron the discoverer's expectations, point of view and interests.
It seems, therefore, that even if there is an objective state of affairs which is partially described by our conceptual
system, the object of discovery is not a totally objective matter. Rather it is a certain aspect of reality, a certain
facet of the state of affairs encountered. Thus, the object of discovery has both epistemological and ontological
aspects; what we discover depends in part on what is there and in part, on our interests and on what we know or
believe. Thus, when someone who does not have any prior beliefs about swans, or believes
< previous page
page_18
next page >
page_19
< previous page
page_19
next page >

Page 19
that there are black swans, observes a black swan, the observation of a black swan does not constitute a discoveryfor-him. On the other hand, when someone who believes that all swans are white observes a black swan, the black
swan constitutes a discovery-for-him which results in a refutation of one of his general beliefs. A discovery has
the effect of adding something to the discoverer's system of beliefs or modifying it. Hence, even the simplest object
of observational discovery is not something which exists out there in the external world independently of our state
of knowledge, awaiting our discovery. When X discovers D, D is not determined only by the state of affairs in the
world, but also on the cognitive state of X.
Yet, not everything which is observed and which is unexpected or interesting constitutes a scientific discovery;
confirmation is part of the discovery process. If we refer to ordinary observational objects or events, there is an
entrenched procedure of confirmation: the act of confirmation is carried out by repeating the observation. It seems,
therefore, that a non-repeatable event cannot be an object of observational discovery. Indeed, "discoveries" such as
the purported discovery of gravitational waves in the early seventies, are discarded by the scientific community
since no one could repeat them (Collins 1975). Yet, a singular or unique event which is not repeatable can
nevertheless be confirmed if it was independently observed by several qualified observers or if similar events were
observed in the past. For example, astronomical events, such as a supernova are legitimate objects of observational
discovery in science.
Thus, the following three factors contribute to the product of observational discovery: (1) A conceptual system or
system of categories constitutes a precondition for making an observation. (2) A background knowledge and prior
point of view (an expectation, a belief, a theory) determine what observational finding would be regarded as a
discovery. (3) The act of confirmation determines whether the object is indeed a discovery. All three factors
"contaminate" the object of discovery with our cognitive intervention. This is obvious for the first two factors. As
we will see in the following chapters, the act of confirmation, too, is not an objective matter; it is not devoid of
psychological and social components. Thus, even the most "naive" objects of scientific discovery, observational
discoveries, are not objective entities in the world.
In science, the effect of factor (2), which determines which observational findings will be regarded as discoveries,
is exhibited in the following way. The objects of observational discovery are objects and events, such as a new
particle, species, chemical element and compound, star, galaxy, particle decay, chemical reaction or supernova. An
object or an event are considered to be a discovery if it is unique of its kind or if only few of its kind have been
discovered. A supernova belongs to the last category. What about a discovery of a new particle? The discoveries of
the neutron, the p-meson and the neutrino were regarded by elementary particle physicists as genuine discoveries
< previous page
page_19
next page >
page_20
< previous page
page_20
next page >

Page 20
since the number of so-called elementary particles was relatively small at the time of these discoveries. However,
since the 1960s, when the number of particles increased, an observation of a new particle or resonance was not
considered a discovery in its own sake. An observation of a new particle or state of matter was treated as a
discovery if it confirmed or refuted a regularity or a theory. Thus, the discovery of the omega-minus particle in
1964 was an important discovery since it strongly confirmed the SU(3) symmetry of strongly interacting particles
(hadrons). Thus, in a science which is in its theoretical phase, an observational discovery may be theory-laden in
one more respect than an observational finding; an observational finding is a genuine discovery only if it leads to
the confirmation of a not-yet established theory or law or to the refutation of an established theory or law.
In astronomy, the standard is different. A new star is considered to be a discovery, although many stars have been
observed. Perhaps the reason for this is the same reason for considering a new biological species as a discovery.
Every new species is unique; there is no law of nature or regularity which will enable us to predict the existence of
a new species. The same thing can be said of stars; there is not a theory or a law of nature which can predict the
existence of stars, as symmetry theories in particle physics or as the Periodic Table of elements predicts the
existence of particles or elements, respectively. Discoveries of unique objects related to the history of humankind
or life in general or to the history of our planet, have a special interest for their own sake. This is understandable
since these sciences are not theory-dominated. Yet, even in theory-dominated sciences, there may be important
discoveries of specific events or objects. Perhaps the most important scientific discovery ever, would be a
discovery of a singular event: how our physical universe was formed (e.g. the "big-bang" event) or how life began.
Moreover, the event or process by which life began would become even more thrilling had we shown that such an
event is unique or that its probability is vanishingly small; no sane scientist would reject a research project which
has high chances for leading to the discovery of this event even if it would be known in advance that it will not
contribute a bit to our general knowledge. However, these kinds of singular events are non-observable and their
discovery is heavily loaded with theoretical inference.
1.2.3 The Plasticity of Theories
We have seen that observational discoveries are contaminated by our cognitive intervention. However, when we
come to scientific theories, the contribution of the discoverer's cognition to the object or product of discovery
becomes much more significant. Theories are created by us, although not entirely by our free imagination; the
process of creation is constrained by the empirical data and by the conceptual resources available to the scientist.
Here
< previous page
page_20
next page >
page_21
< previous page
page_21
next page >

Page 21
the scientist and the artist are in the same boat. The sculptor, as well as the physicist, is restricted by his medium of
creation.
The term scientific theory is employed in a variety of ways in the literature. Obviously, there is no a priori "right"
definition of the term. I will propose a way to characterise this notion by distinguishing it from the notion of
natural law, having in mind contemporary usages by scientists and by philosophers of science.
Laws of nature are specific kinds of regularities. The discovery of a regularity or an empirical generalization does
not involve an introduction of new concepts which do not appear in the observational vocabulary. We make
generalizations when we identify natural kinds which exhibit regular behavior or characteristics. The identification
of natural kinds is a matter of both our experience and our natural endowment. Our cognitive apparatus guides us
in the general pattern of concept formation and our experience in the specific field of investigation guides us in
identifying the natural kinds in that field. Thus, regularities and empirical generalizations are contaminated by our
cognitive intervention in the same manner as observational discoveries of singular events and objects are. And so,
the above mentioned three intervening factors involved in observational discovery are also operative in this kind of
discovery. The only difference is that the act of confirmation is different. It relies on inductive "inference."
Induction cannot be justified by logical or "objective" standards, so we may treat it as another natural endowment
which is part of our cognitive apparatus. Thus, the object or product of discovery is further contaminated by our
belief in induction and in the uniformity of nature. If we still believe that we discover here an entity in the world,
we may say that the object or product of our discovery is an invariable relation between properties in the world, for
example, between the property of ravenhood and the property of blackhood. If we are realists we may regard
relations as entities existing in the world, although in a more abstract sense than objects and events are.
Now, there is a difference between an empirical generalization, such as "all ravens are black," which seems to be
"accidental," and a law of nature, such as the law of universal gravitation, which seems to be "universal." Both
empirical generalizations and laws of nature embody invariable relations in the world. So what is the difference
between them? Philosophers have tried to answer this longstanding question by looking at the logical form of lawstatements, employing the machinery of "possible worlds." However, they have not yet found a satisfactory answer
in this direction. This is one of the examples where philosophers unsuccessfully try to find answers to questions
concerning science in a logicist direction. Perhaps the answer lies in the realm of the dynamics of scientific
knowledge; perhaps it lies in the manner by which science progresses, rather than in the logical form of the lawstatements.
When an empirical generalization is embedded in, or derived from, a strongly confirmed theory which holds over a
wide range of phenomena, then
< previous page
page_21
next page >
page_22
< previous page
page_22
next page >

Page 22
we might treat it as a law of nature. Indeed, in such a case, we can explain why the generalization is not accidental;
if the theory's postulates are true, then the generalization must be true. A unifying theory is confirmed by its wide
range of successful predictions and explanations. So the theory gives the generalization a stronger confirmation;
through the theory the truth of the empirical generalization is linked with the truth of other kinds of established
phenomena. Sometimes an empirical generalization is so entrenched that it is treated as a law of nature even
without being embedded in a theory. But then it is assumed that, eventually, a unifying theory will be found in
which the generalization will be embedded.
A physical law, such as the law of universal gravitation, is an invariable relation between physical magnitudes,
such as masses, distances and forces, which can be thought of as properties. There are other kinds of laws, which
cannot be described as relations. For example, a conservation law, which states the constancy of a given
magnitude, such as energy, or the second law of thermodynamics, amount to a restriction on possible processes, or
to a limitation on the value of a given magnitude or property.
When we say, as it is customary to say, that a law "states" something, we do not mean that the law is a statement
made by us. It plays an analogous role to a judicial law in preventing certain states of affairs to occur, or in
specifying how properties or objects "should" behave. We represent the law by statements expressed, for example,
by mathematical formulae. However, these formulae or statements refer to the law which is part of reality. A law
of nature is not a proposition or a statement. The formula F=kq1q2/r2 is not Coulomb's law, as is frequently stated
(presumably as a shorthand) in textbooks. Rather, it is a law-statement which refers to Coulomb's law; the law is a
relation between the physical magnitudes (properties) of the electric force between two interacting particles, their
charges and the distance between them. In summary, a law of nature may be an invariable relation or a restriction
on possible states of affairs or processes.
It is not within the scope of this book to deal with the ontological status of laws of naturewhether and in what
sense they exist. However, in order to understand in what sense they differ from theories as objects of discovery, I
will adopt the following mildly realistic view. I start from the assumption that objects, properties and events exist
in the world. According to this view, laws of nature are invariable relations between properties or restriction on the
existence of certain events or objects. That is, they do not exist in their own right in the sense objects and events
exist in space and time (or in spacetime), but they correlate properties or dictate what objects or events cannot
exist.
Thus, laws of nature are somehow related to entities which exist in the world, in the same fashion as properties are.
We may view them as properties of higher order, or as restrictions on what can exist or happen in the world. If we
have only two possibilities where to accommodate them, in our cognition
< previous page
page_22
next page >
page_23
< previous page
page_23
next page >

Page 23
or in the external world, then as realists we will choose the second possibility. This is still an intuitive view which
is not fully analyzed. However, for the sake of distinguishing between laws and theories, I will not need more than
this.
Theories are the most important targets of scientific discovery, at least in the physical sciences. What sort of
entities are they? What sort of entity is Newtonian mechanics, the Darwinian theory of evolution or Maxwell's
theory of electromagnetism? In order to answer this question, we have to find out what are the functions of
theories in modern science. Among the functions attributed to a theory we can find the explanation and prediction
of natural phenomena and laws of nature and the unification and systematization of our knowledge. Newtonian
theory explains why planets are encircling the sun according to (a modified version of) Kepler's laws. It also
provides a unification of diverse phenomena such as celestial and earthly motion. It seems, therefore, that theories
fulfill epistemic functions. They stay closer to our cognition than to the objects and events which exist in the
world. We would not say that Newton's or Darwin's theories exist in the world; rather they describe and explain
what exists. Thus, if we do not wish to employ Popper's three-worlds machinery for accomodating theories and we
have only the two above mentioned possibilities to categorize them, we would say that they are elements of
knowledge rather than elements of reality; they are cognitive objects which represent reality.
When we discover a theory, we discover something which is related to external reality; for example, something
which represents reality in our minds or something which helps us in comprehending reality. So what kind of
entity is it? Is it just a description, a statement, or an explanatory instrument? Traditionally, philosophers of
science viewed theory as a set of statements, orsimplyas a statement (which is the conjunction of all the theory's
statements). This might distinguish between a law of nature and a theory, since a law of nature is not a statement.
However, if we believe that a law of nature exists in the world, we would say that a law-statement refers to a law.
We might, therefore, ask what the theory-statement refers to?
Some theories may be viewed as a system of interrelated law-statements. For example, Maxwell's theory of
electromagnetism consists of Maxwell's equations which yield formulas and equations expressing laws of nature,
the laws of electricity, magnetism and electromagnetism. Yet, the system of laws consists of something more than
a mere collection of laws. Within the theory the laws are interrelated via Maxwell's equations. The theory thus
serves as a unifying entity. This might be expressed by saying that the theory refers to relations of a second order:
relations between laws which are themselves relations between physical magnitudes. If we admit relations to our
ontology, there is no reason for rejecting relations of any specified order. Thus, in discovering a theory, we
discover a set of relations between observable properties and measurable magnitudes.
< previous page
page_23
next page >
page_24
< previous page
page_24
next page >

Page 24
Yet, it is doubtful that even Maxwell's theory just refers to nothing more than to a set of interconnected laws of
nature. If we are not instrumentalists or anti-realists we would say that there are ontological claims made by the
theory about the existence of some kind of entities, such as electromagnetic waves. And if one would argue that all
these claims about waves can be reduced to relations, without committing oneself to the existence of anything else,
we might take as another example the kinetic theory of gases, which deals with more tangible objects. This theory
does not exclusively consists of a system of equations. It has clear ontological claims: the existence of molecules in
motion etc. The theory's equations are derived from a model of molecules in motion. If we do not yield to the
already bankrupt positivistic views of reducing everything in theory to the measurable and the observable,
dismissing any ontological commitment, then we would say that the theory states the existence of particles and
their properties and behavior.
This difference between law and theory is substantial. A law-statement does not state explicitly anything about the
existence of any entities in the world. It just describes relations between entities which scientists have accepted as
existing before the law was discovered. A theory, on the other hand, in many cases states the existence of new
entities, such as electromagnetic waves, subatomic particles and forces, i.e. new with respect to our state of
knowledge before the theory was discovered. More generally, the theory is constructed from new ideas, concepts,
models and analogies. These may be qualified as proto-theoretical entities, which are also objects of discovery. In
other words, the novelty of a theory is in the new proto-theoretical entities from which it is constructed, including
the new entities it claims to exist. Thus, here the distinction between a law-statement and a theory-statement is not
related to their form or their content but to their genesis.
In fact, the introduction of new entities, concepts, ideas, etc. is a necessary condition for the theory to be
explanatory. Newtonian mechanics, for example, introduced the concept of mass. Indeed, one of the common
principles of explanation is that explanation cannot be circular. Thus, a phenomenon is not explained by employing
the same predicates used for describing the phenomenon. For example, "the sky is blue" would not be explained by
"the sky consists of blue particles." This might apply to the explanation of laws of nature as well: Coulomb's law
would not be explained, on pain of circularity, by the "theory": "all charged particles obey Coulomb law." Or,
Boyle's law for an ideal gas in a container cannot be explained by the statement that every volume element of the
gas in the container satisfies Boyle's law. Thus, the explanans should include something newe.g., new kinds of
objects, new properties, relations or structureswhich do not appear in the explanandum. This principle is not
mandatory in scientific explanation, since we encounter in science "bootstraps'' kinds of explanation. However, if
we restrict ourselves to atomistic explanations or to explanations by reduction, which constitute
< previous page
page_24
next page >
page_25
< previous page
page_25
next page >

Page 25
the majority of scientific explanations and perhaps the ideal model of scientific explanation, the principle must
hold.
When I say that a theory introduces "new" concepts or entities, I mean new relative to our state of knowledge
before the discovery of the theory was made. Hence, if we view the theory as a final product, independently of the
epistemic circumstances in which it was discovered, no sense can be given to the "newness" of the entities referred
to by the theory or of the concepts employed by it. The theory-statement in this case just employs a set of concepts,
without making explicitly any ontological claim. In this respect a theory does not differ from the statement
referring to a law of nature.
Thus, a substantial difference between the two would arise only if we view a theory from a historical or epistemic
point of view. Only in a historical context can the theory be viewed as making explicit ontological claims. But
then, a theory will be part of a historical process, transcending a mere statement. In saying that one of the
distinctive features of a theory is the new entities it predicts, relative to the previous state of knowledge, we
attribute to the theory an epistemic role of advancing knowledge. Therefore, we cannot view the theory in isolation
from the historical process in which it emerged. Furthermore, in section 1.3, I will propose viewing the theory itself
as a dynamic entity which brings about further discoveries after it was brought to life. According to this view, in
discovering a theory, it is not just a static description, or a statement, which is discovered. Rather, it is a basic idea,
model or picture which guides a research program.
We would expect, therefore, that the process of discovering a theory will differ from the process of discovering a
regularity, for example. In discovering a theory, we discover a guiding tool for advancing knowledge. This means
that the object of discovery in this case is neither something in the world nor a mere statement. A model or a
picture cannot be fully described by a statement. It is not an entity which refers to something in the world. Rather
it is a cognitive or epistemic object which helps us in representing the world or in grasping it in a dynamic
fashion.
This view of theories can by no means be categorized as instrumentalism. If we were instrumentalists, viewing
theories as instruments for organizing and predicting observational data, then the object of discovery would be a
set of directions for deriving predictions, a superstructure for organizing the data, etc. The main difference between
the view expounded above and the instrumentalist view is that according to the above view the theory as a
cognitive tool is mirroring reality, whereas the instrumentalist view does not make such a claim.
From this point of view, the transition from the pretheoretical stage, when a science deals only with empirical
generalizations, regularities and laws of nature, to the theoretical stage, is a radical transition. It does not involve a
change in degree, such as the change embodied in the transition from a less
< previous page
page_25
next page >
page_26
< previous page
page_26
next page >

Page 26
general theory to a more general and abstract theory, as exemplified by the transition from the kinetic theory of
ideal gases to the molecular theory of matter. It rather involves a category-change. Discovering a regularity or a
law of nature and discovering a theory are entirely different things. In one case, we discover something in the
world, in the other casea cognitive object. One difference is that although discovering a regularity or a law of
nature may require cognitive intervention and human ingenuity, a regularity or a law of nature is something we
find in the world rather than create or invent. Proto-theoretical entities and a full-fledged theory, on the other hand,
are our own creations or inventions. They are new tools for acquiring knowledge or new information channels
through which we interact with the world. This categorial difference will be reflected in the process of discovery.
At first sight, the difference seems to be so big that one may wonder why we subsume both kinds of processes
under the same title of "scientific discovery."
Until now I have not distinguished between different stages in the development of a theory. It seems that even if a
growing theory is not a statement, a mature theory comes close to being a statement. When a theory matures and
becomes well established it is relegated to the unquestionable background knowledge of science. In its mature
stage, the basic structure of the theory almost does not change over long periods of time. The theory is mainly
applied for explanation and prediction, and is employed as a premise in scientific reasoning. It would be tempting
to say that at this stage a theory is a statement, referring to a portion of reality. Namely, we may treat the mature
version of the theory as what remains of the dynamic tool. Indeed, when people say that a theory is a (set of)
statement(s), they have in mind a mature version of the theory. Thus, we may view the established version of the
theory as an object of discovery, which may constitute a substantial advance over the initial version. As we will
see when we discuss kinds of discovery processes, the process of discovering a theory may indeed be a prolonged
dynamic process resulting in a mature version of a theory. Thus, in addition to the discovery of the initial version
of a theory, there is another kind of discovery, which is a dynamic process in which the theory is adjusted to the
data and further elaborated. For example, Bohr's discovery of the initial or "naive" version of the structure of the
hydrogen atom was the major discovery. However, the mature version, including elliptical orbits, relativistic effects
and spin, constituted a discovery in its own right.
Yet, even a mature theory need not, and perhaps should not, be seen as a static entity referring to some definite
entities in the world. Classical mechanics, for example, the paradigm of scientific theory, gives a general
description of the worldgeneral equations which can be applied for describing different kinds of systems in the
world. The application of the theory to a new kind of physical system is in many cases a creative task in its own
right which does not involve mere computations. Here we should distinguish two kinds of applica-
< previous page
page_26
next page >
page_27
< previous page
page_27
next page >

Page 27
tion: the application of the theory to a general system, such as the physical pendulum or the planetary system,
which results in a "model" for this general system, and the application of the theory to a concrete, specific,
pendulum or planetary system. The structure referred to by this kind of "model" can be viewed as a generalized
law of nature applying to a general kind of system.
The application of a theory to new systems is an essential part of the development of the theory proper. Indeed, the
theory can be viewed as a theoretical core plus the range of general kinds of systems to which the theory applies.
In the early stages of the theory development, the core is changing in order to adjust it to new kinds of systems. In
the mature stage, the core may change only marginally, but the range of applications expands. Hence, we can say
two things with respect to a mature theory. First, even in its mature stage, a theory is developing. Second, the
mature theory is not a statement about reality. Rather it specifies general ways of treating different kinds of
systems. Its domain of application expands and this is one of the major ways through which a theory grows. Only a
"model" (in the above sense), describing a certain kind of system, functions as a statement referring to a general
structure or to a generalized relation existing in the world.
In summary, the statement-view is not appropriate for describing the two major aspects of the growth of a theory
as a dynamic entity. It is not appropriate for describing a developing theory which is changing in its basic structure.
And it does not do justice to a mature theory extending its domain of application. In both cases the theory serves as
a guide for the growth of knowledge about the world rather than as a statement referring to something in the
world.
Due to the dynamical nature of scientific theories, the process of discovering them requires a high degree of
creativity. Furthermore, the plasticity of theories makes them liable to creative changes and in particular to
unintentional or serendipitous changes. As I will argue in Part II, scientific creativity is equated with
unintentionality and serendipity. And this is what makes scientific progress an evolutionary phenomenon.
1.2.4 Explanations, Problems and Solutions
There are three additional kinds of products of discovery which deserve our attention, since they are related to the
theoretical structure of science and to its evolutionary nature.
Explanations
There are two kinds of scientific explanation: an explanation of the properties and structure of general systems,
phenomena, regularities and laws of nature and the explanation of specific events and properties. The first kind of
explanation is derived from the theory in conjunction with some assumptions regarding the structure or the
dynamics of the explanandum. For example, a modified version of Kepler's laws was derived from Newton's
theory
< previous page
page_27
next page >
page_28
< previous page
page_28
next page >

Page 28
of universal gravitation in conjunction with Kepler's model for the solar system. An example of the explanation of
a phenomenon, i.e of a general kind of process or event, is the explanation of the phenomenon of tides which is
provided by gravitational theory in conjunction with some assumptions about the initial condition of the system
consisting of the earth, the sea and the moon.
In the second kind of explanation, the explanation is given by a law or a theory in conjunction with initial or
boundary conditions, possibly in con-junction with auxiliary hypotheses or assumptions. This is the "deductivenomological" kind of explanation treated by Carl Hempel (1965). However, the first kind of explanation is more
common and more important in a theoretically advanced science. The advance from the phenomenological stage of
empirical generalizations to a theory brings about the explanation of the phenomenological regularities by the
theory.
To find an explanation is not the same as finding a cause. If we adopt the hypothetical realist view, a cause is an
entity existing in the world. The cause of the tides is the moon moving in a particular trajectory. The cause of the
pain I feel in my head is the object which hit it. Both causes are events, which exist in the world. However, the
explanation of the phenomenon of the tides, or of a particular occurrence of a tide, is a human product. If we treat
scientific explanation along the lines proposed by Hempel, for example, we have to find the theoretical premises
and the statements describing the initial conditions, from which we can derive the explanandum. It may happen
that we know all the premises from which such an explanation can be constructed, without being able to construct
the explanation. Discovering an explanation involves finding the right ingredients, and the right combinations
thereof, from our repertoire of theories, laws and facts, from which we can derive the explanandum. We encounter
the same situation in puzzle-solving. Here the discovery is not made by looking at the world, but by looking at our
representation of the world. In Chapter 6, I will discuss a theory which describes the process of discovery as
consisting of quasi-random formation of combinations of "mental elements." The product of the process is a stable
"configuration" which is finally selected. In the above mentioned process of discovering an explanation, the
product of the process is, indeed, a "configuration'' of mental elements, i.e. of ideas, theories, laws, facts, etc.,
which solves the problem.
Problems and Solutions
A very general category of discoveries is a solution of a problem. When we resolve an inconsistency in a theory or
between theories, or a disagreement between a theory and data or when we find an explanation for a puzzling
event or when we explain a set of empirical generalizations by a general theory, we solve a problem. In general,
when our scientific standards require a specified manner of explanation or understanding, we face a problem
whenever we have not yet achieved the goal set up by these standards. Problems may arise in theoretical and
pragmatic contexts.
< previous page
page_28
next page >
page_29
< previous page
page_29
next page >

Page 29
However, an important kind of discovery is the discovery of a new problem. In many cases, the discovery of a
solution of a given problem leads to the emergence of new problems. A discovery of a new problem may lead to a
significant progress of science, sometimes more so than a discovery of a solution of a known problem; such a new
problem may open up a whole new field of investigation. For example, the problem called "the ultraviolet
catastrophe" in blackbody radiation was crucial for the discovery of quantum mechanics. Problem solving can be
construed as an evolutionary phenomenon. We will turn to this point when we discuss the subject of serendipity
and the evolution of science.
1.3 The Kinds of Discovery Processes
Let us divide processes of discovery into two main categories which may be termed "exposure" and "generation."
Paradigm cases of discovery by exposure from everyday experience is when we discover the hidden content of a
closed box by opening it, when we discover something in the darkness by throwing light upon it or when we infer
the hidden cause of an event (discovering who is the murderer in a detective story). An example of generational
discovery in everyday experience is when we discover the maximum tension which a rope can stand by hanging
increasingly heavier weights on it. Or, when we discover how the color of the rope changes when we put it in a
certain solution. Another example is when we discover the effects of growing some plants in a greenhouse. In all
these experiments, the new effects discovered by us are in a sense created by us. However, the salient cases of
generational discovery occur in science (actually, the above examples are on the border-line between ordinary
experience and science), whereas in everyday experience exposure cases are more typical. In fact, generational
discovery is one of the characteristics of modern science. In general, discovery by exposure does not create
anything new in the world or in our representation of the world, whereas generational discovery, in a sense,
perturbs the world, interferes with the natural course of events, creating new, or new kinds of, observational or
theoretical entities. The sense in which generational discovery perturbs the world will be discussed in section 1.4.
The distinction between exposure and generation will be essential to the evolutionary view of discovery: it is
generational discovery which exhibits the characteristics of natural selection.
1.3.1 Exposure
Discovery by Observation
Discovering a new object, event or a new observable property by looking or sensing, or by using observational
tools and methods. Examples: The discovery of Jupiter's satellites by Galileo, using his tele-
< previous page
page_29
next page >
page_30
< previous page
page_30
next page >

Page 30
scope, or the discovery of the structure of a macromolecule, using an electron-microscope. Observation is by no

means a passive act, since it involves looking at chosen directions, employing instruments and making inferences.
However, it is not a generational process, since the act of observation does not create, or significantly effect, the
object of discovery. We have to qualify this statement with regard to the observation of macroscopic systems. In
quantum physics, observation does perturb the observed system.
Discovery by Searching
Scanning a given portion of space looking for some prescribed event or object. Examples: Looking for oil
resources by searching a given area. Scanning bubble-chamber photographs in order to discover certain events.
Searching for a solution to a problem in a space of possible alternatives. This procedure is practiced in problem
solving and in heuristic search in artificial intelligence. Search may be conducted where there is a finite number of
possible hypotheses for explaining a given phenomenon. Here search means eliminating alternative hypotheses, the
product of discovery being the remaining alternative. If there is a finite number of alternative hypotheses for the
general cause of a phenomenon, this method is reduced to "inductive" inference, i.e. eliminative "induction"; if we
come to the conclusion that there are n possible causes for a given general phenomenon, we may eliminate n-1
possibilities by conducting appropriate observations or experiments. The result will be the discovery of the cause of
the phenomenon. In fact, eliminative "induction" of this kind, were the number of possible hypotheses is finite, is a
deductive inference.
Calculation and Computation
Mathematical calculation may lead to important discoveries in everyday life as well as in science. For example, by
measuring the length of a metal rod at two different temperatures (L and L0, at t and t0), we can determine its
coefficient of thermal expansion (a) by using the formula L=L0[1 + a(t-t0)]. If we have a table of thermal
expansion coeficients for different metals, we can use this result to discover the chemical identity of the metal.
This is an example of a procedure in which one starts from some premises which include certain mathematical
formulas and the numerical results of some measurements and arrives at the result by substituting the numerical
results in the formulas and carrying out the required mathematical operations. The mathematical formulas may be
derived from a full-fledged theory or may be just rules of thumb. This procedure is a specific case of deductive
inference when one starts with some premises and arrives at a logical conclusion by following the rules of
inference.
Inference
The question is why do we categorize calculation, and deductive inference in general, as a process of discovery by
exposure rather than by generation? The reason is that the process of deductive inference leads to information
which is already logically contained in, or necessarily implied by, the
< previous page
page_30
next page >
page_31
< previous page
page_31
next page >

Page 31
premises. Thus, deductive inference cannot create any new information or new theoretical entities referring to new
kinds of objects, events or phenomena in the world; it is just a process which exposes to us or uncovers what is
"hidden" in, or implied by, the set of premises.
Deductive inference is the paradigmatic case of discovering by exposure what is hidden in our representation of the
external world. Similarly, discovering the content of a closed box by removing the cover is the paradigmatic case
of discovery by exposure in the external world. In the first case we discover information contained in a set of
statements which is beyond the grasp of our intellect without carrying out a sequence of logical transformations,
whereas in the second case, we discover objects in the world which are beyond the reach of our senses without
pulling up the cover. In both instances, our knowledge grows as a result of the discovery. Indeed, our knowledge is
not deductively closed. If our knowledge is represented by a set of statements K, not all logical consequences of K
are known to us. Otherwise, we would have to accept the claim that a child knows all the theorems of Euclidean
geometry as soon as he learned its axioms.
By deductive inference we can arrive in mathematics not only at new numerical results but also at new theorems.
In physics, we might arrive at new laws of nature expressed in mathematical formulas. For example, in Newtonian
mechanics, the laws of planetary motion were discovered by deriving them from the inverse square law of
universal gravitation. Non-mathematical deductive inference does not carry us as far as mathematical derivations
do but it is still valuable as a means of discovery.
In computer science and artificial intelligence, one may arrive at far-reaching discoveries by logical inference. By
giving the instruction FIND to the word processor I am working on, I can discover a "lost" word throughout my
text. Here we have a case where a physical object "hidden" in the text can be found by logical steps (made by a
machine) which guides a physical search.
However, inference is not restricted to deductive inference. By inductive inference we may discover causes of
phenomena and laws of nature. The most common inductive inference is arriving at a new generalization by
enumeration, i.e. by generalizing from particular cases. What is common to these kinds of inference is that they
lead to new information which is not logically included in the premises. The premises of an inductive inference
which leads to a generalization include information only about a finite number of past instances, whereas the
conclusion refers to an infinite number of past and future instances. Yet, this sort of discovery process can be still
viewed as a process of exposure. First, it does not generate new concepts or theoretical entities referring to new
kinds of objects or phenomena; the conclusion (e.g. "all P's are Q") in an inductive inference does not contain new
predicates which do not appear in the premises (which refer, for example, to single instances of P's which are Q).
Second, the new information generated by
< previous page
page_31
next page >
page_32
< previous page
page_32
next page >

Page 32
inductive inference can still be seen as "hidden" in the premises. Indeed, we can view inductive generalization, for
example, as an inference which relies on a principle of induction which is tacitly assumed, e.g. the principle which
says that if a large enough number of instances of observed P's were found to have the attribute Q, and no instance
of P was found not to be Q, than all P's are Q. If this principle is taken as a tacit premise, than the inductive
generalization follows deductively. Hence, when the principle of induction is transparent, inductive inference can
be qualified as discovery by exposure rather than by generation. Here, too, the choice of the "right" predicates on
which we make our inductive generalization is a crucial stage in the inference. If we use natural language, the
problem may not arise since the system of possible predicates by which we refer to the objects in the world is
given in the language. In science, however, the system may change; this is in fact part of scientific creativity. This
leads us to the issue of natural kinds. Choosing a different system of natural kinds involves a creative act and the
process would be qualified then as generational.
There is another type of inference, where the rules of inference are "material" rules. Here the rules themselves
carry information about the world, and the conclusion of the inference might contain predicates not appearing in
the premises. However, these predicates are not new since they appear in the inference rules. Hence, if rules of
scientific discovery can be represented as material rules of inference, we may treat the discovery process as
discovery by exposure. The issue of material rules of inference will be discussed in Chapter 2.
Dynamic Theory-Construction
As we have noted, in many cases a theory is not constructed in one act. Possibly after an initial discovery of an
"ideal" version, or a first approximation, of the theory, the theory is gradually built through a dynamic process in
which it is adjusted and readjusted to experimental results and to established theories. This process yields
throughout its history a sequence of different theory-versions. The sequence may be called (following Lakatos
1970) "research program." For example, the discovery of the structure of the hydrogen atom, starting with the
Bohr-Rutherford model and ending up with quantum mechanical theories of atomic structure, can be seen as a
research program. In general, when we are guided by an analogy or a background model in constructing a theory,
the process of "spelling out" the model or the analogy is a process of exposure which may contain generational
elements.
1.3.2 Generation
In generational discovery the object of discoverywhether it is an object in the world or in our representation of the
worldis in a sense a product of the
< previous page
page_32
next page >
page_33
< previous page
page_33
next page >

Page 33
discovery process, hence we cannot say that it existed before its discovery in the same full-blown sense as do
objects discovered, for example, by direct observation or by searching.
Experimentation
By devising controlled experiments, which may involve sophisticated instruments and experimental methods, the
scientist may discover things which he would have never discovered in natural conditions. The objects or products
of such a discovery come into existence to a large extent as a result of the experimenter's action. This claim may
be illustrated by the case of watching the behavior of animals in captivity. The patterns of behavior we might
discover in these circumstances do not necessarily reflect the animal's behavior in its natural habitat; to a large
extent these patterns of behavior appear as a result of the artificial conditions we create. As we will see, certain
processes of observation in particle physics belong to the category of generational discovery; in these processes the
observed particle is sometimes produced in the process of observation. Thus, the border-line between observation
and generation is not sharp. As we will see in section 1.4, one of the ways of exposing the deep structure of natural
systems is by generating artificial products, such as new short-lived particles.
Theory-Construction
As we have observed, the construction of an explanatory theory involves the introduction of new concepts or new
theoretical entities which do not belong to the observational vocabulary or to the old theoretical vocabulary. Thus,
the processes of constructing theories, such as the kinetic theory of gases, Newtonian mechanics or the germ theory
of disease is generational.
The direction of scientific explanation or unification is the direction of the growth of knowledge in the strong
sense. In this sense the advance of science is not restricted to the accumulation of new empirical information, or
new information on a given theoretical level. Rather it is characterized by progressing to deeper levels of
explanation and reality. What is common to experimental and theoretical generation is that discovery by generation
is related to probing into deeper levels of reality.
Yet, even when we arrive at a theory by generation, the process of discovery must include a stage of exposure.
Only after testing some of its far-reaching predictions can we discover that the theory is successful. The theory
may be constructed first, and then it is tested and highly confirmed. Or, construction and confirmation may proceed
together. Only after the theory is confirmed does the scientist find out that the generated theory is a discovery. The
confirmation may come in a dramatic event, or it may be a prolonged process. The act or process of confirmation
may be viewed as an act of exposure; we derive a prediction from the theory and discover that it fits observational
data. Ideally, the derivation of the prediction is a process of deductive inference, by which we expose the prediction
which was hidden in the known
< previous page
page_33
next page >
page_34
< previous page
page_34
next page >

Page 34
premises, consisting of the theory and the relevant background knowledge, including the initial conditions.
The Link between Experimental and Theoretical Generation
The above two categories of generational discovery, experimental and theoretical, are closely linked. In order to
put a theory which is remote from the phenomenological level to a stringent test, the scientist has to devise
controlled experiments which create artificial environments. In this way, experimental generational discoveries are
made. On the other hand, in order to explain the results of highly sophisticated experiments the scientist has to
generate highly abstract theories. The result is that highly abstract conceptual systems and constructs which are
remote from our everyday conceptual system and mental representation of the world are generated in order to
comprehend highly artificial phenomena generated in the laboratory which are remote from everyday experience.
Thus, the two processes are fed by each other.
1.3.3 Poincar: The Poverty of Creation
Henri Poincar's observations on discovery focus on mathematical discovery, however they are of value for
understanding the process of discovery of ideas, theories and solutions of problems in general. Poincar maintains
that selection is the focal point of discovery: "Discovery is discernment, selection." Selection is much more
important than generation. Everyone can form new combinations with mathematical entities. However, the number
of these combinations is unlimited and most of them are useless. Thus, the creative part of the process, i.e. forming
the unlimited number of combinations, is valueless. The discoverer chooses the unexpected useful combinations or
relationships or hits upon them. In describing the way these combinations are created, Poincar turns to
psychological speculation. An unlimited number of combinations or hypotheses may be generated in the
subconsciousness. However, in the discoverer's consciousness mostly useful combinations appear. "Everything
happens as if the discoverer were a secondary examiner, who had only to interrogate candidates declared eligible
after passing a preliminary test" (quoted in Taton 1957,17). Thus, what distinguishes the discoverer from everyone
else is that he has an internal power of discrimination between the unlimited number of ideas and candidatesolutions. Most of the selection is done subconsciously. The sudden appearance of inspiration is due to
subconscious activity, mainly of selection of the most fruitful combinations, according to some feeling of
"mathematical beauty" and ''aesthetic sensibility." These feelings are part of the tacit knowledge shared only by the
scientists (in this casethe mathematicians). Perhaps it is a necessary condition for being a discoverer to be capable
of internalizing these shared tacit standards. In Chapter 6, I shall discuss a psychological theory which treats the
above process less speculatively.
< previous page
page_34
next page >
page_35
< previous page
page_35
next page >

Page 35
Two conclusions might be drawn from Poincar's view. First, the more possible ideas or hypotheses we have for
solving a problem or explaining some phenomena, the less we are close to our target. Note that the opposite
statement is not necessarily true: we might have a very few ideas, none of them is useful! Second, in contrast with
the common image of scientific discovery, the creative part of the process of discovery is valueless. However, we
may view the process of creation as generation and selection taken together; generation without selection is
fruitless. Third, the discoverer is somewhat less responsible for the act of discovery than what is commonly
thought. Indeed, the main process is an unintentional or natural (psychological or even biological) process which is
carried out without the intervention of a human conscious act. We have here a first hint towards a specific kind of
generational discovery process, discovery as an unintentional or involuntary phenomenon. We will turn to this kind
of process in Part II. Thus, Poincar's view diminishes the value of method of discovery. Unintentional or
involuntary acts cannot be governed by method.
1.3.4 Eureka Events and Unintentionality
A dramatic moment of discovery may occur in a sudden act of revelation or as a gestalt switch, without any
intentional effort on the part of the discoverer. The story of Archimedes' discovery of his law in the bath, the story
of Newton's discovery of universal gravitation triggered by the falling apple and Kekul's story about the discovery
of the ring-shaped structure of the Benzen molecule while daydreaming about a snake chasing its tail are perhaps
apocryphal. But we do not have to look too far for real cases of this sort. Most people have had the experience of
seeing something in a new light, of sudden understanding of an enigmatic phenomenon, discovering a relation
between previously unconnected phenomena or finding a solution to a problem after many unfruitful attempts.
Again, it is not uncommon to "see" suddenly that an idea or a theory which was employed for other purposes, or
which had been forgotten, solves a new problem or provides an explanation for new phenomena. This association
of an idea or a theory with a problem or with an explanandum may occur as an eureka event.
In order to count as an eureka event, unsuccessful attempts must have already been made to solve the problem or
to explain the phenomenon conceived. Such an event may come as a result of a long process of making intentional
efforts to solve the problem and possibly after a period of incubation.
Not every process of discovery ends up with a momentous eureka event. We cannot always find a simple "moment
of discovery." Sometimes the process of discovery is gradual and can only be identified as a process of discovery
in retrospect; all depends on the object of discovery. If that object is a product of generational discovery, such as a
full-fledged theory, there might
< previous page
page_35
next page >
page_36
< previous page
page_36
next page >

Page 36
not be such a moment. For example, quantum mechanics (that is, the full quantum theory, not only Planck's idea of
the energy quantum) was generated-discovered through a lengthy process. However, there are cases of generational
discovery which culminate with a final eureka event. Indeed, after a long process of trying to construct a theory or
to solve a problem there might remain a missing link without which the theory is not satisfactory. The missing link
may then be found in an eureka event. For example, Darwin describes the creative flash he experienced after he
had read Malthus, which presumably supplied him with the missing link for completing his theory.
Here we encounter again unintentional creative processes for which the discoverer cannot claim full responsibility.
According to Simonton's theory expounded in Chapter 6, it is the same phenomenon described by Poincar.
Norman Storer presents some descriptions of creative people in the arts, which refer to a similar phenomenon
(Storer 1966, 678): Creative people "have felt possessed by their work, which seems to have an independent
existence." A composer describes his experience as if "the work has come through, rather than from, him." He cites
a musician saying: "I feel as if I am an instrument through which something is speaking. ... It's as if I am standing
off from myself and watching while someting else takes over." In these processes the product, or the process, of
discovery has objective existence entirely independent of the discoverer intentions. "It has a vitality of its own so
that it has been able to use the individual in order to 'make itself real.'" The creator is ''more a spectator at the act of
creation than the author who has been fully and consciously in command of the process." The creator who has this
experience tends, therefore, to be modest since he feels that he was only a spectator at the act of creation. As we
will see, this is not the only kind of non-intentional process of discovery of which the individual discoverer is not
in full command. Another kind is related to the social dimension of science.
1.4 The Creative Element in Discovery and the Issue of Realism
1.4.1 Discovery, Invention and Creativity
As we have seen, a major connotation of the prescientific notion of discovery is to expose, to uncover or to
disclose a secret. Hence, if we view science as a tool or method of discovery, and look upon discovery in this
sense, we may be led to the very popular view that the aim of science is to expose the secrets of nature. The
following three conclusions are implied by this conception of scientific discovery:
1.The aim of scientific inquiry is to reveal something objective which exists independently of the
scientist's efforts to bring it to light, just as an inner
< previous page
page_36
next page >
page_37
< previous page
page_37
next page >

Page 37
component or mechanism of a clock exists whether or not there is someone attempting to expose it.
2.There are obstacles facing the scientist in his efforts to find out what is hidden behind the observed
phenomena.
3.The process of scientific discovery is the process of removing or overcoming these obstacles;
whenever the latter are removed, the secret is revealed intact. Thus, science does not create anything
new in the world, it only serves as a unidirectional mediator, or information channel, between the
world and the human mind.
This picture reflects a simplistic conception of scientific discovery embraced by naive realism. It views discovery
as exposure. According to this view, discovery is clearly distinguished from invention. Discovery and invention are
the two ways by which novelty is generated in human culture. Discovery is the process of exposing something
which exists independently of the inquiring mind. The latter is the object of discovery. The product of discovery is
a new item of knowledge which represents the object of discovery. Invention is a process of generating something
new according to human design, to achieve certain purposes. The product of invention is a new tool, in the broad
sense of the word. This distinction can be seen clearly in the discovery of X-rays which was followed by the
invention of X-ray photography for medical diagnosis. The discovery of X-rays was a discovery of a natural
phenomenon. It was an unintentional discovery. It seems, therefore, that the discovery was made without any active
intervention of the human mind which acted only as a passive receiver of the information. The invention, on the
other hand, required a human design, which aimed at utilizing the discovery for certain purposes. Similarly, alpha
particles and protons were employed after their discovery as tools for further research in nuclear and particle
physics, i.e. for bombarding target particles and nuclei. This pattern of discovery-followed-by-invention seems to
be reflected in the interrelation between basic, or pure, science, which is engaged in discovery, and applied science
and technology which are invention-oriented.
Yet, in science, the process of unveiling the secrets of nature leads to the creation of material artifacts and theories.
Using a new observational instrument or method has a similar effect as creating a new artificial environment. Thus,
either the object of inquiry or the observational tool is an artifact. In fact, in many cases the boundary between the
observational tool and the observed system is arbitrary and determined by convention. For example, in the case of
the discovery of radioactivity, the photographic plates may be seen as the observational tools or, alternatively, as
part of the system investigated. In a collision experiment where particle A collides with particle B the process may
be viewed as observing the structure of particle B, where B is treated as a
< previous page
page_37
next page >
page_38
< previous page
page_38
next page >

Page 38
target and A functions as a probe. However, in particle physics, the two-particle system is treated as a whole,
where A is treated as part of the investigated system.
Ordinary observation may involve throwing light upon the observed object, where the light is regarded as a means
of observation. Similarly, in a high energy photoproduction experiment the structure of a proton, say, can be
probed by "throwing" high energy photons on it. Of course, in this case, structure means something different from
the ordinary spatial notion. However, it is mainly the photon-proton system which is studied without distinguishing
between a projectile and a target. In general, the principle is the same: the colliding projectile, serving as a probe,
draws information about the properties and structure of the target object. In photoproduction experiments, for
example, the so-called form factors of the proton can be determined. And these quantites are related to the
structural properties of the proton. However, in this case the process of "observation" may involve the production
of particles and resonances which were not present before the observation. Here the generational effect of the act
of observation is manifest in its highest degree.
On the other hand, our description and explanation of the phenomena is colored by our conceptual systems and
categories. And in a highly theoretical science, which departs significantly from the observational or
phenomenological level, the scientist is not only guided by the available conceptual system and by his prior
expectations in conducting his observations and in describing the phenomena. He also looks for deeper
explanations. In order to expose a hidden cause or hidden structure behind the observed phenomena, the scientist
generates sophisticated instruments and experimental methods, on the one hand, and a highly abstract conceptual
system and an intricate theoretical machinery, on the other hand. Moreover, in trying to expose the secrets of
nature, the discoverer creates, in a sense, some of the objects of his inquiry. The products of this kind of process of
discovery seem to be human artifacts; they seem to be products of human design. In fact, any experimental system
which is designed to carry out controlled experiments can be viewed as an artifact. Hence, their products, the
experimental results, must be artifacts as well. Furthermore, these artifacts serve, in a sense, certain purposes, such
as testing a theory or explaining the data. Thus, the discoveries produced in these experiments seem to have all the
characteristics of inventions.
On the other hand, inventions have some characteristics of scientific discoveries. In many cases, as in the above
mentioned X-rays' case, an invention follows the discovery that a natural phenomenon can be exploited for certain
purposes or that an available tool may be applied for performing a new task. In other cases, the inventor carries out
a process designed for achieving a certain purpose and discovers in the end that the product of the process can be
applied for achieving a different task. In general, an artifact becomes an inven-
< previous page
page_38
next page >
page_39
< previous page
page_39
next page >

Page 39
tion only after the inventor discovers that it achieves some purpose. This is similar to the discovery that a theory
provides a successful explanation. Thus, a discovered theory may be viewed as an invention designed for the task
of providing explanation, whereas an invention can be viewed as a discovery that a certain artifact achieves a
certain task. Furthermore, if a material invention is indeed successful in achieving a prescribed task, it means that it
reflects some elements of the natural or artificial environment in which it operates. The success is achieved only if
the artifact fits certain aspects of the world to a marked degree. Thus, information about the world is embodied in
the product of invention. The very success of the invention may, therefore, constitute a discovery about the world.
For example, the airplane embodies in its structure information about the aerodynamic properties of the atmosphere
through which it was designed to fly. Indeed, in many cases, an invention leads to scientific discovery. For
example, the invention of the steam engine paved the way for the development of thermodynamics. Thus, invention
may lead to discovery and invention may follow discovery, such as in the X-rays case.
Inventiveness plays a major role in the generational discovery of theories. Theories are invented in order to explain
the observational or experimental phenomena and to resolve anomalies. Their success depends on how much they
satisfy the explanatory "needs" or requirements of the scientist in light of their agreement with observed
phenomena. Experiments are designed to help in inventing and testing theories.
In summary, discovery and invention embody both the structure of reality and human creativity. In addition,
inventions are sometimes discovered and discoveries are sometimes results of human inventiveness. Only
discovery by exposure can be sharply distinguished from invention. We would not say that Columbus invented
America, that Mendeleev invented the Periodic Table of Elements, that Boyle invented his gas law, or that
someone who derived a deductive conclusion from a set of premises invented the conclusion.
1.4.2 The case of Particle Physics: An Active Look at Matter
The issue of creativity and reality in scientific discoveryboth observational and theoreticalis most forcefully posed
in particle physics. Two peculiar features which have emerged in contemporary particle physics draw our attention;
one is related to the observational or experimental side and the otherto the theoretical side of this frontier field of
science (see Kantorovich 1982).
Experiment: Creating Material Artifacts
In the experimental arena what immediately captures our attention is that particle physicists seem to be studying
artifacts which they produce in the laboratory. For example, in the interaction p-pK0p, the K meson and the
lambda are produced from a negative pion colliding with a proton. Or, in inelastic electron-proton inter-
< previous page
page_39
next page >
page_40
< previous page
page_40
next page >

Page 40
action e-pe-N*, higher mass resonances, such as N*(1518) and N*(1688), 4 are produced. These resonances, are
short-lived states of matter. For the "outside observer" it seems that in the last three or four decades, particle
physicists have been continuously engaged in producing at high energies more and more short-lived particles or
states of matter which provide the data for their theorizing. So the question is: in what sense are we entitled to say
that high energy physicists investigate and discover the structure of matter? Or, in what sense is it "pure" nature
which they study and not nature which is highly contaminated by human intellective and active intervention?
Here the experimental method has been stretched to its extreme. The experimental physicist has always created
artificial systems or objects in controlled experiments. The chemist or the nuclear physicist even produced new
kinds of material objects, but the artificial material changes were always seen as an analysis, a synthesis, or a
rearrangement of preexisting fundamental constituents of matternot a creation of new ones. For example a
chemical reaction is a rearrangement of atoms; atoms are not created or annihilated. Of course, this description in
terms of fundamental constituents, such as atoms, already relies on some established theory of matter which is
regarded as unquestionable. But this is true for scientific experiments in general; in planning and conducting an
experiment, the scientist always has a background knowledge, or at least an initial hypothesis, about his domain of
investigation. From this basic picture, he starts his theorizing. In particle physics, we encounter a situation where
according to the prevailing conception, the artificial products do not seem to be constructed out of preexisting
material objects; the new artifacts emerge as new states of matter.
The question is, therefore, does the particle physicist discover these short-lived particles or resonances in nature or
does he create them? The same question arises in any experimental science in a milder form; controlled
experiments always create artificial environments and the question is in what sense the scientist discovers natural
phenomena in such an environment. Or why should we say that he exposes, rather than generates, these
phenomena. An answer to this question may help us to understand why a generational discovery is indeed a
discovery. In the following, I shall offer three interrelated answers to this question.
(i) One answer is that if we wait long enough, we would find the new states of matter in natural conditions, for
example, in cosmic rays. Thus, distant galaxies can be viewed as natural laboratories in which these processes
occur and the new states of matter produced. In this sense, the scientific laboratory serves just as a means of
discovering these kinds of processes and states more quickly, and in a controlled manner. The physicist is viewed,
therefore, as reconstructing or "preconstructing" natural phenomena in the laboratory. However, if we wait for
certain phenomena to be discovered in natural circumstances, we might have to wait longer than the life-time of
humankind or
< previous page
page_40
next page >
page_41
< previous page
page_41
next page >

Page 41
of life on this planet. Hence, the physicist's role seems to be more essential than the role of pre-constructing natural
phenomena, since it may happen that without the physicist's intervention, these new states of matter would never
have appeared on earth. Thus, the physicist discovers in the laboratory something which potentially exists in nature,
bringing it from potentiality to actuality. The potential existence is the objective thing whereas the actual
occurrence is contingent, depending on initialnatural or artificialconditions. Hence, in generational discovery such
as this, the physicist discovers something objective which exists independently of the discovery process.
(ii) This brings us to a second answer to the above question. In discovering a new phenomenon in the laboratory,
the only manner by which the physicist intervenes in the natural course of events is in setting up the appropriate
initial conditions; nature does the rest. When the physicist sets up the appropriate experimental arrangements and
supplies the right amount of energy, for example, he creates the conditions which are necessary for the
phenomenon to occur; he thus creates the conditions by which the phenomenon undergoes the transition from
potentiality to actuality. Whenever these conditions are created, the natural process will take place independently of
the experimenter's intentions or design. In this respect, there is no real difference between a Galileo on the tower of
Pisa and the modern high energy particle physicist in Fermilab. In the falling bodies experiment, the experimenter
just brings the body above the ground level and set it free. The body will then fall down along distances which are
proportional to the squares of the time periods. The law of free fall, i.e. the product of discovery, is independent of
the experimenter's intentions. In the same way, when the particle physicist generates electron and positron beams in
an appropriate electromagnetic field and supplies the right amount of energy in a colliding-beam experiment, he
may produce many new states of matter or particles, provided all the relevant conservation laws, such as the
conservation of energy and electric charge, isospin and unitary spin, are obeyed. Through these particle-production
experiments, he may discover some of these conservation laws. Thus, the experimenter sets up the initial conditions
and then nature takes the lead, allowing only the products which obey the laws of nature to appear.
This brings us back to the distinction between discovery and invention. The inventor constructs his artifact
according to his design. He may wish to employ known natural phenomena for achieving the task. However, he is
not waiting for the answers of nature as the scientist does. Unlike the scientist, he does not wish to let nature take
the lead and surprise him, yielding unintended results which will spoil his plans. An unintended result constitutes a
discovery. However, with respect to the original intentions of the inventor, such a discovery is an undesirable
result, whereas for the discoverer, the discovery is of course very much desired, in particular, if the discovery is
something unexpected. Indeed, an unexpected discovery reflects nature's contribution to our
< previous page
page_41
next page >
page_42
< previous page
page_42
next page >

Page 42
knowledge which is relatively independent of our expectations and presuppositions. In such a discovery human
intellective contribution is minimized and nature's contribution is maximized. Thus, the scientific discoverer wishes
to discover phenomena and laws which reflect as much as possible nature's workings rather than human
contribution.
The criterion of surprise may, therefore, distinguish the discoverer from the inventor. The discoverer who aims at
discovering natural phenomena, rather than the products of his own fabrications, should seek the unexpected. This
criterion is related to the view presented in Chapter 5 that serendipity is a major source for scientific discovery.
According to this pattern of discovery, the discoverer sets out to find a given kind of phenomenon and finds
himself discovering a different kind. According to the principle of serendipity, the discoverer should welcome
unintended discoveries. This is an epistemological principle related to the growth of knowledge. The main function
of discovery is indeed epistemic. Invention, on the other hand has a pragmatic role. The inventor aims at
accomplishing a specified task. He, therefore, designs the process of invention according to what he already knows
in order to produce the required tool. The process of invention seems to be intentional, employing for a useful
purpose what has already been discovered.
Yet, important inventions also come about serendipitously; one aims at producing a tool for one purpose and finds
out that his final product can also, or only, be utilized for another purpose. In this case, the inventor actually
discovers unexpectedly that something is useful for certain purposes. Thus, neither creativity and human design,
nor intentionality or the criterion of surprise can clearly distinguish between discovery and invention. Generational
discovery and invention stubbornly exhibit a high degree of similarity. This is not surprising if we adopt the natural
selection model for sociocultural evolution and the evolution of science. According to this view, inventions and
discoveries play the role of "blind mutations" in this evolutionary process. Hence both are not goal directed. This
view will be expounded in Part II.
(iii) The third answer to the discovery vs. generation question is that in producing "artifacts" in the laboratory, the
main aim of the scientist is not to discover the new artificial phenomena or states of matter; as I indicated in the
last section, the production of the artifacts serves mainly the purpose of discovering the laws of the deep structure
behind. For example, after the hope that the electron, the proton, the neutron and the pi-meson are the fundamental
constituents of matter faded and the list of "elementary" particles became increasingly longer, the particles were
not treated as fundamental any more and the word elementary was omitted from the title "elementary particle
physics." As a result, the importance of discovering a new particle for its own sake was diminished (cf. section
1.2). In the sixties and early seventies, the production of new particles and resonances served the purpose of
discovering or testing theories of internal symmetry of hadrons or theories of hadron dynam-
< previous page
page_42
next page >
page_43
< previous page
page_43
next page >

Page 43
ics, such as SU(3) symmetry and higher symmetries or Regge poles and Smatrix theories. The discovery of the
omega-minus particle, for example, was an important event since it provided high confirmation for the SU(3)
theory. Indeed, a typical particle theory predicted, among all other things, the existence of certain particles and
resonances which had not yet been discovered.
In general, one of the aims of the physicist is to find conservation laws and invariants. This can be done by
generating all possible material changes in the laboratory, finding out what remains invariant. Theories of
symmetry in particle physics are generalizations of such conservation laws. Hence, what the particle physicist
produces in his laboratory is not his main object of discovery. These are the laws of nature and invariant relations
which matter in this case, not so much the list of new particles.
Discovering by generation of the various particles also serves to expose the deep structure of matter. This is the
principle which was mentioned in the last section: by generational discovery on the phenomenological or
observable level of matter, the physicist exposes the deep, unobservable, structure of matter. Indeed, some of the
conservation laws related to the internal symmetries of hadrons, such as unitary symmetry, were viewed as
reflecting the deep structure of matter. Thus, when the experimenter produces the variety of particles, or discovers
them by generation, he exposes these conservation laws, or the deeper structure of matter. The deep structure of
matter is the natural, or human-independent, component of the discovery.
Theory: Creating Abstract Entities
In order to explain and comprehend the complex data generated in the highly artificial environment of
experimental physics, the theoretical physicist employs highly abstract theories, such as quantum relativistic field
theories, quantum chromodynamics (QCD) or superstring theory. The wave functions, fields, particles, colors and
strings appearing in these theories have at most only a very faint resemblance to their counterparts in classical
physics or in ordinary experience. These entities "exist" in abstract spaces, such as the infinite dimensional Hilbert
space. The quarks, for example, which are the ultimate constituents of matter, according to contemporary theories
of matter, exemplify this abstractness. They have "color" but cannot be seen. They are "particles" but defy all
attempts to be individually detected as electrons, protons and mesons. Quarks appear only as constituents of
baryons and mesons but cannot be found in isolation. The primordial model of a corpuscle had already been
stripped of its original qualities in the classical theory, where we encounter, for example, point particles or
particles which possess only the quantitative properties of mass and momentum. Quantum theory drew its basic
analogy for the particle aspects of atomic entities from the classical corpuscle model, where the possession of
kinematical and dynamical properties, such as velocity and mass, and the adherence to equations of motion
constituted the so-called positive analogy, whereas pic-
< previous page
page_43
next page >
page_44
< previous page
page_44
next page >

Page 44
turability or amenability to a causal description in space belonged to the negative analogy. Quarks are even further
removed from the commonsense notion of a particle. Thus, in order to comprehend the data of high energy particle
physics, theoretical physicists generate highly abstract theoretical constructs. The attempts to further develop and
test the theory yield more data which need further theoretical development and so forth.
It seems, therefore, that the physicist starts as a naive realist, trying to comprehend what he observes. However, his
attempts to explain the phenomena he encounters lead him to create increasingly more abstract theories whose
testing requires the production of an increasingly artificial phenomena in the laboratory. The question of scientific
realism is thus forcefully posed.
1.4.3 Epistemological Realism: Construction, Transaction and Representation
In recent years realism has been split into many versions, such as hypothetical, metaphysical, epistemological,
convergent, naturalistic, modal and constructive realism. In relation to scientific discovery, it will be useful to
adopt the distinction between metaphysical and epistemological realism (see Stein 1990). Metaphysical realism is a
belief about the world: the belief in objective reality or in an external world which is independent of human action
or thought. Epistemological realism is the view that we can know objective reality. We will be interested in
epistemological realism since it is a view about discovery: the process of discovery provides us with an access to
objective reality. From this it is implied that the object or product of discovery is a real entity, or that it refers to
real entities. Epistemological realism presupposes metaphysical realism. I will concentrate now on epistemological
realism. In doing so, it will be helpful to introduce the naive-sophisticated scale of epistemological realism, which
will encompass some of the other versions of realism.
A somewhat naive realist may view our conceptual systems and experimental and theoretical tools as our means of
interaction with the world; we do not construct the world or invent it, rather we interact with it. We construct or
generate the means of interaction: the experimental devices and the conceptual and theoretical machinery. By
employing new modes of interaction which are products of human creativity, we might expose new relations and
regularities which exist independently of the interaction. Human creativity just helps us in exposing existing
objects, relations and regularities; it helps us in "pulling back the curtain on pregiven," to borrow Bruno Latour and
Stephen Woolgar's expression (1979, 129).
A more sophisticated realist would maintain that our means of interaction give us only certain aspects of reality,
depending on our cognitive apparatus, needs and interests, but we would never know the "things in themselves," to
use the Kantian terminology. By the borrowed term "things in themselves"
< previous page
page_44
next page >
page_45
< previous page
page_45
next page >

Page 45
I refer here to objective reality in two senses: (1) reality in its totalitynot just aspects of reality (which interest us or
which are related to our needs), and (2) unperturbed reality, i.e. reality which is free from our active and cognitive
intervention. These two senses are partially overlapping; when we look at one aspect of reality, the picture we get
is not free from our cognitive intervention. Thus, according to the above view, our picture of reality, which is a
product of our interaction with reality, reflects both our own constitution and the structure of reality.
The anti-realist would say that the means of interaction are not neutral channels of information; they create a new
"reality" which cannot be attributed to something objective that is independent of us. One traditional alternative to
realism with respect to theories is instrumentalism, an approach which holds, roughly speaking, that theories are
just instruments for organizing the observational data and for predicting new events and phenomena. A theory does
not refer to real objects, structures, natural laws or relations existing in the world and does not possess a truth
value. This was the way to save empiricism which maintained that only observational terms refer and are
meaningful. Thus, theoretical terms do not refer; they are parts of the theoretical system which only functions as an
instrument.
One of the recent non-realist views is reflected in the title of Andrew Pickering's book Constructing quarks (1984).
Its main theme can be encapsulated by the claim that scientific knowledge is socially constructed. According to this
view, which is termed "constructivism," "construction" is contrasted with "representation" (Giere 1988, 57).
However, constructivism need not be contrasted with realism nor constructionwith representation. This is evident
from the transactional approach to the epistemological process. I will treat this approach, which refers mainly to
the individual knower, as one component of a wider conception of constructive realism. The other component, the
social component of constructivism, will be treated in my discussion of the social dimension of science. The
similarity between transactionalism and constructivism is reflected in the similarity between the key concepts,
taken from the realm of commerce, that are employed by the two approaches. The key concepts are "transaction''
and "negotiation." The first refers to transactional relations between the knower and the world, including his social
environment. The second refers to negotiations between the members of a community of knowers that result in
scientific knowledge. According to the second approach it seems that science is a trade-game unrelated to the
external world. When I discuss the social dimension, I will try to fuse the two approaches so that the resulting
game will bear essential relations to the real world.
According to the transactional approach, the knower builds his picture of the world through an active transaction
with it. W. Buckley, pursuing this approach and referring to transactionalists like Dewey, Mead and Piaget,
describes knowledge as follows:
< previous page
page_45
next page >
page_46
< previous page
page_46
next page >

Page 46
Knowledge is not passively and finally given merely through information input to the sensory apparatus, but
rather is actively constructed and reconstructed through continual interchange between the individual and
his physical and social environment... the world as we see and act on it is to a great extent created by us, in
a sense that we gradually build up a construction of it by interacting with it. (Buckley 1972, 189, 1934)
The last words hint at a "convergent realist" view, the view that in our learning from experience, we are gradually
progressing towards truth. Piaget's most revealing message is that genuine knowledge is acquired only through
action: "to know an object...is to act on it" (Piaget 1977, 30). Piaget finds a parallelism between the development of
individual knowledge and the growth of scientific knowledge. His classical works on the evolution of perception,
thought and intelligence in the child can provide us, therefore, with clues for understanding the development of
scientific knowledge. Thus, the particle physicist who investigates the structure of matter by manipulating particles
at high energies is playing the same game as the child who gets better results in memorizing designs of little
buttons by acting on them rather than only watching them (Piaget and Inhelder 1971).
A related thesis (attributed to Vico) which says that we know or understand best what we produce by physical
action or by creative thought can shed light on some of the most distinctive constructivist patterns in science. The
Euclidean and Archimedean ideal of deductive systematization, for example, may be viewed in this light, since
deductive systems have been perceived as genuine human products. Thus, philosophers of nature and scientists
throughout the history of science have devised deductive systems of theoretical representation in which they could
reconstruct natural phenomena by deductive manipulations, in order to make the phenomena intelligible. On the
experimental level, we see scientists reproducing or preproducing natural phenomena in the laboratory.
Understanding through material or experimental self production is intimately related to understanding gained by
mental or theoretical construction. In order to be able to reproduce an effect at will, in a controllable way in
diverse circumstances, the scientist has to know the underlying "mechanism" or laws of nature which yield the
effect. Since he has no direct access to the hidden secrets of nature, i.e. to the underlying mechanisms and laws of
nature, he himself may invent them by active theorizing. The theory guides him in setting the appropriate initial
conditions for producing the desired reconstructed natural effects. In case the results are not satisfactory, the theory
may be modified or replaced. In this way, the scientist gradually approaches the understanding of the phenomena
by mental and material creative actions
< previous page
page_46
next page >
page_47
< previous page
page_47
next page >

Page 47
which are interlinked. This applies as well to our ordinary experience. Indeed, this is exactly what Piaget says
about action and knowledge:
To know an object is not to furnish a simple copy of it: it is to act on it so as to transform it and grasp
within these transformations the mechanisms by which they are produced. To know, therefore, is to produce
or reproduce the object dynamically; but to reproduce it is necessary to know how to produce, and this is
why knowledge derives from the entire action, not merely from its figurative aspects. (Ibid., 30)
Knorr-Cetina expresses a similar view:
The constructivist interpretation is opposed to the conception of scientific investigation as descriptive, a
conception which locates the problem of factility in the relation between the products of science and an
external nature. In contrast, the constructivist interpretation considers the products of science as first and
foremost the result of a process of (reflexive) fabrication. (Knorr-Cetina 1983, 11819)
However, according to the transactional view, there is no real contrast between construction and representation
which is implied by the above passage.
There is nothing distinct in science in this respect; the child who plays with clay constructs or manufactures
different shapes and pieces of clay. However, these products are not the final targets of his activity. The final target
is the "law of nature" which he eventually discovers, or indeed fabricates, i.e. the law of conservation of matter, for
example. The child's knowledge is not purely descriptive. Scientific knowledge, a fortiori, is not purely descriptive.
If we grant that its main aim is explanation and understanding, then in generating our "representation" of the world
we cannot help being constructive.
In terms of the distinction between discovery by exposure and by generation, we express the same thing by saying
that generational discovery enables us to expose the deep structure of reality. Those who make the distinction
between construction and representation do not distinguish between the shallow and the deep levels of reality.
The distinction between the shallow and the deep levels of reality is an epistemic distinction. If we view the
growth of scientific knowledge as a stratified process, we can treat every given stage of knowledge as the shallow
level, from which we seek a deeper understanding, which means exposing a deeper level of reality. Thus, the
stratified structure of knowledge is perhaps related to a stratified structure of reality, or of the section of reality
exposed by science.
< previous page
page_47
next page >
page_48
< previous page
page_48
next page >

Page 48
The transition from the classical picture of matter at the end of the nineteenth century to atomic physics and the
transition from atomic physics to nuclear and then to subnuclear (particle) physics may be viewed as such a
process, where in order to understand some anomalous phenomena on a given level, science had to construct a new
level of theoretical knowledge and a new level of experimental technique in order to penetrate the deeper physical
level. In this process, generational discovery on the nth level leads to the exposure of the structure of the n+1th
level.
Thus, we may view observational tools (in the broad sense, including experimental methods, as well as
experimental systems employing high-technology), as the scientist's channels of communication with nature. The
introduction of a new tool means the creation of a new communication channel that provides the scientist with
information about new aspects of reality. The discovery of a new information channel is therefore a creative
process. In a similar fashion, the discovery of new conceptual system and theoretical structure amounts to a
creative discovery of a new comunication channel with the world. As we have noted, both kinds of communication
channelsthe observational-experimental and the conceptual-theoreticalare interlinked.
Thus, generational discovery of new communication channels expose new aspects of reality. As we will see, this
creative process can be viewed as an evolutionary phenomenon.
< previous page
page_48
next page >
page_49
< previous page
page_49
next page >

Page 49
Chapter 2
The Scope Of Method
In this chapter I will assume that scientific discovery is governed by reason and method. I will pretend that the
social dimension is irrelevant to the process of discovery and that unintentional discovery belongs to the realm of
curiosities. In short, I will adopt the inference-view of discovery. As we will see, the inference-view accounts for
some important features of discovery by exposure and even for some kinds of generational discovery. Yet the
attempts to capture the creative elements of discovery by logic or method do not yield significant success.
2.1 The Nature and Function of Method
As we have seen in section 1.2, the objects or products of scientific discovery are very diverse in their kind; they
may be as diverse as the discovery of a specific event and the discovery of a full-fledged theory. Yet when
philosophers of science talk about the method or the logic of scientific discovery, they do not always explicitly
distinguish between the different kinds of discoveries. In this chapter, I will refer mainly to the discovery of laws
of nature, theories and explanations. Even this category is still very heterogeneous; we have observed, for example,
the essential difference between theories and laws of nature.
In this section, I will raise metamethodological questions regarding the nature, function, form and origin of
method, rather than describe or characterize in detail any particular method.
2.1.1 Who Needs a Method?
It is frequently claimed that a scientist arrived at a discovery "by intuition." This is meant to say that the scientist
did not systematically employ a method. Since this occurs frequently, we might treat it as a phenomenon which
may be explained by psychological theory. Indeed, great scientists throughout the his-
< previous page
page_49
next page >
page_50
< previous page
page_50
next page >

Page 50
tory of science either did not employ particular methods in their discoveries or did not tell us about their methods.
In many cases, they left prescriptions which are supposed to constitute the method of discovery. However, it is not
clear that they indeed did employ this method in their actual discoveries. Sometimes it is clear that they did not.
Sometimes their prescriptions seem to be reconstructions after the fact of the discovery process. In this way, they
tried to justify discoveries which were made unintentionally or even by mistake. When I discuss the role of
serendipity in discovery, I will mention some examples of great discoveries which were made as a result of errors
in applying some method or simply errors in calculation.
Some of Newton's "Rules of Reasoning in Philosophy" (Newton 1962, 398400), appearing in Book III of the
Principia, can be viewed as rules guiding discovery. Rule IV, for example, says:
In experimental philosophy we are to look upon propositions collected [inferred] by general induction from
phenomena as accurately or very nearly true, notwithstanding any contrary hypotheses that may be
imagined, till such time as other phenomena occur, by which they may either be made more accurate, or
liable to exceptions.
It is very questionable whether such a rule would assist an ordinary scientist in making a single discovery. It is too
general and some terms appearing in its formulation are too vague. It actually says that the scientist should
discover good inductive generalizations from the phenomena. This is not a very useful direction for making a
discovery. In order to emphasize this point by way of exaggeration, we might compare it to the following useless
rule of discovery: "Construct the best theory which accords with the facts."
In fact, we do not have to invent such a rule. No less a scientist and philosopher than Descartes provides us with a
most instructive example of a method of discovery, which applies to every possible discovery and which is
(therefore!) entirely empty. In his Discourse on the Method, Part II, he provides us with the following four rules of
a method of discovery:
The first of these was to accept nothing as true which I did not clearly recognize to be so: that is to say,
carefully to avoid precipitation and prejudice in judgements, and to accept in them nothing more than was
presented to my mind so clearly and distinctly that I could have no occasion to doubt it.
The second was to divide up each of the difficulties which I examined into as many parts as possible, and
as seemed requisite for it to be resolved in the best manner possible.
The third was to carry on my reflections in due order, beginning with objects that were the most simple and
easy to under-
< previous page
page_50
next page >
page_51
< previous page
page_51
next page >

Page 51
stand, in order to rise little by little, or by degrees, to knowledge of the most complex, assuming an order,
even if a fictitious one, among those which do not follow a natural sequence relative to one another.
The last was in all cases to make enumerations so complete and reviews so general that I should be certain
of having omitted nothing. (Descartes 1967, 92)
The first rule has more implications for justification or validation than for arriving at discovery. The second rule
applies in particular to problem solving. The third rule applies to discovery of regularities in complex systems. The
fourth rule is supposed to apply to the last step in the discovery process. Leibniz reacted to these rules by saying
that they amounted to the following prescription: "Take what you need, and do what you should, and you will get
what you want."
Recommendations related to the general ways of problem-solving are given also by the mathematician Charles
Hermite. He suggests paying attention to exceptions to the rule, to anomalies, to errors and to a gap or fault in a
proof. These suggestions seem to be no better than Descartes'; they do not help us much. He advises us to correct
errors, faults, gaps, and anomalies, but he does not give us even a hint on how to do so. He does not do so since
there is no way to correct all possible errors etc.
We can, therefore, formulate the following metamethodological "rule": The usefulness of method is inversely
related to its degree of generality. In this respect there are two extreme kinds of method: the most general and the
most specific. The most general, as we have seen, is empty. The most specific is very useful in very specific cases
which yield no real discoveries. For example, an algorithmic recipe for baking a cake is very specific and very
useful but yields no exciting discoveries. At most, the cake will taste a little better. The most specific methods of
the recipe type consist of a list of directions for repeating a known process under specified conditions. For
instance, how to grow a certain plant in specified conditions, how to produce a specific chemical reaction with
given quantities of materials or how to solve a given system of linear equations. Not much room is left for
inventiveness or creativity in using a recipe-type method for the exact task the method was designed to fulfill. A
method for growing a general kind of plant under various conditions, or a method for solving a general category of
systems of linear equations, leaves more room for creativity. The more general is the rule, the more creativity is
needed for applying the method in a specific case. Thus, creative minds are needed for applying very general
methods. And the most creative minds are needed for applying the most general methods, which are empty
methods or no methods at all. No wonder that a Newton or a Descartes provide us with the most general methods.
Following this type of argumentation, we arrive at
< previous page
page_51
next page >
page_52
< previous page
page_52
next page >

Page 52
the conclusion that the scientific genius is needed to make discoveries in cases when no appropriate method is
available.
So the question is whether genuine science is an elitist enterprise which can only be carried out by a few great
scientists who do not follow any explicit method, or can it be done by an army of scientists who follow prescribed
methods. The answer is that modern science needs both kinds of scientists: the great discoverers who do not
employ prescribed methods, or invent them in the course of their investigations, and the scientists-technicians who
follow prescribed recipe-methods. Big and democratic science needs more of the latter than elitist science. And on
the scale between the great discoverer and the technician there is a whole spectrum of combinations thereof. Yet
even a technician may hit upon a great discovery "by chance," without employing a method.
When we ask who needs a method, be it a method of discovery or a method of evaluation, we must note the
symptomatic fact that scientists do not study "scientific method" as part of their ordinary scientific curriculum.
They only study the specific methods required for research in specific areas. Scientific method is studied by
philosophers and historians of science. They study it mainly since they think it reflects the rationality of science.
Thus, we may arrive at the seemingly counterintuitive conclusion that perhaps rational behavior in science is
intuitive or unintentional; the rational agent acts intuitively, unaware of the method he is actually using, if any.
The answer to the question whether there is a method of discovery depends on the kind of discovery we are
referring to. If we adopt Kuhn's distinction between "normal" and "revolutionary" science, we might say that in
normal science, where the scientist is engaged in problem solving, great discoverers are less needed than
technicians. Great discoverers are needed in a revolutionary phase, where the known methods of problem solving
are not applicable. Indeed, as we will see later, many of the great scientific discoveries were not made with the
guidance of an existing method. However, the distinction between normal and revolutionary science is not as sharp
as Kuhn presents it; revolutionary developments may gradually emerge from normal research. In some cases, it is
the collective work of many normal problem-solvers which yields a great discovery, although a great scientist may
be needed for recognizing the discovery, i.e. for discovering the new picture which has emerged, or for making the
final decisive step in the process. The discovery of the special theory of relativity was to a large extent this kind of
discovery.
2.1.2 What is a Method of Discovery Supposed to Do?
It should be stressed that we are not dealing here with the so-called scientific method. In the twentieth century,
under the reign of logical empiricism, the term scientific method refered mainly to the methods of justification or
evaluation of the products of scientific discovery rather than to the methods of gen-
< previous page
page_52
next page >
page_53
< previous page
page_53
next page >

Page 53
erating, or arriving at, discoveries. The reason for this attitude is that it was believed that the rationality of science
lies merely in justification and validation, whereas the process of discovery is a psychological phenomenon to be
dealt with by empirical investigations, rather than by logical or epistemological analysis. As well will see in the
following chapter, the distinction between the context of discovery and the context of justification, which has
raised important objections, has lost its currency. So perhaps the term scientific method should refer to the methods
of discovery no less than to the methods of evaluation, restoring its original meaning which was employed by
Francis Bacon and William Whewell, for example.
What, then, do we expect a method to do for us? First, we should ask, what is our goal. We might be interested in
rationality, for example. We might ask what method would guarantee that rational decisions will be made by
scientists. This would apply to questions of selection and evaluation which have implication for the acceptance and
rejection of laws, theories and explanations. The question of rationality would also apply to the manner in which
we arrive at our discoveries. However, rationality might have two connotations: categorical rationality and
instrumental rationality (see Giere 1988). Categorical rationality is independent of any particular goal we have.
Questions referring to the rational way of achieving a specific goal are related to instrumental rationality. In this
case, the method of discovery will provide us with effective means for achieving the goal. For example, we might
ask, what is the rational way to arrive from point A to point B in the shortest time. Or, how should we construct
our theories if our goal is to make successful predictions, to be useful to society or to approach truth. Perhaps we
can relate the two notions of rationality by saying that when our goal is truth, instrumental rationality coincides
with categorical rationality. Thus, our views about the goals of science will bear upon the methods of science,
including the methods of scientific discovery. Traditionally, philosophers of science thought that only the context
of justification or evaluation, in which theories are selected, has a rational dimension. In the next chapter I will
offer some arguments as to why a theory of rationality should have implications for the process of discovery.
Perhaps we would be very pleased if we had an algorithm which, when fed with the data and with basic theoretical
assumptions, would yield a successful explanatory theory. Or perhaps we would be happy if we had possessed
algorithms for solving problems and resolving anomalies. If these were available, then scientific investigations
could be carried out by computers and the era of great discoverers would pass away. Of course, we are far from
this dream. And perhaps it is not such a good dream at all, since many hope that scientific rationality is not
mechanistic and that science is a genuine human enterprise which cannot be replaced by a machine. What would
happen to the ethos of science if a Copernicus, a Galileo, a Newton, a Darwin or an Einstein could be replaced by
a discovery machine, and the discovery of the heliocen-
< previous page
page_53
next page >
page_54
< previous page
page_54
next page >

Page 54
tric world picture and the theories of universal gravitation, natural selection and relativity would not differ in
principle from a production line? Thus, the less we demand from a method of discovery, the more we leave for the
working of human creativity.
The following are some functions we may expect a method of discovery to fulfill and some general requirements
we may expect it to obey.
(a) If the method of discovery should yield a statement, such as an empirical generalization or a statement
expressing a law of nature or a theory (in the traditional sense), we may require that a method of discovery will
lead to a true product. However, this requirement is too strong and its implementation cannot be validated. It is too
strong since it will not account for the discovery of laws of nature which were stepstones in the history of science,
but turned out to be falseas I mentioned in Chapter 1. Perhaps most of the laws of physics are strictly false and this
does not diminish their value; each such false law yields successful predictions and explanations and contributes to
the advance of scientific knowledge. Thus, we might say in a Popperian spirit that the way towards truth is paved
with bold "lies." Moreover, if the product is non-analytic universal statement, we have no way of determining
whether our method leads to truth. Empirical generalization, laws and theories yield predictions about events which
will occur in the future. As long the future is open, we have no way of knowing whether these predictions will turn
out to be true and thus we have no way of knowing whether our purported discovery is true. So that even if we had
a truth-producing method, we would never know that we had it. It is thus senseless to pose a requirement which we
could never know when it is fulfilled. Thus, instead of asking that a method of discovery will yield a true product,
we might require that it will generate a satisfactory product, i.e. a product which satisfies the standards or
desiderata prevailing in science. Such a method may yield a product which is approximately true, plausible,
acceptable, complies with the world picture and established theories or provides a good explanation according to
the prevailing standards of explanation. Each of these standards will determine the nature of the method.
(b) A weaker requirement is that the method of discovery will yield a product with a high probability of being true
or satisfactory. For example, the cannons of inductive generalization may guide us in constructing generalizations
which are highly probable. Also Bayesian theory may advise us in this respect; for example, we will see that when
we construct a hypothesis with a high "prior probability," it will have higher chances of being confirmed. Hence, if
we have a method for constructing hypotheses with high prior probability, it will enhance the chances of making
discoveries. Thus, under this requirement, a method of discovery may produce good candidates for pursuit or for
selection, rather than produce the ultimate product of discovery. Such a method, supplemented by a method of
evaluation and selection, will constitute a
< previous page
page_54
next page >
page_55
< previous page
page_55
next page >

Page 55
method of discovery in the strong sense, which leads to a unique product of discovery.
(c) The method of discovery should yield an efficient discovery process. This is a necessary requirement since it
would be possible to arrive at many discoveries if we guessed blindly or waited long enough for chance discoveries
to occur. In this case, the progress of science would be too slow. For example, let us assume that we confine
ourselves to non-revolutionary science where the process of theoretical discovery employs a given scientific
vocabulary without introducing new concepts. If we conceive our objects of scientific discovery as statements in a
given scientific language L, we might look at the process of discovery as a search process in L. The process would
be searching for theoretical statements expressed in L. Thus, among the unlimited number of strings of words
which can be formed in L, we would search for those which are grammatically correct, which are meaningful, and
which satisfy our requirements for explanation. We will have to wait some time before we discover the strings of
words expressing a simple law of nature such as Boyle's law, and a longer time for the string expressing a theory,
such as, the theory of superstrings in particle physics. Since we will be searching in an infinitely large field, the
task of finding the simplest law might never be accomplished if we do not have some method for narrowing the
range of search. This is the same as waiting for the Eiffel Tower to be constructed by chance events, without
human intervention. To be sure, there is a non-zero probablity for this to happen. Efficiency is, therefore, not a
marginal requirement. Thus, we may view a method of discovery as supplying criteria for limiting the range of
search or for making systematic search.
Yet, even systematic efforts may be inefficient and time consuming. Discoveries can be made by carrying out
tedious calculations and making lengthy efforts. We can make an analogy with arithmetic operations. We can
discover the product of 258983 by carrying out 258 additions of 983, or by using the multiplication method of
multi-digit numbers. The second way is much more efficient and elegant. The method of discovery in this case is a
discovery in its own right. Furthermore, it is more important than the particular discoveries it produces. In
mathematics there are many examples of this kind. A discovery of a simple or short proof or a method of
calculation or computation is sometimes no less important than the discovery of a new theorem. This is true of
empirical science as well. Indeed, there are cases where a solution of a problem or a new explanation is
enthusiastically accepted by the scientific community as a great discovery since it was discovered in an elegant and
simple way, in the sense that the process of arriving at the product of discovery was quick and efficient. Thus, this
requirement applies to the process of discovery rather than to its product.
Two important conclusions can be drawn here. First, an efficient or elegant method of discovery is a discovery in
its own right. Second, an efficient or
< previous page
page_55
next page >
page_56
< previous page
page_56
next page >

Page 56
elegant method would contribute to the positive evaluation of the product generated by using this method.
Therefore, the product may be considered to be a discovery mainly because the method of generating it was
efficient. This is one of the cases where the ''context of discovery" is relevant to the "context of justification" (see
next chapter).
A method of discovery which fulfills requirement (b), i.e. which generates plausible products or products with high
probability, is more efficient than a method which does not fulfill this requirement, since the former generates a
smaller number of candidates for pursuit which have higher chances to be selected, so that time is not wasted on
too many candidates.
(d) An important requirement is that the method will be general enough. A method which is designed to generate
only one discovery or a specific kind of discoveries in a restricted domain, cannot be treated as a method of
scientific discovery. For example, a method for discovering the chemical composition of a material, or of a certain
kind of material, is not a method of scientific discovery. This should be made clear if we ask what is the purpose
of the philosopher of science in investigating the issue of discovery. When philosophers of science look for the
method of scientific discovery, they look for something which reflects the nature of science per se, rather than for
a method in a specific area of science, which is derived from the specific knowledge of that field. Philosophers
who are interested in the epistemology of science or in the rationality of science would not be interested in
methods of solving Schroedinger's equation or in methods of discovering the chemical composition of a material.
This is the territory of the scientists themselves. Philosophers would rather be interested in the general methods
reflecting rational ways of knowledge acquisition.
In this respect, the interest of the philosopher of science may differ from the interest of AI scientists or technoscientists. The latter are interested in finding a method which will assist them in solving their specific problems.
Both AI scientists and techno-scientists would benefit from the discovery of a general method of scientific
discovery. However, since they are not interested in epistemology or in the nature of science per se, they would be
happy if they had area-specific methods of solving problems. This is exactly what is provided in AI by expert
systems. The heuristic method of experts for solving problems in a specific field is programmed into an expert
system which directs problem-solving in that area. The shortcoming of such an approach is that every research area
needs a different expert system. Thus, the interest of the AI scientists in general methods or in the general
characteristics of expert systems is only pragmatic. By the same token, the techno-scientists do not need a general
method of discovery since they are interested only in solving their specific problems.
Nevertheless, in recent years many philosophers of science have been interested in content-specific methods of
discovery. This trend has emerged as
< previous page
page_56
next page >
page_57
< previous page
page_57
next page >

Page 57
a result of the unsuccessful attempts to find a universal logic of discovery. It can be seen also as an influence of AI
research in heuristic-search and expert systems on the methodology of science. This trend is presented sometimes
as one of the characteristics of the new approaches to the philosophy of science. The exclusive engagement with
content-neutral logic or method of science is sometimes even claimed to be a mistake of traditional philosophy of
science. Methodological pluralism is one of the popular slogans of these approaches. However, if the philosopher
of science deals with area-specific methods, he abandons the task of investigating the nature of science and treats
science as an ad hoc grouping of different areas of research. The only thing which might then be left to
characterize science is its subject matter, i.e. the study of nature, human society, etc. However, this will not be a
distinctive characterization, since science deals with almost everything in the world, natural, human and artificial.
It is thus the method or the nature of science, rather than the subject matter, which distinguishes science from other
fields of knowledge or other human activities.
Furthermore, the scientists engaged in active research are the best experts in investigating their area-specific
methods. Thus, no room is left for the philosopher or methodologist of science except perhaps in recording and
explaining these methods for the benefit of philosophers or historians, and at most in systematizing them. This by
itself would not justify their occupation as philosophers of science. Only when they study the area-specific methods
in order to find their common features, might their investigations be of genuine interest and value. But if they study
the common features, they are engaged again in the general methodology of science, i.e. in the study of the
scientific method.
(e) A general function of a method of discovering theories, and scientific method in general, is making the growth
of scientific knowledge minimally continuous. No discovery of theory comes "out of the blue." Any new idea,
theory or a solution of a problem should bear some relationship to the present body of knowledge and the
standards of knowledge. The present body of knowledge and the standards of truth, plausibility or good
explanation set up restrictions on the product of discovery. Even a revolutionary discovery is subject to some
minimal restrictions. Thus, the method of discovery should supply a link between the old and the new. Otherwise
we would not be entitled to talk about "growth" or "progress."
(f) Finally, a more fundamental question should be raised: Do we expect the method of discovery to be normativeprescriptive or descriptive? If our methodology of discovery is derived from an a priori philosophical theory, such
as an epistemological theory or theory of rationality in the traditional sense, then it would attribute a normative
import to method. Thus, a partial answer to this question will be given in the next section, when I will discuss
possible sources of method. A fuller answer will be given in Chapter 4.
< previous page
page_57
next page >
page_58
< previous page
page_58
next page >

Page 58
2.1.3 The Origin of Method

Now that we know what functions a method of discovery is supposed to accomplish, we may ask what might guide
us in finding the method(s) of discovery.
(a) Logical Sources Deductive logic may help us in exposing the logical content of our statements but will not help
us in arriving at novel truths. Inductive logic and theories dealing with probable and plausible inference will help
us in devising methods of discovering plausible statements. Efficient methods may be derived from decision
theory. All these are formal theories dealing with inference and choice and do not refer to the content of the
product of discovery.
(b) Epistemological Sources Philosophical views from which the methodology of discovery may be derived are of
two kinds: epistemological and metaphysical. Since discovery has an epistemic dimension, philosophical theories
of knowledge may direct the philosopher of science to construct a methodology of scientific discovery. Empiricist
theories of knowledge might recommend, for example, constructing theories from observational statements;
whatever "constructing from" means. Certain versions of empiricism would have implications for the product of
discovery. For example, they might recommend avoiding the usage of theoretical statements which cannot be
reduced to observational statements. They might requirein the spirit of logical positivismusing only statements
which are verifiable. The Popperian theory of knowledge would recommend trying "bold" falsifiable conjectures.
Verifiability and falsifiability are examples of criteria which would have implications for the method of
discovering a law of nature or for constructing a theory. They would specify what kinds of building blocks are
allowed for constructing the product of discovery. Thus, when the scientist tries to construct a theory, he would
have in mind these criteria as part of the method of discovery; they would partially guide him in the process of
discovery. He would intentionally try to construct his theory in such a way that the product will obey these criteria.
(c) Metaphysical Sources Metaphysical beliefs about the nature and the structure of the world will have
implications for the process of hypothesizing laws and theories as well as for the acceptance or rejection of
hypotheses. Those who believe that the structure of the world must be simple, would try to construct the simplest
possible laws and theories. For example, the metaphysical principle that nature always chooses the simplest path
guided the medieval scholar Robert Grosseteste to suggest a law relating the angle of refraction and the angle of
incidence for a light ray entering into a denser medium. His argument was the following. The law of reflection
states that the
< previous page
page_58
next page >
page_59
< previous page
page_59
next page >

Page 59
angle of reflection equals the angle of incidence. Hence, the ratio 1:1 was already "occupied." The next available
relation in the order of simplicity was 1:2. Hence, he suggested the law that the angle of refraction equals one half
of the angle of incidence. Another medieval scholar, William of Ockham, adopted another principle, known as
"Ockham's Razor," which stated that our theories about nature should be simple. Nature itself is God's creation, and
there can be no rules restricting God in creating the world. We would say in our terminology that God does not
need method to guide him in inventing the world. We need rules for guiding us how to discover what God created.
However, theories about the world are our own inventions, so that we can demand of them to be simple. A
methodological rule implied by this principle would instruct us not to introduce more concepts in our theories than
are absolutely needed for explaining natural phenomena. Indeed, this principle is reflected in the first of Newton's
"rules of reasoning in philosophy" which says: "We are to admit no more causes of natural things than such as are
both true and sufficient to explain their appearances" (Newton, ibid.). Thus, whereas Grosseteste's methodological
principle originated from a metaphysical belief, a belief about the nature of nature, Ockham's principle of
conceptual parsimony may be related to an epistemological view, a view about the means of acquiring knowledge
about nature.
The birth of modern science is signified by another metaphysical belief, the Pythagorean outlook, which was
adopted by Copernicus, Kepler and Galileo. The Pythagorean view was expressed by Galileo's famous dictum that
the book of nature is written in the language of mathematics. Both Copernicus and Kepler believed that
mathematical harmony is what really exists behind the appearances. Kepler was guided in discovering his laws by
the Pythagorean mathematical models.
Moreover, the very fact that scientists look for laws of nature and theories is based on the metaphysical belief in
the uniformity of nature or the belief that natural phenomena are governed by laws. This is the most fundamental
metaphysical belief behind natural science. This heuristic principle paved the way to the discovery of the laws of
nature. It stands, therefore, implicitly at the basis of every method of discovery in science. The principle of the
uniformity of nature posits one of the major goals of science: to find the laws and invariable relations governing
natural phenomena. The method of discovery should, therefore, tell us something about how to find the regularities
and laws of nature. For example, methods of generating inductive generalizations tell us something about this task.
Descartes aimed at deriving the laws of nature from metaphysical principles. Modern science does not adopt this
approach. Metaphysical views sometimes serve as heuristics for discovery. For example, we may view atomism as
a metaphysical view which modern science inherited from Greek philosophy. Atomism guided chemists and
physicists in discovering some of the most suc-
< previous page
page_59
next page >
page_60
< previous page
page_60
next page >

Page 60
cessful theories of matter. Thus, we may treat metaphysics as supplying science with a repertoire of ideas for
constructing theories. However, metaphysics does not dictate to science what ideas to adopt; it is science which
selects the ideas from the metaphysical repertoire, according to the methodological criteria prevailing in science.
Thus, metaphysical views affect scientific discovery on two levels: (1) On the subject matter level, views about the
structure of the world (e.g., atomsim) supply optional ideas for generating discoveries (e.g., composite models). (2)
On the methodological level, general views about the structure of reality (e.g., the uniformity of nature) yield
definite prescriptions for generating the method of discovery (e.g. inductive methods).
(d) Learning from the History of Science Another source for method is the history of science. Scientists adopt
methods and models which prove successful. According to Kuhn, normal science is characterized, among all else,
by certain models which everyone tries to imitate. Thus, science learns from its past, in the domain of method as
well as in the domain of content.
(e) Naturalistic Theories of Science If we treat science and scientific discovery as a natural phenomenon, or as part
of the domain dealt with by science, various theories of science itself will guide us in constructing our
methodology. For example, psychology may aid us in devising methods of learning from experience and for
acquiring new knowledge. Sociology will be a natural source for method if we believe that science has an essential
social dimension. And so is anthropology, in case we view science as a cultural phenomenon.
2.2 Inferring and Reconstructing
Now we come to the methods of discovery themselves. I will sometimes use the word method in a broad sense to
refer also to the logic, procedure, heuristics or strategy of discovery.
If a method is effective, it means that scientific discovery is not entirely a matter of chance, intuition or a flash of
insight. However, we can hope that a method-guided discovery and creativity do not exclude each other. We can
order all kinds of methods according to the degree of novelty they can produce. By deductive inference, we can
only expose information hidden in the premises of an argument. It, therefore, does not produce new information. It
may only produce epistemic or psychological novelty. Inductive inference produces empirical generalizations
which contain new information, e.g. about the future or the past. A method of generating a theory should generate
novel concepts and ideas, whereas the most radical methods, if they exist at all, may generate new conceptual
systems or new world pictures.
< previous page
page_60
next page >
page_61
< previous page
page_61
next page >

Page 61
2.2.1 Reasoning vs. Creativity

There is a widespread belief that scientists reason to their products of discovery. According to this view, the high
esteem given to the discoverer is due to his bright arguments and sharp reasoning. In this respect, the physicist or
the biologist would not differ from the logician, the mathematician or the philosopher. If great discoveries in
natural science were due to chance or unexplicable intuition, the proponent of this view may ask, why should the
discoverer be more admired than the successful gambler or the fortune teller. In general, we do not attribute an
intellectual or cognitive ability to someone who is a successful gambler. Extrasensory powers, and even
extraordinary memory or computational ability, are treated as interesting phenomena rather than as intellectual
attributes. Human faculties which draw our esteem are only those to which we hold the individual possessing them
responsible. A high power of reasoning is among these faculties, whereas luck or extrasensory perception, and even
extraordinary memory or computational ability, are not. The latter would belong to the category of interesting or
curious phenomena. We would not admire, for example, someone who can mechanically repeat everything he
hears or reads; a parrot or a machine can do the job. In this respect, intuition is on the borderline between cognitive
ability and a natural phenomenon. Sometimes we may admire a scientist having an intuition in the sense of having
an insight which is not derived from reasoning and which cannot be fully communicated. However, this notion of
intuition is too general and too vague. It only indicates that we cannot explain how a scientist arrived at his
understanding or at his discovery.
However, a power of reasoning is not the only thing required of a scientist. It seems that discoverers in science
need to have some additional capability. One can reason well within a given framework of thought, i.e. within a
given world picture or a given conceptual system, employing given methods, tools or patterns of problem solving
and research. Reasoning in these cases consists of devising arguments within the existing framework. This requires
an analytical power rather than imagination and creativity. However, in many important cases problems are solved
and understanding is gained by going beyond the existing framework, e.g. by generalizing or abstracting, by
making connections or analogies with other phenomena or domains or by inventing new concepts or new methods
and tools of investigation. In order to depart from the existing system, imagination and creativity are needed; the
discoverer does not only play with the existing concepts, ideas and tools, he also invents new ones. Thus, creativity
is necessary for arriving at new ideas and systems. The power of reasoning is needed for deriving the implications
of the new ideas and for finding the relations they have with existing ideas. Creativity, as well as analytic
capability, is essential to the processes of problem-solving and theory-construction.
< previous page
page_61
next page >
page_62
< previous page
page_62
next page >

Page 62
Creativity, no less than analytical power, draws our high esteem, since unlike luck or extrasensory perception, we
regard the individual possessing this faculty responsible for it. Unlike extraordinary memory, computing capability
or psychic powers, creativity involves making judgments and decisions. It seems that creativity is a natural
endowment which cannot be learned or instructed by method. In Part II I will discuss theories of creativity which
might help us in cultivating this natural talent. We have already encountered one possible mechanism described by
Poincar. It consists of generating combinations of ideas and selecting the good ones. Poincar pointed out that
generating new combinations or new ideas is a trivial and valueless task. However, creativity does not merely
involve combinatorical capability or a wild imagination or association. The creative act involves finding or
constructing a new combination which is useful and fruitful, which solves a problem or explains some phenomena,
and which "no one has thought about before." Poincar's view is that the discoverer's subconscious process of
selection emits ideas which are already preadapted so that the remaining task is a relatively easy task of
consciously selecting the final idea which is the product of discovery. This view would diminish the value of the
conscious act of discovery and the discoverer would not be regarded as fully responsible for the discoveries he
produces; rather he would be treated as a human discovery-generating machine.
2.2.2 Discovery as Inference or Reasoning
If the discoverer arrives at, or generates, his discovery via a process of inference, it means that he arrives at his
object or product of discovery in the first place by arguing to it. There are cases where the scientist tries to justify
the product of discovery by arguing to it only after he has arrived at the discovery by chance, by error or as a result
of a process which does not seem to support the product of discovery. This might be done when the discoverer
wants to make sure that the product is supported by some firm argument, or when the discoverer wants to persuade
the scientific community to accept his product, which otherwise will not be considered to be a discovery. But then
the inference does not fulfill the role of method of discovery, i.e. a method of arriving at a discovery, exposing or
genetrating it in the first place.
If the product of discovery is arrived at, or generated, by proper inferencei.e. by applying established rules of
inference and starting from reliable premisesthe product is born justified due to the inference procedure. In this
case, the inference has both generative and justificatory functions, yielding a full-blown discovery. In the case
where it seems that no proper inference was made in generating the product, an inference is brought in to carry out
the justificatory role. Let us make a comparison with mathematical discovery. The mathematician may arrive at a
theorem, by empirical investigations (in case of a theorem in Euclidean geometry, for example), by intuition or by
chance,
< previous page
page_62
next page >
page_63
< previous page
page_63
next page >

Page 63
believing he has exposed or generated a true theorem. However, in order to convince himself and others in the
truth of the theorem he has to discover a proof. Only after the theorem is proved, is the discovery completed.
However, unlike in the case of mathematical discovery, in natural science the context of justification has an
empirical dimension. An empirical generalization or an explanatory hypothesis would not be declared a discovery
unless it undergoes some empirical tests. The reason for this is that in these cases the process of discovery is
ampliative or content-increasing. In other words, the product of discovery contains information which was not
contained in, or logically implied by, the state of knowledge before the discovery was made. Thus, in natural
science, unlike in mathematics, a proper ampliative inference cannot by itself generate a discovery. This was not
the situation before the advent of modern natural science. Some philosophers of nature believed they could derive
the laws of nature from general metaphysical principles. Even some of the founders of modern science still
believed in this. Descartes, who witnessed the birth of modern science and was himself an active scientist, sought
to discover the laws of nature by deductively inferring them from some "a priori" or "necessary" truths. The
justification of the laws thus derived was, therefore, granted, without any need for a posteriori, or empirical,
testing. To be sure, an important role was given to observation in Descartes' scientific method; in order to derive a
statement about a particular event or phenomenon, one has to include among the premises statements describing
the observed initial conditions of the particular process, as well as statements expressing laws of nature. However,
these predictions could not serve as tests for the laws of nature since the latter drew their validity from the a priori
truths from which they were derived.
Thus, in viewing the whole process of discovery in natural science as an inference, we must include the a
posteriori tests as part of the process of inference. This can be done by adding the data obtained through the tests to
the premises of the inference. Yet, the a posteriori part of the inference will not be generational. Hence, if we
maintain that the discoverer argues to his discovery, then one part of the argument is generational and the other part
is justificatory. Both parts are essential to the discovery process. The justificatory part in turn divides into a pretesting ("a priori") part and an a posteriori part.
In his discovery-generating argument, the discoverer starts with a set of premises p and ends up with the discovery
d as a conclusion. The process can be symbolized as pd, or "p entails d." Thus, in order to represent the process of
discovery as an inference we have to specify what are the premises p and what are the inference rules. If in a
particular historical case of scientific discovery, we think we know what were the premises and what were the
inference rules but we see that d cannot be derived from p, we have two options. The first option is to conclude that
the inference view of discovery is refuted by this case. The second option is to adopt what may be called an ad hoc
strat-
< previous page
page_63
next page >
page_64
< previous page
page_64
next page >

Page 64
egy. If we believe in the inference paradigm, we can do what the scientist would do in the occasion of a
disagreement between his paradigm and the observed phenomena. If, for example, the Newtonian astronomer
observes a deviation of a planet's orbit from what is predicted by his established theory, his explanatory paradigm
would guide him to look for a yet unobserved heavenly body which might have caused the deviation. Similarly, we
would introduce ad hoc modifications or auxiliary hypotheses in the description of the phenomenon, i.e.in our
casein the description of the inference. The logician of science has at his disposal an arsenal of possible ad hoc
maneuvers. In most cases, scientists do not specify all their assumptions or presuppositions. Thus, the logician of
science who believes that the scientist argued to his discovery, might hypothesize that there were some missing or
suppressed premises which were not explicitly expressed by the discoverer or that were overlooked in
reconstructing the process of discovery. Another option at his disposal is to introduce modified or new rules of
inference which, as he might claim, were effective in the process of discovery; these rules are rarely stated
explicitly and in most cases one has to conjecture what rules might have been implicitly used. Of course, in order
to be acceptable, these ad hoc hypotheses should yield an appropriate explanation of the process of discovery
according to the logician's standards of explanation. For example, the ad hoc modifications might be required to be
relatively simple and fruitful in explaining additonal cases of discovery. Thus, the task of the logician of science is
to solve a problem within the framework of his inference-paradigm. If we adopt the Kuhnian conception of normal
science, we might say that this is the same situation which confronts every scientist when confined to normal
science activity.
Now, let us review the different inference procedures at our disposal.
(a) Discovery by Deductive Inference As a first candidate for our logic of discovery we might start with deductive
logic. On the scale of novelty-generating methods, deductive inference will be situated at the bottom, since it
cannot generate new information at all. As we have seen, it can serve as a method of exposing the informationcontent hidden in our set of premises p. To be sure, this kind of exposure is a very important kind of discovery.
For example, the prediction of the existence of electromagnetic waves was exposed in Maxwell's equations by this
kind of inference, without generating any new information which was not contained in the equations. However, the
discovery of Maxwell's equations themselves was a generational discovery. Historically, the above discovery by
exposure was part of the process of discovering Maxwell's theory. Indeed, every generational discovery of a theory
is followed by the derivation of predictions which confirm the theory and which are discoveries in their own right.
An example of deducing a law of nature from known premises is Newton's deduction of the inverse square law of
gravity from Kepler's Third Law of planetary motion (see Zahar 1983).
< previous page
page_64
next page >
page_65
< previous page
page_65
next page >

Page 65
Deductive logic played a central role in the Cartesian system where in the deductive hierarchy of statements the
laws of nature were inferred from the a priori principles. However, even if the laws of nature were theorems in a
deductive system, deductive logic will not suffice for discovering them. Theorems can be generated by successive
applications of the rules of deductive inference, beginning with the postulates or the axioms of the system.
However, if we apply the rules at random, without the guide of some method or heuristic, or with the guide of an
unsuitable heuristic, we would very rarely hit upon an interesting or useful theorem. Pythagoras's theorem, for
example, was first discovered empirically (for specific triangles), with the guidance of land measurement
experience, and only later was a proof discovered within Euclidean geometry. As we have noted, a method of
discovery should make discovery efficient. Deductive logic unaided by heuristic principles is not an efficient way
for discovering interesting theorems. Actually, it is not a method at all! Indeed, someone who knows all the
necessary rules for deducing a given theorem in Euclidean geometry would not necessarily discover the theorem if
he does not have the right intuition or creative power. And he might not discover the proof of the theorem when
the theorem is given. Indeed, the discovery of a deductive proof is not a deductive process. One has to find or even
invent the steps in the proof, applying the various rules in an appropriate order. It is no different from the situation
which faces the chess player who knows all the rules of the game but does not know how to overcome his
opponent. It is again a heuristic search or an intuitive or creative act. Thus, intuition, creativity and heuristic
principles are sometimes needed for making discoveries through deductive inference, although these are discoveries
by exposure which do not produce new information. Thus, discovering the logical content of p is not entirely
analogous to the opening of a closed box and exposing its content. It is more like digging for gold or drilling for
oil. The availability of oil drilling techniques are only a necessary condition for finding oil. The oil driller should
consult the geologist in order to enhance the chances of finding oil. The "premise" here is earth, but the oil driller is
not interested in most of the material he finds hidden in the ground; everything, except oil, found hidden in the
ground is indeed there, but is useless for him. If p is true, everything deduced from p is true, but most of the truths
hidden in p are uninteresting, and therefore are not discoveries.
(b) Discovery by Inductive Inference The next candidate for the logic of discovery is inductive logic which is an
ampliative logic. The conclusion "all ravens are black" goes, indeed, beyond the premises which describe all ravens
observed until now, none of which has been non-black. Thus, induction by enumeration is a novelty-generating
rule of inference. However, since in a "valid" novelty-generating argument it is possible that the conclusion is false
while the premises are true, we may not use the term valid for describing a
< previous page
page_65
next page >
page_66
< previous page
page_66
next page >

Page 66
good inductive argument. Such an argument may be better termed inductively ''strong" rather than inductively
"valid." Hence, the truth of the premises does not guarantee the truth of the conclusion; it only endows the
conclusion high probability, whatever this means.
(c) Retroductive or Abductive Inference Inductive inference can generate empirical generalizations, but not
explanatory theories (in the traditional sense of statements). Newton's theory of universal gravitation cannot be
inductively inferred from the data on planetary motion and even not from Kepler's Laws. Although it can be shown
that the inverse square law can be deductively derived from the Third Law, Newtonian theory includes novel
concepts beyond the mathematical relation expressed by the law; for example, it includes the concept of mass and
it applies to all material bodies, not only to the Solar System. Induction by enumeration or inductive generalization
cannot lead from the data on gas behavior, or from the empirical gas laws, to the kinetic theory of gases. For one
thing, the explanatory theory employs new concepts which do not appear in the pre-theoretical vocabulary. These
concepts cannot be generated by inductive generalization, since every predicate appearing in the conclusion of an
inductive argument must appear in the premises. It is customary to think that the first stage in a development of a
science is the descriptive or phenomenological stage, which is concerned with inductive or empirical
generalizations, and the next stage is the stage of theoretical explanation, which involves conceptual growth. Thus,
since there seems to be a method governing scientific discovery on the phenomenological level, i.e. the inductive
method, it is tempting to look for a method which may govern scientific discovery on the theoretical-explanatory
level. This method should lead from the data, the phenomena or the empirical generalizations requiring explanation
to the theory which will explain them. The name "retroduction" which is sometimes given to such a method may
give the impression that it belongs to the same family of scientific methods or logics to which deduction and
induction belong. However, beyond the hope of finding such a method, which presumably looks like an inference,
nothing substantial is standing behind the name. Although the content of this method is very poor, yet another title
was given to it by Charles Peirce: "abduction." On the scale of novelty-generating methods, retroduction, or
abduction, is situated above induction, since it is supposed to generate not only new information but also novel
concepts.
Retroduction (RD) can be compared to the hypothetico-deductive (HD) method (see, for example, Hanson 1958,
52-3). According to the HD scheme, observational consequences are deductively derived from premises containing
a given hypothesis h and a known set of initial conditions i. If the predictions agree with the observations, the
hypothesis is said to be confirmed. The notion of confirmation is not a logical notion. Indeed, if from h and i one
deduces a
< previous page
page_66
next page >
page_67
< previous page
page_67
next page >

Page 67
statement which turns out to be true (or at least accepted as true), then logically, one can say nothing about the
truth value of h even if i is accepted as true. In practice, scientists accept theories as discoveries in case they yield
successful predictions, especially if these are far-reaching or unexpected predictions. By the notion of confirmation,
philosophers of science try to explicate this intuitive notion.
Since confirmation is an essential part of the process of discovery, HD can be viewed as a non-generational part of
the method of discovery. The RD method is supposed to complement the HD method, as the method of generating
the explanatory hypothesis. The HD method provides a confirmation to h according to the following inference
pattern:
HD:
(1) h predicts an unexpected phenomenon e (e is entailed by h&i).

(2) e agrees with observation.
(3) Therefore, h is highly confirmed.
We assume here that the statement i expressing the initial conditions is true or highly confirmed.
In a somewhat analogous fashion, we may represent RD according to the following pattern:
RD: (1)e is a surprising phenomenon.
(2)e would be explained as a matter of fact if h were an accepted theory (e would be entailed by
h&i).
(3)Therefore there is reason to accept h as an explanatory theory.
The above RD scheme is a paraphrase of Peirce's theory of retroduction. One difference between his scheme and
the above scheme is that the latter refers to an accepted theory rather than to a true theory.
In the HD scheme, h is given (i is accepted as true), whereas e is discovered as a result of testing the prediction.
(Note that the HD scheme also accommodates the case where e is already known but no one thought that h might
account for it). The discovery of e contributes to the acceptance of h itself as a discovery. The discovery of e is by
no means a straightforward matter. One may have to devise sophisticated experiments, the results of which may
have to be analyzed and interpreted. For example, the SU (3) symmetry of hadrons was confirmed following the
discovery of the W- particle. However, the latter discovery was an enterprise involving highly sophisticated
experiments, and the interpretation of the results required elaborate theoretical and statistical analysis.
In the RD scheme, e is known, whereas h is discovered as a result of trying to explain e. The scheme does not
instruct us how to generate h. Hence, it
< previous page
page_67
next page >
page_68
< previous page
page_68
next page >

Page 68
is not an inference which generates an explanatory theory. As in the HD method, h appears in the premises of the
inference pattern. Indeed, Peirce's RD theory was criticised for not being a method of discovery at all. Behind this
criticism lurked the attitude that a method of discovering a theory is supposed to generate the theory. RD enters
into the discovery game only after h is already available. At most, RD may help us in selecting a hypothesis among
several available candidates. Yet, the process of selection is no less important than the process of generation, as
Poincar would view it. Moreover, sometimes it is the main stage in the process of discovery. Indeed, the process
of discovery is in many cases a process of finding the theory among a limited number of available hypotheses. In
addition, RD is a method of confirming h and confirmation is part of the process of discovery. Both HD and RD
schemes can be viewed as inferences from observed phenomena to the confirmation or acceptance of a hypothesis.
Thus, they are ampliative inference schemes which are extensions of inductive inference. In fact, Mary Hesse
(1974, 98) refers to HD as a rule of inductive inference.
2.2.3 The Quest for Certainty or: How Ampliative Inference Can Be Converted into Deductive Inference
Inductive inference causes many difficulties for the philosopher. It can be justified only on pains of circularity. It is
not clear what a "valid," strong or good inductive argument is. We base our inference on the data collected until
now, but the future is open. So the question is how the conclusion of an inductive argument is supported by the
available evidence. All attempts to construct inductive logics failed. The tension between reason and noveltygeneration is evident here. If we want to arrest novelty-generation in an inference pattern, the price paid is that the
inference becomes somewhat awkward. If we want, at all cost, to represent discovery as inference, we must either
conclude that discovery is non-creative or maintain that discovery has inferential as well as creative dimensions.
The first option is counterintuitive. Thus, what is left for us to do is to isolate the inferential component of
discovery, pinpointing the places where the creative component might enter into the process.
If the discovery process carries us beyond what is already known or implied by what is known, we do not have a
deductive support for the discovery. However, we cannot start from secured premises anyway, since no knowledge
is absolutely warranted. Hence, even deductive inference would not secure truth; it will necessarily lead to truth
only if we start from true premises. It is, therefore, equally insecure to proceed via deductive or via inductive
inference. We must rely on some assumptions. We may assume that certain statements, such as observational
statements, are true, reliable, or warranted, and then proceed via deductive inference. We may further assume that
inductive inference is reliable and proceed also by inductive inference. Both
< previous page
page_68
next page >
page_69
< previous page
page_69
next page >

Page 69
ways are risky, whereas in the latter case we are taking more risk since we rely on one more assumption. However,
without taking risks we can make no epistemic profit. With the rise of modern science, the no-risk policy, or the
quest for certainty, was recommended by an influential philosopher such as Descartes. Rationality was equated to
cautiousness. However, Cartesian philosophy did not reflect the nature of modern science. The rise of modern
science is characterized by boldness and imagination. Modern philosophy of science only in recent decades has
freed itself from the formula of rationality equals no imagination, no creativity and no risk, a formula which was
carried to its extreme by the movement of logical positivism.
Scientific rationality consists perhaps of proper dosages of both risky and "responsible" behavior. Responsibility
implies that we start with conventional methods. The first step may, therefore, be to see what we can gain from
deductive inference. We may try to see whether inductive inference can be reduced to deductive inference. If we
are deductivists and we want to bridge the inferential gap in a seemingly inductive argument, we may adopt the
above mentioned ad hoc strategy, representing inductive inference as an incomplete deductive inference. The
bridging might be provided by a statement which would express the principle on which we base our inductive
inference. If we could formulate such a statement, we could include it as a premise in every inductive inference,
thus converting the inference into a deductive inference.
It is sometimes said that the principle of the uniformity of nature is the principle which underlies our inductive
inference. But how shall we express this principle in a statement? The statement "nature is uniform" is not very
informative. It is an a posteriori "rule" which describes the fact that science has discovered regularities or
uniformities in nature. It does not tell us in advance what are the uniformities which are projectible, i.e. which
correspond to natural laws. We can identify the phenomena and the properties in our field of observation in an
infinite number of ways. The regularities we find depend on the manner we make this identification. Thus, the
regularities we find depend on our point of view.
Instead of inserting a world-embracing principle among the premises, we might insert more modest principles. We
might start our argument, for example, from the premises stating that all the large number of observed objects
identified as P's have been found to have the property Q, with no exception. We will then deductively infer the
statement "all P are Q" if we added the following premise expressing the principle of induction by enumeration:
IE: "for any X and Y, if a large number of objects identified as X's are observed to have a property Y, and no
counter-example is observed, then all X are Y."
This is how we might make an inductive inference to appear as a deductive inference. The method is to take the
inductive inference rule and convert it into
< previous page
page_69
next page >
page_70
< previous page
page_70
next page >

Page 70
a premise. If we had satisfactory inductive rules of inference, then inserting these rules among the premises of
inductive inference would indeed amount to deductive inference. However, IE by itself is useless. It is effective as
a rule of discovery only in case we have a definite set of natural kinds by which we can categorize the phenomena.
But this is exactly where the creative component of discovery enters. In our ordinary experience, we categorize the
objects and phenomena in the world by using concepts which are available in our natural language. However, this
conceptual system is not appropriate for describing the phenomena investigated by modern science. Scientists,
therefore, have to invent new conceptual systems, referring to new natural kinds, such as the quantum mechanical,
field-theoretical or cognitive-theoretical conceptual systems. Discovering the new natural kinds is outside the
scope of inductive inference.
Alan Musgrave employs a similar strategy for converting a variety of ampliative arguments into deductive
arguments. He uses bridging principles which are not universal (such as IE) but domain-specific (Musgrave 1988).
His method is to find the suppressed assumptions in the discovery-arguments expounded by scientists. Indeed,
there are always common presuppositions and beliefs shared by the members of the relevant community which,
therefore, need not be explicitly stated in scientific discourse. Musgrave adopts this strategy in order to convince us
that there is a method of scientific discovery, applied to all kinds of discoveries, and it is no less than deductive
logic. Let us follow his examples.
The first example is from everyday experience. When we want to put forward a hypothesis about the color of
emeralds, we do not guess blindly and test our guesses one by one. This would be a very inefficient way to
proceed. We might guess that emeralds have no common color. We might hypothesize that in the winter they are
yellow and in the summer their upper part is blue and lower part is black. Popper tells us in his Logic of Scientific
Discovery that there is no logic of scientific discovery and recommends making bold conjectures. However, if there
is no method restricting our imagination, we would be wasting our time and might arrive at no discovery. In
practice we start with an assumption or a premise such as p1: "all emeralds have some common color." This
assumption is not world-embracing; it is rather domain-specific. In this manner, Musgrave avoids the above
mentioned problem of inventing the natural kinds; he considers a situation where the natural kinds are already
available. The premise p1 can be reduced to a deductive conclusion of the following more general domain-specific
premise p'1: "emeralds belong to a family of kinds of precious stones whose members have a common color." Of
course, there is no justification for p1 or p'1. But the point is that an argument starts with some assumptions and its
rationality (according to the traditional interpretation) resides in its validity rather than in the justification of the
premises. Another premise, p2, which is drawn from observation, says that some particular emeralds are green. We
thus have the following argument:
< previous page
page_70
next page >
page_71
< previous page
page_71
next page >

Page 71
p1: All emeralds have some common color.

p2: A particular emerald is green.
Therefore C: All emeralds are green.
Thus, the "inductive" argument whose premises are statements referring to observed green emeralds and whose
conclusion is C is in fact a deductive argument with a missing premise p1, for example.
The conclusion C does not constitute a novelty with respect to the premises p1 and p2. Consequently, C is certain
relative to the premises. The uncertain element, however, did not disappear; it was pushed to the premise p1. Thus,
we have here a method of discovery based on deductive logic. This method presupposes that the discoverer starts
with some working hypotheses which are plausible and established in his mind.
Not much novelty is generated by the above inference, since the novelty-generating premise p1 or p'1 is a very
common kind of assumption which proved successful in ordinary experience. Creativity will be needed when we
turn to an unfamiliar environment, occupied with unfamiliar objects and phenomena. Creativity is needed in order
to find what natural kinds are there, on which inductive generalizations, such as p1, can be made. No wonder that
inductive inference such as in the above example, which presupposes a stable set of natural kinds, can be "dressed
up" as deductive inference. This is a process of discovery by exposure. We have p1 in mind and then a discovery
of a single green emerald amounts to the discovery of the generalization "all emeralds are green."
Musgrave's second example is of generating a hypothesis about the relationship between two measurable quantities
L and M. We make the general hypothesis (derived from considerations such as Grosseteste's principle of the
simplicity of nature) that the relation is a linear one. Then we make two pairs of measurements and find out the
exact relation. The deductive argument might be, for example, the following:
q1: L and M are lineary related, i.e.L=a M + b for some real numbers a and b.
q2: When M=0 then L=3.
q3: When M=1 then L=5.
Therefore, L=2M + 3.
Here we discover a specific relation by making a general mathematical hypothesis. Again, this is not a noveltygenerating discovery, since the major working hypothesis is included in the premises.
Encouraged by his success in "dressing up" inductive arguments as deductive ones in cases where no novelty was
generated and no creativity was needed, he turns to another example which is a typical generational discovery.
< previous page
page_71
next page >
page_72
< previous page
page_72
next page >

Page 72
The example is Ernest Rutherford's discovery of the structure of the atom. Unlike the first two examples, this
discovery resulted in a great novelty. Since this example is instructive I will describe it in some more detail.
Rutherford Discovering the Emptiness of the Atom 5 In 1908 Hans Geiger used the scintillation method to measure
the scattering of alpha particles. When a thin sheet of metal was inserted between the slit which limited the beam
and a phosphorescent screen, scintillations corresponding to particles deflected from their straight paths were
observed. It was found that the number of particles deflected through a given angle decreased rapidly as the angle
increased. It was found also that the number of deflections increased with the thickness of the foil, and the heavier
was the deflecting atom, the greater was the deviating effect.
Later, young Ernest Marsden joined and, at the suggestion of Rutherford, searched for alpha particles scattered
through a large angle. Rutherford later recalled in a lecture he gave:
I may tell you in confidence that I did not believe that there would be, since we knew that the alpha particle
was a very fast massive particle, with a great deal of energy, and you could show that if the scattering was
due to the accumulated effect of a number of small scatterings the chance of an alpha particle being
scattered backwards was very small. Then I remember two or three days later Geiger coming to me in great
excitement and saying, 'We have been able to get some of the alpha particles coming backwards.'...It was
quite the most incredible event that has ever happened to me in my life. It was almost as incredible as if you
fired a 15-inch shell at a piece of tissue paper and it came back and hit you. (Ibid., 111)
Why had the results struck him as so strange? The model of billiard balls which worked so well in the kinetic
theory of gases led to the belief that atoms are behaving like solid particles. Rutherford expressed this belief as
follows: "I was brought up to look at the atom as nice hard fellow, red or gray in colour, according to taste" (ibid.,
115). However, the picture changed as a result of the experiments made with cathod rays. Philipp Lenard made a
very small hole in the side of the Crooks tube covered with aluminum foil thin enough to let through the electrons
(the cathod rays). He found that swift electrons passed through comparably thick foil. Calculations, which relied on
the known number of atoms in a given volume and the approximate size of the atom, showed that only if the
electrons passed freely through the body of atoms in the foil could the observed penetration be possible. This led
Lenard to conjecture that atoms were made of particles which he called dynamids. Each dynamid con-
< previous page
page_72
next page >
page_73
< previous page
page_73
next page >

Page 73
sisted of an electron closely associate with a unit of positive charge, being as a whole neutral. He showed that the
dynamids of solid platinum must occupy no more than 10-9 of its volume.
Later, J. J. Thomson suggested his "plum pudding" model of the atom. According to this model, the atom was a
sphere of positive electricity, with electrons imbedded in it, arranged in a series of concentric rings in one plane.
This structure corresponded roughly to the periodic chemical properties.
The work of Geiger and Marsden was completed in 1909. Only early in 1911 Rutherford found an explanation to
the results. He came to the conclusion that each of the large-angle deflections of the alpha particles must be due to
a single collision with a very small and very massive charged particle, the nucleus.
Musgrave's Reconstruction The whole process is encapsulated by Musgrave in the following argument:
A1: The same (similar) effects have the same (similar) causes.
A2: Atoms and the Solar System behave in the same "dense and diffuse" way with respect to bodies entering them
[i.e. most bodies entering them pass straight through them, but a few collide violently with them].
A3: The Solar System "dense and diffuse" behavior is explained by its structure, a relatively small but massive
body orbited by much lighter bodies.
Therefore C: Atoms are structurally similar to the Solar System...
Here Musgrave "dresses up" an argument by analogy, which is considered to be an inductive argument, as a
deductive argument. This is a typical example of how a process of discovery can be reconstructed without giving
us a clue about the method which could have guided the discoverer to his discovery before he made the discovery.
A2 is the crucial premise in the above so-called inventive argument. However, it is a premise which can be stated
only after the main step was made in the process of arriving at the hypothesis. The fact from which the discoverer
started here, and which he wanted to explain, was the "dense and diffuse" behavior of atoms with respect to bodies
entering them. So whence sprung the idea about the Solar System into the argument? Everything is similar to
everything in some respect. The problem is to find a fruitful similarity. The fruitful similarity between the atomic
structure and the structure of the Solar System was the creative step in Rutherford's discovery. The reconstructor
already knew this. Had we not already known about Rutherford's discovery, we would not know how to reproduce
it here, since we are not presented with any method to direct us in how to arrive at this particular model. There are
infinitely many systems in the physical world. There were many systems which had been successfully described
and explained by
< previous page
page_73
next page >
page_74
< previous page
page_74
next page >

Page 74
physical models and theories before Rutherford's discovery. The discoverer's task was to find the particular system
which supplied the successful explanation. Hitting upon the similarity between the atomic structure and the
planetary system was the creative act in the discovery and no method is given to us as to how to make such a
discovery. Probably there is no method for arriving at such a creative association. It took Rutherford more than a
year to free himself from his entrenched belief and to hit upon the idea.
<><><><><><><><><><><><>
In the above examples, the method is applied to a marginal step in the process of discovery. The general method
here is to convert the discovery to deductive inference where the premises contain certain hypotheses which bridge
the inferential gap. But the creative step in the discovery is generating these hypotheses. When these have been
discovered, the process indeed does not generate any novelty, it just exposes information implied by the premises.
2.2.4 The Hierarchy of Material Logics
Another ad hoc strategy referred to by Musgrave is to convert deductive inference with a suppressed premise into
an inference with "material" rules of inference. Here, the inference rules are domain-specific. For example, in the
inference: "gravitons are massless, therefore gravitons move with the velocity of light," the missing premise p is:
"massless particles move with the velocity of light.'' Since the last statement is common knowledge among
physicists, there is no need to mention it. In a community of experts there are many suppressed assumptions, some
of which are tacit. This is the reason why the novice in the field would not understand many discussions between
the experts. It is, therefore, tempting to categorize these suppressed premises as material rules of inference of the
if-then form. The rule of inference r corresponding to the above missing premise p would be: "From 'x is a
massless particle,' infer 'x moves with the velocity of light.'" Unlike deductive rules of inference, which are formal
or content-free, material rules of inference are content-dependent. They are, therefore, ampliative, or contentincreasing. However, these rules are not content-increasing relative to the background knowledge of the expert who
has internalized them. Thus, in the material logic of the physicists the inferential rule r is not ampliative relative to
the background knowledge of the community which uses this logic; relative to this background knowledge, if the
antecedent in the rule r is true, the consequent must be true.
Musgrave raises the following possible objection to material logic. Logic is not an empirical science so that the
material rules of inference, which seem to be loaded with empirical content, cannot constitute a logic. He indicates
< previous page
page_74
next page >
page_75
< previous page
page_75
next page >

Page 75
that the distinction between empirical and logical matters belongs to the heritage of logical positivism. However,
according to logical positivism, the "heuristic principles" on which the material rules of inference are based are not
factual statements; some of the heuristic principles are empirical generalizations which would hardly pass the
criteria of factuality. Heuristic principles which are theoretical statements definitely would not count by the logical
positivists as factual. Some heuristic principles are metaphysical principles, which are the antithesis of factual
statements. Since the heuristic principles are not falsifiable, they would also not be considered empirical according
to the Popperian tradition, which regards only falsifiable statements as empirical.
I propose to order domain-specific logics, or community-dependent logics, in a hierarchical order. The most
general logic is deductive logic. It is the logical theory underlying the inferential practice of the broadest human
community whose members can communicate with each other. We might view even this domain as domainspecific, although the domain is the widest domain of knowledge and experience shared by human beings. It is
specific to humankind. Perhaps the corresponding logic of intelligent creatures which evolved in another galaxy
would be different. Inductive logic is also shared by all human beings. However, a precondition for applying it is
the identification of the natural kinds on which inductive projections can be made. Some of the natural kinds will
be common to all human beings. In more specific domains of experience, one should learn to identify the natural
kinds. Hence, although the inference rules are not domain-specific, their application is domain-specific.
A community sharing a narrower domain of specific experience may develop an inferential practice whose
underlying rules of inference are specific to that domain. For example, every community of professionals or
experts may have such a logic. This includes the whole scientific community or narrower scientific communities,
such as the community of biologists or physicists or even of a narrower community such as that of particle
physicists. Those who do not belong to a particular logic community would have difficulties in communicating
with members of the community. A non-scientist will not understand scientific discourse. And this is not only
because of the lack of knowledge, but mainly because of the suppressed premises or the material rules of inference.
Communication is difficult even between different subdisciplines. We know, for example, that physicists who want
to contribute to molecular biology, have difficulties in communicating with the molecular biologists, although they
know everything they have to know in the field. The reason for this is that they were brought up in a different logic
community. This idea may remind us of the Kuhnian notion of incommensurability between different paradigms.
But the similarity is only partial since the "rule" is that people belonging to a narrower logic-community share all
the wider logics in the hierarchy. Thus, there is a one-way communication along the hierarchy, whereas no
communication is possible across paradigms.
< previous page
page_75
next page >
page_76
< previous page
page_76
next page >

Page 76
Every community of experts has tacit knowledge and implicit assumptions underlying their discourse. Some of the
suppressed heuristic principles are included in the tacit knowledge and are not listed anywhere. They are rather
acquired by the novice via the process of apprenticeship. We encounter the same situation in ordinary reasoning.
Deductive logic explicates the rules behind ordinary inferential practice. However, ordinary people who are
participating in this practice do not learn deductive logic. They acquire the tacit rules by an apprenticeship-like
process, when they learn their mother tongue and when they participate in ordinary discourse. Perhaps the ability to
reason correctly is genetically determined. However, the main point is that for the non-logician, the inference rules
are tacit. Thus, when we identify suppressed assumptions which are regularly employed in arguments made in a
given community, we have one option of treating them as missing premises, and another option of treating them as
representing material rules of inference.
Noretta Koertge makes the distinction between inference patterns which are legitimate in all possible worlds and
those which are heuristically successful in our world. She maintains that the first kind of inference patterns are
logical whereas the second kind of inferences are rational (Nickles 1980, 48). Presumably, logical implies rational
but not vice versa. According to the above view about the hierarchy of material logics, logical (read deductive)
inferences are valid in all the possible worlds which can be comprehended by us, whereas heuristically successful
inference patterns are rational in their specific domains. Thus, rationality in this sense is domain-specific.
Rationality in Koertge's sense is categorical, referring to "our world" without distinguishing different domains of
experience. According to my approach, rationality is domain-specific. In any given domain, it is manifested by the
material logic of that domain, whereas categorical rationality is manifested by obeying the rules of the material
logic governing the discourse in the widest domain of human experience, i.e. deductive logic.
In Chapter 4, I will develop an approach according to which the justification of the rules of deductive logic resides
in part in their being an explication of the rules underlying our inferential practice. Furthermore, logical theory will
be treated there as a theory in natural science. In this sense, even deductive logic is empirical, as every theory in
natural science is. This approach might, therefore, be suitable for treating the hierarchy of material logics, which
correspond to the inferential practice of the different communities.
If we identify the material logic underlying the discourse of a specific scientific community, we might understand
their reasoning patterns. We would be, therefore, able to decide whether and how they reason in arriving at their
hypotheses. However, we have to be cautious in distinguishing between generative and justificatory reasoning.
< previous page
page_76
next page >
page_77
< previous page
page_77
next page >

Page 77
2.2.5 The Discovery Machine

There are discovery processes which are carried out effectively by scientists and which can be reconstructed or
simulated in simple cases by a real machine. On the other hand, there are discoveries which are more effectively
carried out by a computer rather than by a human agent, e.g. recursive computational processes of discovery.
Machines can create novelty; new statements, information and formulas can be generated at random or by a
heuristic-guided process. We might simulate, for example, natural selection by generating quasi-random
hypotheses, some of which will be selected according to predesigned criteria. The selected hypothesis will be a
product of a creative act. For example, the machine may be fed by data and the task will be to find a mathematical
function which will be best fitted to the data. The programmer may design a hypotheses-generator which will
generate mathematical functions quasi-randomially. The first function which will be "caught up" as providing a fit
which falls under the predetermined range of variation tolerance will be the product of the process. If, as
evolutionary epistemology postulates, science simulates the process of natural selection, there is no reason for
excluding the possibility of simulating scientific discovery which transcends an established paradigm by a machine
which simulates natural selection.
In general, we can equip the machine with criteria and procedures for deciding whether a novel product is useful
or achieves a predesigned goal, such as explaining some phenomena or solving a certain problem. We can even
equip the machine with a repertoire of problems which the machine will scan, with different selection procedures,
so that it can decide what product solves what problem. In this manner, the machine may even generate solutions
for unexpected problems. We may also devise the program in such a way that the criteria of selection will change
following the ongoing experience of the machine. Of course, to translate this possibility "in principle" to a working
system which makes judgments and which learns from experience is achievable, for the time being only for a very
restricted range of problems.
Imagine a discovery machine programmed by a programmer who knows everything known to physicists just
before Planck solved the problem of blackbody radiation. We can supply the machine with the major ideas and
methods employed by physicists during the last half of the nineteenth century. One of these ideas, was the idea of
calculating certain integrals in thermodynamics by equating the energy with integer multiples of a certain fixed
quantity, and then calculating the limit where this quantity goes to zero. This was a computational device
employed by Boltzmann. After despairing from solving the problem in other ways, Planck used this trick without
going to the limit (see Chapter 5). How can we instruct the machine to find this idea or method? The programmer
should be wise enough to include Boltzmann's trick in the repertoire of ideas
< previous page
page_77
next page >
page_78
< previous page
page_78
next page >

Page 78
supplied to the machine, and then the machine should hit upon the idea and generate the required modification of
the idea. The probability for this is very small. The probability will increase only if the programmer will narrow the
search field. But in order to do this in the right direction, the programmer should already know how the discovery
was made by Planck.
At the present, the computer lacks an important dimension of the discovery process: the social dimension. We can
envisage a network of intercommunicating computers replicating the scientific community, including its
institutions, but for the time being, this belongs to the realm of science fiction. When this "vision" is fulfilled, we
will not be far from replacing human society by a computer society. The social dynamics of the scientific
community includes the judgments made by each scientist on whom he can rely or what weight can be given to
each of his colleagues. In such a process, rhetoric, personal influence and other faculties which (for the time being)
cannot be mechanized play major roles. In view of the cooperative or historical dimension of discovery, the
machine cannot be as creative as science is. Until the social dynamics or a historical process of discovery can be
simulated by a machine, a major dimension of creativity will be missing in machine discovery. No hardware
supplemented by any amount of software can replace the whole scientific community. All major discussions of
machine discovery have ignored this aspect of discovery.
Yet, although the discovery machine cannot simulate the whole process of discovery, it might aid scientists in their
decisions and acts at particular stages of the process. A scientist is aided by calculators and computers in his
calculations and data processing. The discovery machine might be another aid to the scientist in case a decision
which involves taking into account too many factors should be made.
Let us follow a specific example, presented by Herbert Simon (1987). Simon describes a computer program,
BACON, developed in collaboration with Pat Langley, Gary Bradshaw and Jan Zytkow. Let us see how this
program discovers Kepler's Third Law of planetary motion. BACON is supplied with data on the periods of
revolution (P) and the distances (D) of the planets from the sun. It is applied to the data according to the following
recursive heuristic rule: REC: "If two variables co-vary, introduce their ratio as a new variable; if they vary
inversely, introduce their product as a new variable and test it for constancy." With this rule, BACON first notices
that P and D co-vary. It thus computes P/D, which is found not to be invariant. Then REC is applied recursively to
the new variables P/D and D, which are found to co-vary. Their ratio P/D2 is found not to be invariant. Then
BACON finds that P/D2 vary inversely with P/D, so it multiplies them, obtaining P2D3, which is found to be
constant. The constancy of this variable is indeed an expression of Kepler's Third Law.
In this example, the discovery machine is doing only part of the job. The first important step is choosing the
variables P and D. The choice of the "right"
< previous page
page_78
next page >
page_79
< previous page
page_79
next page >

Page 79
variables sometimes constitutes the main step in the discovery, after which the regularity is immediately exposed.
In this particular case the programmer and Kepler alike did not have many available alternatives to choose from; P
and D were inherited from the prevailing scientific tradition of circular planetary motion. This situation where the
variables are available and no new ones are introduced is typical for a discovery which does not involve the
construction of a new explanatory theory in which novel theoretical terms are introduced.
Simon claims that the new variables constructed from the original ones are theoretical terms. However, neither P/D
nor P/D2 can be treated as genuine theoretical terms. The reason for this is that a theoretical term should appear as
part of a unifying or an explanatory theory. Both variables do not have any role in any theory; they are formed just
as steps in the computation. They do not refer to any physical phenomenon or to a significant physical magnitude.
They do not appear in any law of nature. In the process of developing a theory, many expressions are obtained
along the way. We would not call all these expressions "theoretical terms."
Yet, Simon mentions another variable which is created in the process and which corresponds to a new theoretical
concept. By using the word concept, he presumably means that this is a theoretical term having a physical
significance. The new concept, which he calls "gravitational mass," is created in the following way. In a given
planetary system, the magnitude P2D3 has a constant value K. If BACON is applied to different planetary systems,
such as the satellites of Neptune, different values of K will be obtained. In this way, the concept of gravitational
mass will be discovered, since in Newton's theory of universal gravitation, K is proportional to the gravitational
mass of the central body in the planetary system. This seems to be a creative discovery since a new physical
magnitude appears to have been discovered here. However, this is only an apparent discover. If a machine or a
playing child who are supplied with two physical magnitudes such as P and D, would form from them a new
combination which turns out to play a role in a theory such as Newton's mechanics, it by no means means that the
child or the machine discovered the new concept. Had BACON discovered a theory or a law, in which
gravitational mass plays a significant role, could we say it discovered the new concept. The concept of gravitational
mass has more content in it than just being related to a certain combination of P and D. BACON plays the role of a
"Kepler machine" but not of a ''Newton machine."
The process carried out by BACON is not an inference. Indeed, the recursive heuristic rule (REC) programmed in
BACON is not equivalent to a rule or a set of rules of deductive or inductive inference. Nevertheless, REC guides
a discovery process which is not generational but is a process of exposure; it exposes a regularity hidden in the
data. If a mechanical procedure generates discoveries in a data-driven process, it means that the heuristic rules are
good ones. Thus, an important step of the discovery is the discovery of the
< previous page
page_79
next page >
page_80
< previous page
page_80
next page >

Page 80
heuristic rules. BACON is by itself a product of a creative discovery or an invention. It is a machine which when
fed with the right data, discovers regularities hidden in it. We can make the analogy with an observational
instrument. For example, after the telescope was invented or discovered, Jupiter's moons were discovered by
exposure. The telescope magnifies our sensual capabilities whereas the heuristic-instructed machine amplifies our
capability of discovering regularities. A successful heuristic rule is, therefore, an instrument for the discovery of
regularities, just as the telescope is an instrument for observational discovery. Thus, in order to make significant
discoveries by exposure, we sometimes have to discover first an appropriate instrument; the discovery of the
exposing instrument is a generational discovery or an invention. If we look at the whole process of discovery in
these cases, it is not a pure discovery by exposure but a generational discovery which leads to discovery by
exposure.
BACON is thus an example of how the machine can magnify our discovery capabilities. Thus, the computer may
help the human discoverer in the case of recursive procedure. The computer also magnifies our computational
capabilities and data-processing capabilities. These are examples where the computer is an important device for the
process of discovery. However, from this we cannot draw the sweeping conclusion that (in all, or even in most,
cases) "discovery can be mechanized."
2.2.6 Theory-Construction and Research Programs
A typical pattern of theory-generation in a framework preserving setting is when the theory is constructed
according to a general model, or a general heuristic principle, available in the field. For example, in Ptolemaic
astronomy, there was only one basic explanatory model, the epicyclic model. Planetary motion was described in
terms of a given configuration of epicycles. When a planet was found to deviate from its prescribed circular
motion, the heuristic rule would tell the astronomer to solve the problem by adding new epicycles in order to
comply with the principle of circular motion. Yet, the heuristic rule did not specify how to do this.
Another example of heuristic-guided theory-construction is the evolution of theories of the structure of matter
since the nineteenth century which seems to have been guided by the Meyersonian heuristic principle. According to
this principle, theories of material changes, e.g. theories of chemistry, have to be constructed out of conserved
"substances," or "material causes," such as atoms and fundamental particles, and conservation laws; the general
"recipe'' is that whenever some new phenomena do not obey the conservation laws, science has to look for new
conserved substances and new conservation laws (Meyerson 1908). For example, Dalton's atomic theory, which
appeared at the beginning of the nineteenth century and which was completed by Avogadro,
< previous page
page_80
next page >
page_81
< previous page
page_81
next page >

Page 81
yielded a set of conserved substances, the atoms of the chemical elements, and a set of corresponding material
conservation laws which stated that the number of atoms of each element should remain constant throughout
material changes. In 1896 Becquerel and Curie discovered the phenomenon of radioactivity which showed that
chemical elements are not conserved. Eventually, this phenomenon was explained with the emergence of new
conserved substances, the nucleons and the electrons, and with the corresponding new conservation laws stating
that the number of nucleons and the number of electrons are conserved. These conservation laws were found to be
violated and the developments in the physics of particles and fields also followed the Meyersonian recipe (see
Ne'eman 1983). We have here an example of a heuristic rule which is domain-specific and content-free. It is
confined to a specific domaintheories of material change. However, it does not specify the content of the conserved
substances and the material conservation laws. It is the task of the discoverer to invent them.
The above discovery pattern may be represented as a "semi-inference"; i.e. an inference, in which the inference
rules are replaced by the heuristic rules. The premises of such an argument include the new anomalous data.
However, there may be more than one "conclusion" to the argument. The heuristic rules do not uniquely determine
the product of discovery. For example, the Meyersonian heuristic would leave the discoverer with more than one
possible explanatory hypothesis.
In general, in a framework-preserving discovery, the discoverer will have at his disposal a very limited number of
basic explanatory ideas or models; the prevailing world picture and background knowledge would narrow the
number of possible or plausible explanatory models. Thus, in some cases, whenever the discoverer encounters
problems awaiting a solution, or phenomena awaiting explanation, he might not be required to construct a new
theory afresh, but only to adjust the general explanatory model available in the field to the explanandum. When
there are several explanatory models in the field, it sometimes becomes immediately evident what model is the
appropriate one, or it is a matter of simple inference to determine which model will explain the phenomena; several
observational results plus the background knowledge uniquely determine the appropriate model among the small
number of models available. The process of discovery is then reduced to "induction by elimination": When we have
n hypotheses-candidates, n-1 of which are eliminated by observational or experimental results, the remaining nth
hypothesis is declared a discovery. This is, in fact, a deductive inference conditional on the assumption that there
are only n possible hypotheses capable of explaining the explanandum. If we denote the hypotheses by hi, the
background knowledge by B, and the conjunction of falsifying observational results by e, we can represent the
discovery argument by the following deductive pattern:
< previous page
page_81
next page >
page_82
< previous page
page_82
next page >

Page 82
p1: B entails (h1vh2v...vhn-1vhn).

p2: e entails (~h1&~h2 &...&~hn-1).
p3: e&B.
Therefore hn.
Due to premise p1, this is a proper inference and the product of discovery is unique. In the special case where n=2,
or after n-2 hypotheses have been eliminated, e is a result of a so-called crucial experiment which serves to decide
between the two competing hypotheses.
Next, let us consider the very common situation when we are already equipped with a theory T and a new piece of
data e brings about a new theory T' which is a result of adjusting T to the new data. T', the product of discovery,
may be regarded as a modified version of T. The process can be symbolized according to the following formula:
The heuristic HEU guides us in modifying the theory. Again, the product T' is not uniquely determined. This
discovery pattern can be viewed as dynamic theory-construction, where we start with an initial hypothesis and
modify it in order to adjust it to the data. In the field of AI, such procedures are carried out for discovering
regularities and laws. In a typical experiment, a robot arm mixes chemical substances according to some initial
hypothesis. Following the results obtained during the night, the hypothesis is changed in the following morning
according to certain heuristic rules (Buchanan 1982). This is another example of a recursive procedure.
The RP pattern is typically a data-driven process; it is the new data which brings about the need for modifying T.
The data-driven process of dynamically constructing a theory is very close to the notion of research program, or to
the notion of a dynamic theory (see Kantorovich 1979). The latter notion refers to a theory which is expanded and
modified from time to time in order to adjust it to new observational data. (This notion of dynamic theory should
not be confused with the ordinary notion referring to a theory dealing with dynamic phenomena; here it is meant
that the theory itself, rather than its subject matter, is dynamic.) In the process of adjustment and expansion, the
theory retains its name and identity and its central claims. Within the Popperian tradition the Lakatosian notion of
research program refers more or less to the same thing. A Lakatosian research program is a methodological entity;
it replaces the falsifiable theory as the fundamental unit for appraisal in science. A research program is a succession
of falsifiable theories which is characterized by a theoretical hard core, which remains by convention unfalsifiable
and unchangeable, by the methods of research and by the resources of theory building. The latter are referred to as
the "positive heuristic" of the program.
< previous page
page_82
next page >
page_83
< previous page
page_83
next page >

Page 83
We may combine the notions of dynamic theory and research program by saying that a dynamic theory is
developed in the framework of a research program. The process of constructing and modifying the theory can be
progressive or degenerative, according to the amount of successful predictions it yields. Thus, the process of
generation and evaluation goes hand in hand.
The difference between the notion of research program and the notion of dynamic theory is that the latter does not
necessarily contain a predetermined hard core. Lakatos assigns to the hard core those parts of the theory under
evaluation which are protected from falsification whereas the so-called protective belt, consisting of those parts of
the theory which undergo changes, includes the auxiliary assumptions and other "softer" components constructed
around the hard core. The fact that certain elements of the dynamic theory remain unchanged throughout the
process might testify for their adequacy, but it does not necessarily mean that an explicit decision was made at the
start of the process to protect them from refutation (see also McMullin 1976).
Indeed, a dynamic theory is not identified by a hard core. In the case when a model stands behind the developing
theory, it is the basic idea or the basic picture underlying the model which characterizes the theory. This idea or
picture is not fully expressed by statements. Hence, the dynamic theory is not a closed system. Rather it has open
points. As we will see below, the so-called neutral analogy of the model supplies such open points. The basic idea
or the basic picture regulates the development of the theory. It replaces both the Lakatosian hard core and the
positive heuristic. The model supplies the heuristic which guides scientists in developing the theory.
A dynamic theory cannot be falsified in a purely logical sense, but it can lose its viability when repeated ad hoc
attempts to protect it from clashes with observational data result in a degeneration of the research program. A
research program degenerates if it does not produce new predictions or if none of its prediction is successful. Since
the Lakatosian methodology is an outgrowth of the falsificationist tradition, it treats a research program as a
succession of falsifiable theories, each of which is falsified in its turn and replaced by its successor, while the hard
core remains untouched. In the notion of a dynamic theory, I depart from the falsificationalist tradition and talk
about a theory as a dynamic entity that retains its identity while being realized by successive theory-versions; the
identity is established by the basic picture or ideas behind the theory.
A dynamic theory or a research program evolves within the framework of normal science or a paradigm. It is
developed within a certain conceptual system or world picture and relies on some background knowledge. Within
this general framework, the research program intends to solve specific problems. This is done by introducing and
evaluating a specific model or a set of ideas. Examples are the Bohr-Rutherford planetary model of the atom or the
billiard ball model of the kinetic theory of gases. Both the general world picture
< previous page
page_83
next page >
page_84
< previous page
page_84
next page >

Page 84
and the specific model contain tacit elements, intuitive ideas and metaphors, which are being verbalized, cleared
up, explicated and formalized throughout the development of the research program. As Max Black put it, "every
science must start with metaphor and ends with algebra; and perhaps without the metaphor there would never have
been any algebra" (Black 1962, 242).
The process of "explication" or "clearing up" involves a continuous interaction of the theory with observational
data. One may say that the basic ideas and metaphor from which the dynamic theory starts to develop gradually
materialize into an explicit theory only through the interaction with observational data. The research program starts
with the discovery of a basic model or idea which is initially formulated as the first version of the theory.
Sometimes this first version is referred to as a "first approximation," or an "ideal" or ''naive" version of the theory,
such as the first version of Bohr's model of the hydrogen atom, which referred to circular electronic orbits. The
process continues to evolvethe theory is continuously constructed and reconstructedwhere further observational
information keeps it going, producing more and more advanced theory-versions which accommodate the flow of
information. On the other hand, the process of testing the new versions of the theory generates more data. The
development of a dynamic theory is thus brought about by the interplay between the basic ideas or the model and
the observational information. This is a process where a metaphor or an idea "ends with algebra" or with an
explicit theory.
Let us examine what happens in the prototypical case when the development of a dynamic theory is guided by a
model. The first theory-version which is generated by the research program may be seen as a simplified or a crude
version of the model, where an analogy is made between elements of the object system and some elements of the
model, whereas other elements of the model are considered to have no analogy with elements of the system; these
cases are "positive" and "negative" analogies of the model, respectively, to use Mary Hesse's terminology (Hesse
1966). Some elements of the model remain unutilized explicitly. These elements constitute the "neutral analogy,"
which serves as a guide for modifying and further developing the theory. For example, Bohr's model described the
hydrogen atom as consisting of a central positively charged nucleus encircled by a negatively charged electron
which can occupy descrete energy-levels, each of which corresponds to a certain radius of revolution. Energy
could be absorbed or emitted by the atom in the form of photons which carry energies equal to the difference
between energy levels corresponding to electron jumps between different orbits. Thus, the central motion and the
classical laws governing this motion belonged to the positive analogy of the model. The classical laws of
electromagnetic radiation and features of the Solar System, such as the heat radiated by the sun, belonged to the
negative analogy. In hindsight, we can say that the rotation of a planet around its axis belonged to the neutral
analogy, which later was utilized to invent spin.
< previous page
page_84
next page >
page_85
< previous page
page_85
next page >

Page 85
This simple model explained some basic features of the hydrogen's spectrum. The naive theory had to be adjusted
to further data, such as the fine and the hyperfine structure of the spectrum. Some of the new features of the
spectrum were explained by new versions of the theory. In one of the new versions, the notion of spin was
introduced. The heuristic principle which led to the suggestion of this hypothesis is the above mentioned neutral
analogy of planetary rotation around the axis, which, as a result, was converted from neutral analogy to positive
analogy.
Thus, with the guidance of the model, the new data generate a succession of theory-versions, where at each step,
more of the neutral analogy turns into positive or negative analogy. The neutral analogy serves as a source and as a
guide for developing the theory, and, therefore, it may be regarded as a heuristic which is derived from the original
hypothesis. In Bohr's model, we see also that an initial negative analogy may become a positive analogy, as
happened to the circular orbits which became elliptic in the developmental stage of the model called the BohrSommerfeld model. Due to the presence of the neutral analogy, the original hypothesis, i.e. the model minus the
initial negative analogy, cannot be described as an explicit statement. Only additional data and the imagination of
the scientist will determine the fate of the neutral analogy in the subsequent development of theory. The scientist's
imagination is inspired by the original hypothesis but not always in an unequivocal and explicit way. The various
associations and metaphors which are evoked by the model cannot be analyzed by the logicist's tools in the way a
well defined set of propositions can, e.g. by deducing testable predictions from it at the outset. The heuristic,
derived from the neutral analogy plus relevant parts of the background knowledge, leaves more than one possibility
for modifying the theory. Thus, even if we have a relatively well defined model, we have to decide at each stage
what to do with the neutral analogy, since the data do not dictate a unique way to proceed. Although we may
symbolize the growth of a dynamic theory by formula RP, it is not an algorithmic method of generating discovery;
it only describes a pattern of discovery which leaves room for creativity and for unpredictable developments.
2.2.7 The Calculus of Plausibility: Logic of Pursuit
A logic of pursuit determines whether an already given hypothesis is plausible in view of our background
knowledge and beliefs, and thus, whether it is worth pursuing and testing. Norwood Russel Hanson is renown for
proposing such a logic or method. He gives the example of Leverrier who hypothesized the existence of a planet
Vulcan in order to explain the deviation of Mercury's orbit from a perfect ellipse. Hanson reconstructs the reasons
which led Leverrier to his suggestion as reasoning by analogy. The irregularities in Mercury's orbit are similar to
those in Uranus' orbit. The latter were successfully explained by
< previous page
page_85
next page >
page_86
< previous page
page_86
next page >

Page 86
hypothesizing the existence of Neptune. Therefore, it is reasonable to conjecture the existence of a new planet,
Vulcan. This argument can either be represented as an inductive argument (by analogy), or as a deductive
argument (as Musgrave represents Rutherford's discovery). Again, as in Rutherford's example, to present this
argument as an hypothesis-generating argument would be a mistake, since, the argument is brought in after the
analogy is already given. The argument does not show how Leverrier arrived at the analogy. However, Hanson
does not claim that this argument is generational. He presents it as the reasons for suggesting the hypothesis. These
are not the reasons by which the hypothesis was generated; rather they are the reasons for choosing a particular
hypothesis out of the hypotheses available to the investigator: "...many hypotheses flash through the investigator's
mind only to be rejected on sight. Some are proposed for serious considerations, however, and with good reasons"
(Hanson 1958, 85). This reminds us again of Poincar's claim about the subconscious process of generation and
selection. Yet, in order that the process of selection will have chances to hit upon a successful hypothesis, Hanson
would have to adopt Poincar's claim that the candidates which consciously appear in the discoverer's mind are
already products of subconscious selection. However, it would be more plausible to assume that in the historical
context in which Leverrier made his suggestion, perhaps this was the only hypothesis which came to his mind,
following the success of the Neptune hypothesis.
Reason enters into the game after the hypothesis is created, whereas the act of creation is "a matter of psychology."
This does not diminish the value of reasoning if we adopt Poincar's claim that the generation of ideas or
hypotheses is valueless; selection or discernment is the crucial factor in the process. Thus, the reasons for
suggesting a hypothesis are reasons for selecting or choosing the hypothesis from a set of already generated
hypotheses. Indeed, in many cases the discoverer describes the act of discovery as an act of finding the hypothesis.
Hanson says: "Our concern has been not with giving physical explanations but with finding them" (ibid., 72). The
expression "I found a solution" is sometimes used synonymously with ''I generated a solution." However, the
literal meaning, which refers to searching and selecting among preexisting candidates, may reflect the true nature
of the process. This brings us back to the notion of discovering X as literally finding, or exposing, X. In view of
what has been said in the last section, in cases when all candidate hypotheses are at hand, the logic of pursuit is a
logic of selection and it is identical with the logic of arriving at the discovery. Thus, in these cases the distance
between exposure and generation is not so big.
In cases when the investigator has one successful model by which he successfully solved a previous problem so
that when posed with a similar problem he immediately turns to this model, there is no need for selection or choice.
Thus, Kepler conjectured that Jupiter's orbit is noncircular, following his discovery that Mars' orbit is noncircular
and the similarity between the
< previous page
page_86
next page >
page_87
< previous page
page_87
next page >

Page 87
two cases. In other cases, the reasons for suggesting a hypothesis may be pragmatic: e.g. the availability of the
model, the familiarity with the model, the ease by which one can calculate with the model, etc. In Chapter 8, when
I discuss the role of tinkering in the evolution of science, a new light will be shed on the significance of these
kinds of pragmatic considerations.
Note that the above examples refer to reasons for suggesting a certain type or kind of hypothesis (rather than a
particular hypothesis), e.g. a hypothesis referring to the existence of a new planet, a noncircular planetary orbit or a
planetary system kind of model. A particular hypothesis is generated by applying and adjusting the general kind of
model to the particular problem in question. Thus, the logic of pursuit is a logic for pursuing kinds of hypotheses.
However, it was suggested by Hanson that plausibility depends not only on the content of the hypothesis, but also
on the credentials of the person proposing it or on his rhetorical powers. In these cases, the logic of pursuit is not
directed only towards the kind of the proposed hypothesis. Only if we treat the origin of the hypothesis or the
manner by which it was produced as characterizing a kind may we still treat the above case as a kind of
hypothesis, e.g. the hypotheses proposed by X.
Hanson does not present us with a detailed logic of pursuit. Rather he gives us a programmatic picture. The reasons
for suggesting a hypothesis or a kind of hypothesis are mainly reasons incurred by analogy or by considerations of
simplicity. In fact, all these reasons amount to saying that the suggested hypothesis provides a potential explanation
for the explanandum. Thus, if the hypothesis is confirmed, it would be accepted as an explanation. We have to
remember that there is an unlimited number of hypotheses which can (deductively or statistically) entail a finite
body of data, while another necessary condition for an hypothesis to become an explanation, and, thus, a discovery,
is that it comply with the criteria of explanation. Hence, the good reasons for suggesting a hypothesis are those
reasons which indicate that the hypothesis complies with the criteria of explanation and with the already available
data. Some of these criteria, such as fruitfulness, predictability and simplicity, are content-neutral. Other reasons
are content-dependent, for example, compliance with current established explanatory models and theories, or with
the scientific world picture.
Hanson has been accused, together with Peirce, for confusing the process of generating a hypothesis with the act of
the initial evaluation of an already given hypothesis. However, as we have seen, in cases where there are a limited
number of ideas in the idea pool of the scientific community, the reasons which lead scientists to a hypothesis are
not generational; rather they are the reasons for selecting it from the pool, or finding it as a good candidate for
explanation. These reasons also provide the initial evaluation of the hypothesis. Thus, Hanson's account of
discovery is appropriate for describing a wide range of cases, where the hypothesis is drawn from a given idea
pool. In other
< previous page
page_87
next page >
page_88
< previous page
page_88
next page >

Page 88
cases, where an entirely new and successful idea is created in the mind of the discoverer, Hanson's and Peirce's
schemes still provide a partial account of the process of discovery. The creation of a new idea might be explained
by Poincar's hypothesis; a new idea is selected among many combinations of ideas which are formed in the
discoverer's mind. Again, according to this view, selection is the most important part of the process and this can be
described by the logic of pursuit.
The task of the logic or the method of pursuit is, therefore, to narrow the range of candidates. Without such a
method, we have an infinite number of hypotheses from which a given set of data can be deduced (given the initial
conditions). As I have stated above, not every statement from which the explanandum can be deduced, is
considered to be an explanation. Thus, I adopt here the following conception of theoretical explanation: A theory T
explains the explanandum E only if the following two conditions are met: (1) E can be deduced from T, (possibly
in conjunction with some initial conditions or auxiliary assumptions), (2) T complies with the prevailing world
picture, including the explanatory models which have proved successful, and with the criteria of explanation,
including the requirement of predictability. Thus, even if the trajectory of a planet can be derived from a certain
configuration of epicycles, this will not provide the contemporary astronomer with an explanation of why the
planet's trajectory is what it is, whereas it will be considered an explanation by the Ptolemaic astronomer, since it
complies with his paradigm of explanation. Thus, the logic of pursuit accounts for condition (2). The confirmation
of the hypothesis, according to this conception, depends on both pre-testing considerations, encapsulated in
condition (2), and on a posteriori considerations of evidential support or successful testing.
The logic of pursuit purports to determine the degree of plausibility of a hypothesis before it is submitted for tests.
The degree of plausibility of a hypothesis is determined, among other things, by the explanatory paradigm, by the
prevailing world picture and by the previous successes of hypotheses of this kind. I will describe now such a logic,
based on a Bayesian model of probability. The model is very instructive for our purpose, since it treats plausibility
considerations in a manner which is modeled on logical inference. Hence, we can draw sharp conclusions from the
model. This is a heuristic model. I do not claim that scientists actually calculate probabilities before they accept a
theory for pursuit. The status of the model in relation to scientific practice is similar to the status of formal logic
with respect to ordinary reasoning, or ordinary inferential practice. In ordinary discourse, people do not formalize
their arguments and they do not draw their conclusions by explicitly following the rules of propositional or
predicate calculus. However, formal logic provides a good explication of our intuitive inferential practice. The
process of explication provides us with a formal system, based on a small number of axioms, which reconstruct
our intuitions in a clear and concise manner. If the explication is
< previous page
page_88
next page >
page_89
< previous page
page_89
next page >

Page 89
fruitful, we will reconstruct with it most of our intuitions, we will find out whether they are consistent and we may
predict some new rules of inference which might turn out to be insightful and useful. The rules which are
explicated by a simple and fruitful formal system are not justified in the traditional sense which seeks a priori
justification. However, the formal system summarizes our intuitions in a concise manner and lends them formal
strength. In Chapter 4, I shall discuss the concept of explication in some more depth.
The Bayesian model can be employed for explicating the notions of plausibility, evidential support and
confirmation. According to this model, the probability of a proposition is interpreted as a subjective probability
which measures one's degree of belief in the proposition. The probability p(h,e) of a hypothesis h, conditional on
the evidence e, is a measure of the evidential support e lends to h, or the degree of confirmation of h, given that e
is the total evidence relevant to h which has been accumulated until now. The degree of confirmation, which is
changing with the accumulated data e, will determine whether h will be accepted as a discovery. According to
Bayes' formula, this probability is a function of p(h), the prior probability of h, i.e. the probability of h before the
evidence was brought into consideration, and p(e), the probability, or the degree of expectedness, of e before the
evidence was observed. The prior probability determines whether the scientific community will accept h for pursuit.
The condition that p(h) will be above some minimal value thus corresponds to the above mentioned condition (2)
for explanation. In the following, I will describe the model in some detail.
The fundamental entity in the model is the probability p(a) of a proposition a, given the background knowledge or
information, whereas the conditional probability of a, given b, is defined for any pair of propositions a and b as
If the initial or prior probability of a hypothesis h at time t0 is p0 (h) and a new empirical evidence e is acquired at
time t1, such that p1 (e)=1, then according to the so-called dynamic assumption of the Bayesian model (see, for
example, Hacking 1967, 314) the posterior probability, or the degree of confirmation, of h after e was observed is
The assumption is that when the only reason for changing the degree of belief from p0 to p1 is the observation or
the acceptance of e, then a rational change should be represented by a conditionalization upon e. In other words,
only the new evidence should affect our belief-change. This reflects an empiricist attitude. Bayes' formula, which
can easily be derived from eq. (1), relates the prior and the posterior probabilities of h. (From now on, the time
subscripts will be omitted from the probability functions at t0):
< previous page
page_89
next page >
page_90
< previous page
page_90
next page >

Page 90
If we consider an observational statement e which is entailed by h, then p(e,h)=1 and we get the simpler formula:
Eq. (4) applies to the case when e is predicted by h (given the background knowledge at t0 which includes the
initial conditions, etc.).
According to this model, we should be able to calculate in advance at t0 what our degree of belief in h will be,
when we accept e as true as a result of observation. Theories of probability such as that of Keynes (1921) or
Carnap in his earlier views (1950), attribute unique conditional probability p(h,e) to a given pair of statements h
and e. These approaches do not leave room for any arbitrariness in the inductive inference from e to h which may
be caused by extra-evidential factors, such as metaphysical beliefs or aesthetic and simplicity criteria. In the
Bayesian model, the dependence on these factors enters through the subjective prior probabilities, whereas the
dependence on empirical evidence enters through the dynamic assumption. The prior degrees of belief may vary
from one person to another but a rational person is ready to change his beliefs in accordance with the dynamic
assumption. It is claimed by some Bayesians that under these conditions, if a hypothesis has non-zero prior
probabilities, its posterior probabilities converge to 1 or 0 with an increasing amount of empirical evidence
(Edwards, Lindman and Savage 1963).
The relations of entailment and contradiction correspond to special values of the quantity p(h,e). If and only if h
can be deduced from e, it is evident that p(h&e)=p(e) and consequently, as can be seen from eq. (2), p(h,e)=1.
Indeed, if and only if h can be deduced from e, the truth of h is conditional upon the truth of e. Hence, if and only
if one believes in the truth of e, one should believe in the truth of h, i.e. he should give it probability 1. When h
contradicts e, then p(h,e)=0.
Now we can see how Bayes' model provides an explication of our intuitive notions of plausibility, confirmation or
evidential support. When a hypothesis is proposed, its prior probability corresponds to its degree of plausibility. A
hypothesis h is plausible or implausible only if p(h)>1/2 or p(h)<1/2, respectively. There are factors determining
the value of p(h) which are common to all or most members of the scientific community. The prior probability
would be high only if the following conditions were met: (1) h conforms to the general beliefs and world picture of
the scientific community, (2) h accords with the accepted standards of simplicity or aesthetic value. Another
condition which may raise the prior probability of h is when hypotheses of its kind succeeded in the past in similar
domains. Of course, all these factors might be judged differently by different members of the community so that
the values of
< previous page
page_90
next page >
page_91
< previous page
page_91
next page >

Page 91
the prior probabilities will diverge. There may be additional factors related to the credibility of the person or the
group who proposes the hypothesis. Hanson, for example, maintains that, other things being equal, the plausibility
of h would be higher if we have confidence in the scientist proposing it due to his successes in the past or because
of the status of leadership he enjoys in the community. Another factor is related to the way h is presented. This
factor belongs to the realm of rhetoric, which is by no means absent from scientific discourse.
The notion of "e confirms h" may be explicated by the conditions: p(h,e)>p(h). In the case when e is entailed by h
(in conjunction with some initial conditions), we find from eq. (4) that when e is accepted as true, e would confirm
h only if p(e)<1. This means that if e would be expected to be true before it was observed, then its observation
would not lead to the confirmation of h. Whereas if p(e) is small, or when e is unexpected, its observation would
confer a high degree of confirmation on h. This accords with the intuitions of scientists. Note that the latter
intuition was not taken into account in devising the model. It is, therefore, a successful prediction of the model
which confers higher confirmation on the model itself. However, in this case, the confirmation of the model is not
brought about by a prediction of a methodological rule which was unexpected when the model was proposed; on
the contrary, the rule was intuitively accepted by many scientists. This is an example for the following rule of
confirmation: A hypothesis is confirmed by an evidence which was known when the hypothesis was proposed, but
was not taken into account in constructing the hypothesis (see discussion of this rule in Chapter 3). This is indeed
one of the weak points of the model, since it does not account for this rule of confirmation. We might, therefore,
reject the model. However, if we have good reasons for believing in the model, we might try to modify it in order
to accomodate the above intuitive rule.
We can draw additional conclusions from the Bayesian scheme, some of which accord with the intuitions of
scientists. However, our interest in the logic of pursuit will lead us to investigate mainly the role played by the
prior probability. The prior probability depends on three kinds of factors: those related to the hypothesis itself,
those related to the process by which the hypothesis was generated and those related to the rhetoric of its
presentation. Traditional approaches, such as logical empiricism, would regard the first factor as dependent only on
objective standards and would deny any dependence on the second or the third factors. In the Bayesian scheme the
conditionalization upon the evidence balances the latter factors. This balancing well reflects the practice of science,
where psychological factors are balanced by empirical evidence, but they are not totally dismissed. Metaphysical
influence is indispensible, since it enriches the idea-pool of the scientific community. Psychosociological factors
are important, since in their cooperative enterprise, scientists cannot give the same credit to everyone. Thus, the
Bayesian model explicates this delicate balancing between empirical and non-empirical factors.
< previous page
page_91
next page >
page_92
< previous page
page_92
next page >

Page 92
The moral we can draw from the above story is that the process of uncovering the secrets of nature is by no means
a straightforward mechanical process of pulling up some curtain. It is a very intricate process in which
psychological, social, rhetorical and metaphysical machinery is involved as well as mathematical proofs and meterreading. However, to complicate the situation, it should be noted that even elaborated mathematical proofs and
sophisticated experiments sometimes have rhetorical power, which might divert scientists from the rational track.
For instance, the von Neumann proof regarding the impossibility of hidden-variable theories in quantum mechanics
was accepted by the scientific elite without thorough checking because of the sophistication of the mathematical
proof and because of the reputation of von Neumann as a brilliant mathematician. Only after many years was it
shown that the proof assumed hidden-variables axioms which were not compatible with Bohm's hidden-variable
model which was supposedly "refuted" by the "proof." The self-explanatory title of Trevor J. Pinch's article on the
subject is most appropriate to describe the situation: "What does a proof do if it does not prove?" (Krohn et al.
1978, 171215).
2.2.8 Discovery as a Skill: The Invisible Logic
The trained scientist who has experience in his field will recognize and discover things which the layman will not
be able to recognize. Discovery may therefore be viewed as a skill. Since skill cannot be taught by giving a list of
instructions, discovery would remain in this case beyond the reach of method. No description or recipe can replace
the expert. This is evident from the practice of expert systems in AI. Computer scientists try to translate the
experience of the expert into a set of machine-oriented instructions. They try to watch or to interrogate the experts
in order to draw sets of heuristic principles which might be translated into sets of instructions. However, at the
present state of the art the success of this method is very limited. A physician, for example, may diagnose the kind
of illness his patient has by watching him, by listening to his bodily sounds and by feeling or sensing his organs.
Although there has been some success in mechanizing limited kinds of medical diagnostics, the most sophisticated
software cannot faithfully replicate this skill.
Terry Winograd and Fernando Flores (1987) show how Martin Heidegger's analysis may account for the limitation
of expert systems. Heidegger distinguishes a domain of action from a domain of description. In bringing tacit
knowledge into action, we do it without being aware of the knowledge we employ. In the translation of action into
description by an external observer, something is lost. An expert system is a description provided by an external
observer of the expert's action. Since the translation is incomplete, the expert system does not function properly in
new situations. It is non-adaptable to new tasks, as the human experts is.
< previous page
page_92
next page >
page_93
< previous page
page_93
next page >

Page 93
Practical skills, such as riding a bicycle or baking a cake, are carried out by human beings almost "automatically,"
without paying attention to them; furthermore, when one tries to pay attention to the way he is carrying out the
task, the performance may be disturbed.
A skill involves making the right judgments and performing the proper acts in a given domain of practice. It is
acquired by experience. Past experiences are not stored in the memory such that in performing a task one simply
recalls them. Bo Goeranzorn and his colleagues studied the nature of human skill and how a skill may be affected
by the use of different technologies (Vaux 1990). In the framework of this study, a professional photographer
describes his experience as follows: "all of these earlier memories and experiences that are stored away over the
years only partly penetrate my consciousness" (ibid., 57). Thus, the rules an expert follows in performing a task are
not expressed by propositions; they are expressed directly in action. This view is in line with Heidegger's
observations. It also agrees with Ludwig Wittgenstein's view on tacit knowledge according to which following a
rule does not mean following a set of instructions; it means doing something in a practical way (Wittgenstein
1968). One acquires a skill through apprenticeship, by imitation and by non-verbal communication.
In the context of scientific discovery we may conclude that the skill and discerning power of the discoverer is
restricted to the domain in which he has acquired experience. Thus, a scientist may be a great discoverer in one
scientific field but not in another. This may be related to the different material logics employed by different
scientific communities. As I have said, a material logic is mainly part of the tacit knowledge which governs the
reasoning practice and action of a given community. Thus, part of the discerning power of a discoverer in science
is drawn from the internalization of the tacit material rules of inference. There may be rules governing a skill, but
only the experienced expert can apply them correctly. For example, in devising a mathematical theory in physics,
one might be guided by a rule of simplicity. But only the experienced theoretical physicist would know how to
apply the rule in constructing a theory, in choosing one or in adjusting it to new observational data.
Transparency, Invisibility and Black Boxing The tacit knowledge which the scientist internalizes includes the
presuppositions and background theories which are taken for granted and shared by the members of the relevant
research community. These presuppositions appear as suppressed premises in scientific discourse and
argumentation. In this sense, they are "invisible" to the expert; from the expert vantage point they are "transparent."
The expert who employs these presuppositions or suppressed premises does not "see" them; he considers them to
be ''self-evident," and he is not fully aware of them. This is the reason he has difficulties in explaining to the nonexpert what he is doing. Indeed, we encounter many cases in which an eminent scientist is considered to be a "bad
teacher."
< previous page
page_93
next page >
page_94
< previous page
page_94
next page >

Page 94
The notion of transparency has been used mainly in relation to the practice of using observational instruments (see
section 1.2). In this context, it refers to "the attribute an instrument possesses when it is treated as a reliable
transmitter of nature's messages" (Gooding et al. 1989, 3). After the scientist has acquired the skill of using an
instrument, the procedure of using it become transparent. Gooding et al. employ this notion when they describe the
historical development of the practice of using instruments such as the glass prism or the telescope. The notions of
invisibility and transparency might refer as well to the usage of the most advanced experimental equipment, such
as fast electronic detection systems in particle physics, where a much more intricate practice is involved. The
process of establishing the reliability of the instrument is called by the above authors "black boxing." When an
instrument becomes black-boxed, it is treated as transparent and the information it conveys is treated as the
messages of nature. The scientist treats the instrument as if it were an extension of his organs. Thus, when the
particle physicist looks at a bubble-chamber photograph he sees particle trajectories. When the instrumentation
becomes transparent, "only the phenomena remain" (ibid., 217) and the process of discovery becomes discovery by
exposure, although the black-boxed procedure may be highly generational relative to everyday practice, or relative
to the previous state of knowledge. Black-boxing converts discovery by generation into discovery by exposure.
Thus, we may say that observation and discovery are skill-laden (Nickles 1980, 300). If we adopt Polanyi's
distinction between focal and subsidiary awareness, we may say that the scientist has only a subsidiary awareness
of his practice in using the instrument. Only the phenomena remain under his focal awareness. When we use a tool
for performing a certain task, we are focally aware of the task. We have only subsidiary awareness of the tool
(Polanyi 1958, 55).
The notion of invisibility can be applied to the use of theoretical tools, as well. In constructing his theoretical
arguments, the scientist relies on suppressed premises or material rules of inference, which are invisible to him, in
the above sense. For the trained scientist these theoretical tools are transparent. The scientist treats them as if they
were part of his cognitive apparatus. In this sense, his theoretical argumentation looks sometimes like deductive
inference. Scientific argumentation is contaminated with suppressed premises. This is the reason why in many
typical cases, when the scientist attempts to solve a problem, he can choose among a very few hypotheses; he does
not have to choose among the unlimited number of logically possible hypotheses. It is the invisible paradigm
which narrows the range of possible solutions. The scientist who has internalized the presuppositions of the
paradigm, takes them for granted. He is not focally aware of them.
Let us consider as an example the discovery of the planet Neptune. In view of Newtonian theory, the anomalies in
the motion of Uranus were explained by Adams and Leverrier by assuming the existence of an unknown
< previous page
page_94
next page >
page_95
< previous page
page_95
next page >

Page 95
planet, Neptune, perturbing Uranus' motion, besides the perturbations caused by Saturn. The assumption about the
existence of a perturbing planet was guided by the presupposition that the motion of a planet is only affected by
gravitational forces due to the sun and other planets. This heuristic was not derived from Newtonian theory but was
tacitly assumed. No one thought about other possibilities, such as the possible existence of other forces affecting
planetary motion besides gravitation, or a change in the law of gravitational force. The theoretical presupposition
and the guiding heuristic did not have any logical validity but they were parts of the prevailing paradigm. We can
therefore treat this presupposition as an invisible assumption which was employed as a missing premise in the
inventive argument which led to the conjecture about the existence of Neptune. The problem regarding Uranus'
motion was solved by calculating the position of Neptune on the basis of Newtonian theory and the available
astronomical data. Although the mathematical solution involved some guessing, Adam and Leverrier arrived
independently at the same result since they relied on the same presupposition.
Thus, whenever a theoretical assumption becomes established, it becomes transparent. Sometimes a presupposition
is so entrenched that a revolutionary move is required to replace it, as in the example of Rutherford who "was
brought up to look at the atom as nice hard fellow, red or gray in colour..." Sometimes it is so established that the
scientist is using it without being aware of alternative paths, as was the case of Herbert Simon who was not aware
of the possibility of using different variables in reconstructing Kepler's problem. The totality of presuppositions,
methods and tools which are transparent may be viewed as the prism through which the scientist sees the world.
Some of these are the material rules of inference which determine the community-specific logic. The notions of
transparency, invisibility and black-boxing enable us in certain cases to view generational discovery as inference
or exposure. Yet, the process of discovering a new (material or conceptual) tool is a creative discovery. As we have
observed in section 1.4, this process amounts to discovering a new communication channel with nature. When the
new channel becomes transparent, the skilled discoverer treats it as a tool for exposing new aspects of reality.
< previous page
page_95
next page >
page_97
< previous page
page_97
next page >

Page 97
Chapter 3
Why Did Traditional Philosophy Of Science Ignore Discovery?
3.1 The Distinction between the Context of Discovery and the Context of Justification
The distinction between the context of discovery and the context of justification (D-J distinction) is one of the
cornerstones of logical empiricism. This is one of the weak points of traditional philosophy of science, which has
provided one of the major reasons for ignoring discovery.
3.1.1 John Herschel's Distinction: Consequentialism
The context of discovery refers to the actual processes leading to a new idea, and the context of justification refers
to the ways in which we evaluate the new idea. John Herschel was the precursor of the modern distinction between
the two contexts. He departs from Baconian inductivism according to which we arrive at laws and theories by
applying inductive rules on observations. Herschel maintains that there is a second way to arrive at laws and
theoriesby creative hypothesizing. He gives the example of Ampere's theory of electromagnetism (Herschel 1830,
2023). Ampere explained the attraction and repulsion between magnets by the existence of circulating electric
currents within the magnets. The empirical laws of electricity and magnetism known at that time could be arrived
at by applying inductive rules on observations of electric and magnetic phenomena. However, Ampere could not
arrive at the theory of circulating currents by applying inductive rules on these empirical laws. He had to apply his
creative imagination when he made the association between the magnetic phenomena and the phenomena of
attraction and repulsion between electric currents. Herschel maintains that this creative process is not governed by
rules. Had Ampere arrived at his theory by inductive inference, the process of generating the theory would confer
partial justi-
< previous page
page_97
next page >
page_98
< previous page
page_98
next page >

Page 98
fication on the product. However, the actual process of generating the theory was irrelevant to justification. The
confirmation or justification of the theory would come only as a result of testing its predictions.
Here we encounter a clear distinction between two different approaches to justification which are labeled by Larry
Laudan (1978) as "generationism" and "consequentialism." The generationist claims that theories can be
established, or justified, only by showing that they can be derived from observations. In fact, we can generalize this
notion: an idea or a theory is generationistically justified by showing that it can be derived from established
"premises," may they be observations, first principles or established theoretical foundations. Thus, according to
generationism, a hypothesis is justified in either of two cases: first, if it was generated in fact by derivation from
established premises, second, if it can be so derived, i.e. by "rational reconstruction'' of the process of discovery.
In the latter case, the role of justification is to justify a hypothesis in case we have arrived at it without explicitly
taking into account all logical steps leading to it as a conclusion of a valid argument, or without establishing the
consistency of the idea with the rest of our system of beliefs. The act of justification in this case literally means
that we try to justify an idea or to argue for it. This notion of justification is thus similar to that of a proof. Nickles
(1987) gives the name "discoverability" to this criterion of justification. A discoverable hypothesis, i.e. a
hypothesis which can be derived from what we already know, is justified. Generationism is therefore very close to
the inference view of discovery. The proper way to derive a discovery is by valid inference. If the discoverer has
arrived at his hypothesis by accident he has to show that it is discoverable by rationally reconstructing the process,
i.e. by proving the hypothesis.
The consequentialist claims that theories are tested only with respect to the success of their predictions. The
hypothetico-deductivists belong to this category. If we talk about evaluation, rather than justification, we can
include also Popper in this category, since a theory is refuted by its false predictions, irrespective of the manner in
which we arrived at it. Using this terminology, Herschel is a consequentialist; he maintains that there is no way to
"prove" a theory (which is not an inductive generalization) or to derive it from secure premises. So that it can be
confirmed only by the success of its predictions.
According to the generationist interpretation, we argue to a hypothesis we propose and we are, therefore, entirely
responsible for it; if it turns out to be inconsistent with the known facts or with some of our other beliefs, or if we
argued for it fallaciously, we are to be blamed for making a mistake. This interpretation reflects a rationalist
attitude. However, in proposing a hypothesis about the world, it is not enough to show that it is consistent with our
other beliefs. We also have to show its empirical adequacy, i.e. that its predictions agree with observations. If they
do not, we should not be blamed for making a mistake, as Popper, for example, implies when he says that scientists
"learn
< previous page
page_98
next page >
page_99
< previous page
page_99
next page >

Page 99
from their mistakes" when their hypotheses are refuted. When our conjectural law of nature or theory is refuted by
observation, our position is closer to that of the unlucky gambler rather than to that of someone who makes a
logical mistake. The view that a refuted hypothesis constitutes a mistake is a remnant of the inference view of
discovery; we argue to our hypotheses and if the argument is invalid it means that we have made a logical mistake.
Yet, even if we do not make a logical mistake, our hypothesis may still not be empirically adequate; our system of
beliefs may be perfectly consistent but empirically false. Consequentialism accounts for this situation. We cannot
derive a theory from secured premises. We can only confirm it by testing its predictions. The requirement that the
theory should be consistent with our other relevant beliefs is only a necessary condition.
3.1.2 Reichenbach's D-J Thesis: Generationism
The modern logicist version of the D-J distinction was introduced by Hans Reichenbach (1938). In addition to the
irrelevance of the actual process of discovery or generation to justification, this version also implies that the task of
epistemology and philosophy of science is to deal only with the context of justification, whereas the context of
discovery is left to psychology, which can deal with the actual processes of thinking. It should be noted that
Herschel did not exclude the context of discovery from the domain of the occupation of the philosophy of science
since he believed that some laws are arrived at by induction, which is rule-governed. The following passage
summarizes Reichenbach's thesis:
Epistemology does not regard the processes of thinking in their actual occurrences; this task is entirely left
to psychology. What epistemology intends is to construct thinking processes in a way in which they ought
to occur if they are to be ranged in a consistent system; or to construct justifiable sets of operations which
can be intercalated between the starting-point and the issue of the thought-processes, replacing the real
intermediate links. (Ibid., 5)
Thus, epistemology is prescriptive (this is indicated by the usage of the word ought) by virtue of its logical force,
whereas psychology would describe the actual ways we arrive at ideas. We may add also sociology, anthropology
and even biology to the descriptive sciences which can deal with discovery. According to this view, Poincar's
hypothesis about the subconscious processes of discovery or the social processes leading to discovery (which will
be discussed in Chapter 7) should be treated by psychology and sociology; they are not a subject for the
philosophy of science.
In the above quotation, Reichenbach refers to justification by logical
< previous page
page_99
next page >
page_100
< previous page
page_100
next page >

Page 100
reconstruction, rather than to empirical confirmation. Thus, unlike Herschel, he is a generationist, rather than a
consequentialist. However, in contemporary discussions on the D-J issue with respect to hypotheses about the
world, the term justification is used also with the consequentialist interpretation, referring to the empirical
confirmation of the hypothesis. Logic still plays a central role in this wider interpretation: the new observational
results should bear some logical relation with the hypothesis in question.
There are three theses built into the D-J distinction: (a) that there is a sharp line separating the two contexts, (b) that
only the context of justification is amenable to logical analysis and (c) that descriptive science, such as psychology,
is irrelevant to the context of justification. Thus, the context of discovery can be dealt with by descriptive science,
whereas the context of justification is prescriptive or normative. The proponents of this view maintain that the
philosophy of science is a logic of science which should, therefore, deal only with the context of justification. The
most important consequence is: (d) information regarding the context of discovery is irrelevant to justification.
There are two reasons for this: first, the context of discovery is a-logical and, therefore, it does not have epistemic
or justificatory force, second, justification is concerned with the final product of discovery which is a statement or
a set of statements. Hence, it does not matter how the discoverer arrived at his product. Logic will give us the
whole information about its truth. Thus, logical empiricism regarded the study of the process of discovery as an
empirical inquiry to be dealt with by empirical science. The only respectable engagement of the philosopher of
science was considered to be the logical analysis of the products of scientific discovery.
There are two paradigmatic cases which seem to confirm the above thesis. The first is the discovery of a
mathematical theorem. One can arrive at a theorem intuitively, by a flash of insight, for instance. But the proof is a
matter of a logical act and it does not depend on the way the discoverer hit upon the theorem. The second is the socalled chance discovery or an unintentional discovery. It would be implausible to claim that an unintentional
process of discovery is amenable to logical analysis and that the context of discovery is relevant in that case to the
justification.
It should be stressed that the so-called context of justification encompasses all kinds of evaluative implications for
a hypothesis. These include, in addition to confirmation and acceptance, also disconfirmation and refutation. Thus,
the term justification is somewhat misleading. Indeed, Karl Popper, who rejects justification and accepts refutation
only, joins in on the claim that the context of discovery is irrelevant to the logic of science. This is one of the
cornerstones of his philosophy, expressed in The Logic of Scientific Discovery (which denies the existence of its
subject matter): "The question of how it happens that a new idea occurs to man...may be of great interest to
empirical psychology; but it is irrelevant to the logical analysis of scientific knowledge"
< previous page
page_100
next page >
page_101
< previous page
page_101
next page >

Page 101
(Popper 1959, 31). The term context of evaluation may, therefore, be more appropriate than "context of
justification."
Thus, according to the proponents of the D-J distinction, a philosophy of science is a logic of science and a logic
of science is a logic of justification or evaluation. This is perhaps one of the major assumptions which has shaped
most of the twentieth-century philosophy of science, which had been dominated by logical empiricism and its
offspring.
3.2 Objections to the Distinction
3.2.1 Justification and Discovery are Inseparable
The most obvious objection to the D-J distinction is that the two contexts are inseparable; each context is
"contaminated" with elements of the other. This is an objection to thesis (a) and, as a consequence, also to the other
three theses, which depend on the validity of the first one. The context of discovery is contaminated with
justification or evaluation. Indeed, as I have emphasized, evaluation is an integral part of the process of discovery,
since an entity would be considered a discovery only if it was proved to be true, successful, a solution to a
problem, etc. In particular, if we refer to a theory, we would not say we discovered it unless we have confirmed it.
In scientific practice, as we have observed, if a scientist proposes a theory and did not prove or confirm it, he
would not always be considered to be the discoverer of the theory even in case it was later proved or confirmed by
another scientist. He would be credited only for participating in the process of discovery. By the act of
confirmation, we discover that the theory is indeed a discovery. Moreover, sometimes we discover that an idea or a
theory which was generated in an attempt to solve a different problem, solves our present problem. The act of
discovery consists here of association and confirmation. These two acts may be simultaneous, especially if the
discovery is an eureka event. Indeed, many ideas pass through our minds without any notice. However, we make a
discovery when we see that a particular idea in this flux solves a certain problem. Poincar would tell us that
discovery consists mainly of selection, which is an evaluative act. There is no shortage of ideas, but there is a
shortage of successful ideas. Thus, we would not say that Democritus discovered atoms. Dalton did, since he
provided us with a confirmation of the conjecture in a particular context.
Let us consider the validity of the D-J distinction in the two kinds of discovery: exposure and generation. In
discovery by exposure, the context of discovery mainly consists of justification. When, for example, we discover a
new star with a telescope, the act of discovery may be identical with the act of observation. If we are empiricists,
who take observation to be a warrant for truth, the act of justification is thus identical with the act of discovery. If
we deduce an interesting or unexpected conclusion from a set of accepted
< previous page
page_101
next page >
page_102
< previous page
page_102
next page >

Page 102
premises, such as a theory, we make a discovery by exposure. The act of discovery is identical here with the act of
deductive inference. It is again identical with the act of justification. Thus, discovery and justification coincide.
In generational discovery, the process of generation may precede the act of confirmation. In this case, the theory is
constructed and then tested and confirmed. We can then separate the context of generation from the context of
justification, rather than the context of discovery from the context of justification. This is an ideal case for the
logician of science, when the theory is generated full-blown and then tested and confirmed by the data. However,
in many, if not most, cases the theory is dynamically constructed as was described in section 2.2. Thus, the theory
is constructed by adjusting it to the data. The process of adjustment to the data, which belongs to the context of
generation, is also part of the process of justification. In each step of the process, we modify the theory in such a
way that the new version will be generated partially justified. Then we submit it to further tests and the process
continues, where the dynamic theory may gradually become increasingly more confirmed. Hence, in this case
generation cannot be separated from justification.
Now, what about an unintentional process of discovery? In subconscious or involuntary processes of discovery, a
final stage of justification must come after the process of generation, since the latter process does not provide
justification in the traditional sense. Thus, unintentional generation (rather than discovery) and justification are
separated. This point will be further elaborated in section 7.5.
3.2.2 Justification is Not Aprioristic
The attempts to construct a logic of justification, such as inductive logic or probabilistic theories of confirmation,
have failed. Induction and confirmation cannot be justified by, or reduced to, deductive inference. We may
conclude, therefore, that there is no valid algorithm of justification. Indeed, confirmation is not a matter of logic; it
does not have the status of a logical proof. Rather it depends on the scientist's system of beliefs and on
psychological and sociological factors. Hence, both contexts are not guided by logically valid principles and,
therefore, there is no epistemic priority to justification over discovery; both have the same epistemic status. Hilary
Putnam argues that if any of the two contexts are guided by rules, these have the status of maxims (Putnam 1973,
268). The above argument can be raised against thesis (b) of the D-J distinction. It is not only the process of
discovery which is not amenable to logical analysis; also justification is not.
However, even if we could show that justification is a matter of logic, we might still question the justification for
the rules of logical inference. If we do not treat the principles of deductive logic as a priori justified, then how
would we justify these very logical principles which confer justification upon our
< previous page
page_102
next page >
page_103
< previous page
page_103
next page >

Page 103
ideas? In the next chapter, I will propose a view which maintains that the rules of logical inference are by
themselves justified in virtue of their being a good explication and explanation of our intuitive inferential practice.
Hence, those logical principles, which supposedly eliminate descriptive psychology from epistemology, are by
themselves justified by reference to actual thinking processes. Which discipline then, if not empirical psychology,
can determine what are the rules behind our inferential practice? Thus, both contexts are not amenable to logical
analysis and even if they were, this is not an aprioristic logic but a logic which is susceptible to empirical
investigation. This argument can be raised against theses (b) and (c).
Harvey Siegel (1980, 301) raises the following objection to Putnam's claim: "the point of Reichenbach's distinction
is that information relevant to the generation of a scientific idea is irrelevant to the evaluation of that idea; and this
distinction between generation and evaluation (or discovery and justification) can be instructively maintained
despite the fact that both contexts are guided by maxims."
Before I will propose my objection to Siegel's objection, I will offer an example which seems to be favorable to
his claim. Suppose the construction of a theory is guided by maxims of adherence to a certain world picture or to
some general established theoretical principles. If, on the other hand, the maxims of evaluation depend only on the
formal syntactic relations which hold between the theory and certain observational sentences, and not on the
content of the theory, then justification, indeed, does not depend on the maxims of generation, which refer to the
content of the theory. A simple example for such a situation would be the case when the maxims of generation
demand that an explanatory theory of gas behavior be corpuscularian and adhere to the mechanistic world picture,
whereas the maxims of justification demand only that the logical implications of a theory match the observational
results according to some formal confirmation theory. Here validation is indeed independent of the context of
discovery or generation. However, if validation by logical (syntactic) standards is not aprioristic, there is no reason
to treat these standards as giving an absolute warrant for the truth of the theory.
Thus, by claiming that "the point" of the D-J distinction is that the context of discovery is irrelevant to the context
of justification, Siegel ignores the epistemological import of the distinction. According to Reichenbach and his
followers, the two contexts differ in their epistemic status; the context of discovery does not confer an epistemic
warrant to the discovered idea, whereas the context of justification, because of its logical force, does. Hence, if
both contexts are maxim-guided, the epistemological spirit of the thesis will be absent. The context of justification
will have no normative import, i.e. it will not have any epistemic superiority over the context of discovery as the
thesis requires. Putnam's objection refers, therefore, to the absence of epistemic superiority for the context of
justification when it is maxim-guided.
< previous page
page_103
next page >
page_104
< previous page
page_104
next page >

Page 104
3.2.3 Information about Generation is Necessary for Evaluation: Predictability and Novelty
If the scientist knows what are the content-dependent criteria of plausibility of hypotheses, he will construct the
theory, taking into account these criteria. This is the subject matter of the logic of pursuit discussed in section 2.2.
If the hypothesis complies with these criteria, its plausibility, or prior probability, contributes to the final degree of
confirmation, as can be seen through Bayes's formula. Thus, in this case, an information about the generation of
the hypothesis is relevant to the evaluation of the hypothesis. But then, the maxims of hypothesis-generation are
derived from the maxims of evaluation.
If the standards of evaluation are taken into account when the theory is generated, information about generation
might be relevant to, but not necessary for, evaluation. When a hypothesis is proposed, we do not have to look for
the criteria which guided the scientist who generated it. We simply inspect the hypothesis itself and find out
whether it complies with the world picture, for example. Indeed, the D-J thesis implies that although information
about generation might be relevant to evaluation, it is not necessary for evaluation. However, as I will show in
what follows, there are important cases where such information is necessary.
A very important question which bears on the confirmation of a scientific theory is whether or not a certain event
or phenomenon which is predicted by the theory was known at the time the theory was proposed. In scientific
practice it is well known that when the predicted event is observed, or becomes known, only after the theory was
proposed, then the theory's degree of credibility rises considerably, provided that the event has not been expected
on other grounds. Typical examples are the discovery of a new planet (e.g. Neptune) or a new particle (e.g. the
omega minus), which were predicted by physical theories. This intuitively accepted methodological principle
cannot be explicated in a confirmation theory which depends only on syntactical or formal relations between the
theory and observation statements, since such relations are timeless, i.e. are insensitive to time priorities. I shall
illlustrate this point by the following example. Let us consider Nicod's rule of confirmation which states roughly
that statements of the form ($x)(Ax & Bx), i.e. "there exists an object of the kind A with a property B," confirms
the law-like statement (x)(Ax Bx), i.e. "every A is B." This rule is insensitive to the question of when each
statement became known to the scientist who proposes the hypothetical law. If our maxims of justification or
confirmation are of the formal kind, like this one, then information as to how a hypothesis has been arrived at is
irrelevant to its confirmation. Typically, a formal rule of confirmation takes into account the final products of
scientific discovery, i.e. the statements which describe laws and theories, without taking into account the history
which led to their discovery. Thus, the D-J distinction thesis fits in well
< previous page
page_104
next page >
page_105
< previous page
page_105
next page >

Page 105
with the ahistoric epistemological view, which maintains that all that is relevant to the validation of a claim is its
formal structure and relations to other claims.
If we wish to stay close to actual inferential practice in science, however, we have to devise confirmation rules
which will do justice to the above historical or generational consideration. With such rules, the question of whether
or not a theory has been constructed with the knowledge of certain facts or phenomena will be relevant to the
evaluation of the theory. The higher the percentage of facts previously unknown to the discoverer which are
successfully predicted by the theory, the higher will be the epistemic status conferred upon the theory by these
rules. In such a case we can say that we have gained something from the theory.
The fact, or the methodological requirement, that a theory gets strong support from its novel predictions is put into
question by several authors (Donovan, Laudan, and Laudan 1988). According to these authors, the case studies
they present show that theories can be accepted without meeting the above requirement. Finocchiaro finds no
evidence that Galileo demanded that the Copernican worldview should yield novel predictions (Finocchiaro 1988,
4967). Hofmann claims that Ampere did not make this demand as a touchstone to his electrodynamics (Hofmann
1988, 201217). And Zandvoort finds that surprising or novel predictions played no role in the acceptance of nuclear
magnetic resonance (Zandvoort 1988, 337358).
These cases indicate that in the absence of novel predictions, scientists may accept a theory if it successfully
explains known facts. But there are degrees of acceptance. The following generalization is supported by examples:
if, in addition to the explanation of known facts, the theory also predicts a novel fact and this fact is later
discovered, the effect of this event on the credibility of the theory is much more dramatic, and the acceptance of
the theory is much stronger. For example, the unitary symmetry theory had been accepted for pursuit by particle
physicists before the discovery of the W- particle, since it had successfully explained a variety of known facts and
relations much better than its competitors, such as the Sakata model or the G2 symmetry group. Yet, the discovery
of the W- (Barnes et al. 1964), which had been predicted by the theory, led to the final acceptance of that theory,
with a dramatic impact.
The requirement that a theory should be judged by its ability to solve problems it was not invented to solve, or to
explain phenomena it was not invented to explain, is supported by many similar examples. According to this
requirement, the credibility of a theory may rise when it explains things already known, provided they were not
taken into account when the theory was constructed. But from the vantage point of the discoverer there is no
difference between the case in which the predicted fact was not known at the time of discovery and the case known
but not taken into account. The question is whether the discoverer generated his theory without taking into
< previous page
page_105
next page >
page_106
< previous page
page_106
next page >

Page 106
account a certain fact. If the fact was not taken into account and it is afterwards explained by the theory, it makes
no epistemic difference whether the fact has been already known or not. Yet, there may be a difference if an
already known fact subconsciously effected the process of generating the theory. Since we have already entertained
the possibility that part of the process of discovery is subconscious, we should take this possibility into account.
This issue will be further discussed in Chapter 6.
Now, I would like to introduce the dynamic view of confirmation which is intimately related to the notion of
dynamic theory. The proponent of the D-J distinction would claim that it does not matter how the theory was
constructed. The act of justification is the logical act of comparing the theory's predictions to the data. However,
since the theory can always be modified so that it would fit the data, there is no sense to this notion of static
confirmation or justification, since the theory may be always be kept "justified" in this sense. Justification should
therefore be a dynamic notion. According to the dynamic approach which I propose (1978, 1979) a theory is
confirmed only if in our efforts to adjust it to the data we make what I call epistemic profit. Namely, if we gain
from the dynamic theory more empirical knowledge than we have invested in it. By our investment and gain I
mean the following. In the dynamic process we adjust the theory to already known data which constitute our
investment. The new versions of the theory may yield successful predictions of empirical data, which constitute our
gain. Thus, it makes the whole difference in the world if all, or most, of data explained by the theory was known
and taken into account before the theory was constructed and the theory was adjusted to it, or if the theory
predicted all or most of this data. So it is not the final result which counts, but the way it was obtained. In addition,
the function of unification should be taken into account. Known data or phenomena which have been epistemically
isolated are unified under the theory's explanatory umbrella. This may still be regarded as an epistemic profit in a
wider sense. The dynamicist view may be expressed by the following words of Frederick Suppe: "Full epistemic
understanding of scientific theories could only be had by seeing the dynamics of theory development" (Suppe
1974, 126).
The above approach may be called dynamicism. It can be contrasted with generationism and consequentialism,
which seek a warrant for the claims our hypotheses make. Dynamicism seeks criteria for judging the potentialities
and fruitfulness of our hypotheses, rather than a warrant for truth. A false theory may be very fruitful in providing
novel and successful predictions.
3.2.4 The Context of Generation Has an Epistemic Dimension
If we believe that the search for truth is not totally blind (as we will see, even evolutionary epistemologists believe
so), i.e. that among the infinite number of the logically possible hypotheses humans frequently arrive, in particular
in
< previous page
page_106
next page >
page_107
< previous page
page_107
next page >

Page 107
science, at hypotheses which prove successful, it is unreasonable to exclude this fact from our epistemological
discourse. We may require, therefore, that epistemology and philosophy of science account for this fact. According
to this view, science is not only an evaluator of ideaswhatever these may bebut more importantly, a generator of
successful ideas. Hence, the ways by which humans, and in particular scientists, come to new ideas are
epistemologically important. There must exist a rational way to generate good ideas in a reasonable frequency. We
would otherwise be engaged in testing all kinds of hypotheses with no preferred direction, the chances for progress
being accordingly diminished.
It may be instructive to draw an analogy with training and judging in sport. The sport of running, for example, has
two "contexts," the context of training and the context of judging. According to the training-judging (T-J)
distinction thesis there are no exact criteria as to how to produce a good athlete and to improve results. Only the
measurement of the results achieved in a particular competition and the decision as to who is the winner are guided
by exact criteria. Thus, the process of scientific discovery is analogous to the process of training. The product of the
first process is a theory (or a law). The product of the second process is a trained athlete. The evaluation of a
theory, relative to the competing theories, is analogous to the evaluation of the achievements of the athlete, relative
to his competitors. The quality of a theory is judged with respect to its success in explaining and predicting
observational data and phenomena, relative to other theories. The quality of an athlete is judged with respect to his
scores in competitions. Furthermore, in the case of a sport such as running, we cannot object to the claim that the
methods of training and producing good athletes are irrelevant to the way of choosing the winner in a competition.
Hence, this is a perfect analogy to the case where the D-J distinction is valid. Here, the method of measuring the
results is analogous to the logic of justification, whereas physiology and psychology of sport, for example, are
analogous to the psychology of discovery. Thus, the following theses are true: (a') there is a sharp line separating
the context of training and the context of judging, (b') only the context of judging can benefit from the the methods
of measuring results, (c') physiology and psychology of sport is irrelevant to the context of judging and (d') the
context of training is irrelevant to the context of judging.
However, the T-J distincion thesis does not imply that a theory or a methodology of running should concentrate
only on the methods of judging. This thesis would surely seem to offer too narrow a view of athletics, since it
ignores the major goal of athletics, i.e. the goal of generating good athletes and breaking records. Moreover, if the
aim of athletics is ever-improving achievement, it would be irrational for someone who wishes to understand this
human activity to be solely engaged with judging, and not at all with the methods of improving achievement. It is
of course entirely rational to be interested
< previous page
page_107
next page >
page_108
< previous page
page_108
next page >

Page 108
in methods of judging, but why call this a ''methodology of athletics"; a better name would be "methodology of
judging athletics." Thus, if the aim of the theory of athletics is to understand this human activity and to improve it,
the analogue of the conclusion adopted by logical empiriciststhat the theory of athletics should conentrate only on
the context of judgingwould sound absurd here.
The moral in this for the philosophy of science is that it will miss the essence of science if it concentrates solely
upon evaluation. Rationality in science resides not only in the activity of testing theories, but also, and perhaps
mainly, in the activity of generating theories which are good candidates for testing. If the aim of science is
progress, it would be irrational to be engaged in testing ideas with no regard to their content or origin. If science
had been engaged only in testing theories, with no regard to which theories it was testing, it is very doubtful that it
would have arrived at its spectacular achievements in understanding and mastering natural phenomena.
Traditional rationality is categorical rationality. It is concerned with truth. Nothing is wrong with truth. However,
according to this conception of rationality, we would be perfectly rational if we had been only engaged with
tautologies or with very shallow truths which can be proved beyond any doubt. We also would be perfectly
rational if we aim at arriving at absolute truth, without any chance of success. In both cases we would know
nothing about the world, but we would be perfectly rational. A philosophy of science which is only interested in
truth and in evaluation would adopt the categorical conception of rationality and would secure the truth of
scientific claims. However, if we want to know something about the world, it would be irrational to adopt this
conception of rationality.
Karl Popper goes to the other extreme. He requires that scientists will make bold conjectures. Thus, they would not
be engaged in tautologies or in shallow truths. He says (1969, 215) that "continued growth is essential to the
rational and empirical character of scientific knowledge." He stresses that the way of growth as he conceives iti.e.
by conjectures and refutationsis responsible for the rational and empirical character of science. However, he
maintains that it is not the business of epistemology to study the process of arriving at a new conjecture. His only
requirement on hypothesis generation is that a new hypothesis will be bold, i.e. that it will significanly depart from
what is already known or expected. But if the generation of conjectures were not guided by a mechanism of some
epistemic merit, scientists would be engaged in criticizing slack hypotheses which lead nowhere. Bold hypotheses
may lead in unfruitful as well fruitful directions. Since the fruitful directions constitute a very small percentage of
all possible directions, scientific knowledge would have very little chance to grow in this way. A rational gambler,
who gambles with truth, would not be satisfied with the advice to bet "boldly," since under this advice he would
have more chances to lose than to gain a for-
< previous page
page_108
next page >
page_109
< previous page
page_109
next page >

Page 109
tune. Hence, one of the tasks of epistemology should be to investigate this mechanism for generating conjectures,
which facilitates the growth of scientific knowledge and which therefore, according to Popper himself, guarrantees
the rational character of scientific knowledge.
We may conclude that understanding the process of discovery is not only a legitimate occupation for the
philosopher of science, but is essential for understanding the phenomenon of science.
We are facing a seemingly paradoxical situation. On the one hand, the logician would tell us that there are too
many (infinitely many) possible hypotheses for explaining anything. On the other hand, every scientist would tell
us that it is sometimes impossible to find even one explanation. We can always invent a theory from which we
could derive any given explanandum, may it be an event, a phenomenon or a regularity. And when we adopt one
theory, we can protect it from refutation as long as we wish, by making modifications in other portions of the body
of knowledge. This is what the "Duhem-Quine thesis" says (see Lakatos 1970). Some philosophers conclude from
this that science is and should be pluralistic (see, for example, Feyerabend 1978). However, at least in natural
science, pluralism and explanation do not come together; if we are left with more than one possible explanation, we
do not have any!
Thus, the difficulty of the logician is to dispose of all his candidate theories but one, whereas the difficulty of the
scientist is to discover one possible explanation for a given phenomenon or anomaly. The reason for this "paradox"
is that scientific explanation is not a logical notion. The logician is "blind" to the content of the theories. Not every
statement from which we can derive the explanandum is a possible explanation. An explanation should obey
certain requirements. Some of the requirements are logical, formal or methodological, such as predictablity or
simplicity. The more important requirements are the context-specific ones, which cannot be captured by the
logician's tools. The progress of science is signified by narrowing the range of possible hypotheses. This means, for
example, that the explanans should comply with the established knowledge which has been formed in this process.
And this is a question of content, rather than of form. This is exactly where the method or the theory of discovery
might help us. As we will see in the next chapter, in a naturalistic approach to the philosophy of science, the theory
of discovery may yield a method of discovery. A method of discovery would tell us how to generate, or how to
arrive at, hypotheses of the kind which have more chances to succeed under the extra-logical requirements. A
theory of discovery would provide us with an explanation of the fact that we arrive in a finite time at a single
successful hypothesis among the infinite possible ones.
< previous page
page_109
next page >
page_110
< previous page
page_110
next page >

Page 110
We would understand this phenomenon if we had a method for generating successful hypotheses at a relatively
high rate and if we can show that science, in fact, adopts this method. Alternatively, we could view scientific
discovery as an involuntary, or natural, phenomenon and look for a scientific explanation for the success of
science. If we adopt, for example, a view such as Poincar's, we would have to explain the mechanism which
weeds out the large number of ideas created in our minds before the better ones rise above the threshold of
awareness. In Part II, I will try to view the philosophy of science as an explanatory discipline. As such, it will be
capable of providing an explanation for the phenomenon of discovery.
< previous page
page_110
next page >
page_111
< previous page
page_111
next page >

Page 111
PART II
DISCOVERY NATURALIZED
< previous page
page_111
next page >
page_113
< previous page
page_113
next page >

Page 113
The Prepared Mind: Cultivating the Unintentional

Both conceptions of discovery by exposure and discovery by generation imply that the process of discovery is
carried out intentionally by the discoverer and that it remains at all stages under his control. However, this is an
idealized, and even a distorted, picture. The following are the most obvious kinds of discovery processes which are
unintentional or involuntary. The first kind is the intrapsychic process of creation, such as the process of
incubation, where the discovery is a product of a subconscious activity. Furthermore, according to the theory which
will be expounded in Chapter 6, every discovery includes subconscious stages. Second, the process may be a
cooperative-historical enterprise, where individual scientists do not intentionally generate the product of discovery.
According to the social oriented view of discovery that will be developed in Chapter 7, unexpected discoveries are
liable to occur in a cooperative-historical setting. Third, in the exercise of skill, as we have seen, the discoverer in
not fully aware of the process of discovery. All these kinds of phenomena are involuntary or unintentional. Either
the discoverer participates in such a process or hosts it. Involuntary discovery may be generational. Many theories
throughout the history of science have been generated in a cooperative and/or historical process. An incubation
process may yield a novel idea. It may be a process of exposure as well. The discoverer may expose a deep
structure, or a solution to a problem, by a flash of insight. However, a flash of insight, or an eureka event, may be
a culmination of a subconscious processan incubation process, for instance. So, although it seems to be a discovery
by exposure, the whole process is generational.
It is customary to confront method-governed discoveries with so-called chance discoveries. However, it is not clear
what "chance" means here. Archimedes' discovery of his law, and Fleming's discovery of penicillin are sometimes
quoted as examples of chance or accidental discovery. However, chance favors the prepared mind, to use the
phrase coined by Pasteur. In each of these cases, the discovery required a prepared mind. Thus, a particular
observation, experience or thought triggers the process of discovery only if the discoverer is prepared for it. Hence,
these cases can be included in the category of "involuntary" processes. The theory of discovery which will be
developed in the following chapters will give a specific interpretation to the notion of the "prepared mind."
The involuntary or unintentional discoveries are the most creative; these are the processes which enable science to
generate novelty and to make
< previous page
page_113
next page >
page_114
< previous page
page_114
next page >

Page 114
radical progress. Hence, since the algorithmic or computational conception of method applies only to intentional
processes, it will not apply to major scientific discoveries. Method and inference may apply only to discovery by
exposure and to some kinds of framework-preserving generational discoveries. And yet, even the latter involve
unintentional steps.
However, the discoverer may contribute to unintentional or involuntary processes of discovery. We cannot supply
the discoverer with methods, algorithms or rules for making unintentional discoveries. Rather we can provide him
with a set of suggestions or recommendations for nourishing and cultivating the process. This cannot be regarded
as a method leading to discovery, since unintentional processes cannot be directed or guided. The suggestions aim
at preparing optimal conditions for the process of discovery. I would compare these recommendations to directions
for growing plants or animals and even for educating children. The process of growth in these cases is a natural
process. However, we can affect the process and its products by cultivation. Only if we properly plough and
fertilize the soil, will the plant grow and reach an optimal size and shape. In growing animals, we have to feed
them properly, and we may train them. In raising children, we are also engaged in education.
The best way to comprehend the notion of "cultivating discovery" is by using the example of the farmer or the
gardener who cultivates the soil, scatters the seeds and irrigates the field. The products of the process are the plants
which grow up, some of which may exhibit novel characteristics. Since some seeds may have been hidden in the
ground or were blown by the wind from other fields, the farmer or the gardener may literally discover unexpected
wild plants which yield unexpected fruits. Thus, the farmer scatters seeds and affects the environment, whereas the
final products depend on both seeds and cultivation, according to the "nature vs. nurture" dichotomy.
What is the moral of this story for cultivating discovery? The prepared mind is the cultivated soil and the ideas are
the seeds. Only the prepared mind can benefit from ideas which it encounters. A psychologically oriented view of
discovery, such as Poincar's, can be illuminated by the cultivation metaphor. According to this theory, the most
important part of the process of discovery takes place under the discoverer's threshold of awareness. It is important
to note, therefore, that the discoverer acts just as the host of the process, without consciously contributing to it. The
discoverer cannot be held fully responsible for the subconscious stage of the process. This is the same as one
cannot be held fully responsible for any involuntary physiological process, such as sweating, heart functions and
breathing, which takes place in his body. He can, however, do things which will improve or affect the quality of
the involuntary functions or the subconscious processes.
The following recommendations may be drawn from this psychological conjecture. The "instructions" here are not
very sophisticated, and are part of
< previous page
page_114
next page >
page_115
< previous page
page_115
next page >

Page 115
common knowledge. The discoverer should be knowledgeable in the field, seek information about observational
results, be open to ideas from other fields and be familiar with the specific kind of problems for which a solution is
sought. All these may nourish the involuntary process of discovery. Another kind of advice may be given for
preparing the optimal conditions for the process. For example: if you have difficulties in solving a problem, leave
the problem aside, do other things, try solving another problem, take a rest. These suggestions aim at paving the
way for involuntary and subconscious activities. We encounter many descriptions by scientists of how they arrived
at their discoveries, which are in line with these pieces of advice. Additional recommendations will be given in
Chapter 6, when we discuss a psychological theory of discovery.
Other suggestions may be derived from the social nature of science. In this case the aim of cultivation is to prepare
the so-called collective mind of the scientific community. This notion will be specified in Chapter 7.
The concept of "cultivation" is not a logical or a methodological concept. I have attempted to clarify it by using a
naturalistic metaphor. Indeed, the above view may be labeled a "naturalistic" view of scientific discovery. It is
diametrically opposed to a logical approach and to the possibility of mechanizing scientific discovery. In the
following, I will develop a naturalistic theory of science and scientific discovery which will bear upon the process
of cultivating discovery. The above recommendations or rules are drawn from psychologically and sociologically
oriented theories of science, which will be expounded in Chapters 6 and 7.
To conclude, the alternative to method-governed discovery cannot be described as "chance" discovery. Rather, it is
involuntary discovery. The latter kind of discoveries are those which favor the "prepared mind." Cultivation is the
act of "preparing the mind." Although we may expect the highest degree of novelty to be created by involuntary
processes, the philosophy of science has been totally ignorant of them, treating them as anecdotes or relegating
them to the realm of curiosities. In the following chapters I will attempt to show how this deficiency may be
corrected. But first, I will develop my conception of a naturalistic or an explanatory philosophy of science.
< previous page
page_115
next page >
page_117
< previous page
page_117
next page >

Page 117
Chapter 4
Philosophy Of Science: From Justification To Explanation
In recent years, the philosophy of science has undergone radical changes. With the decline of logical empiricism, it
is not believed as widely as before that the source of scientific rationality can only be found in some system of
formal logic or methodology. The philosophy of science, however, has not yet settled on a new widely accepted
path. Thus, fundamental questions are raised with respect to its scope, tasks and methods. For example, what
should supplement or replace the logical analysis of science? Should the philosophy of science be closely linked to
the history of science or should it perhaps be converted into a science of science? There is however one prior,
more fundamental, question which has engaged traditional philosophers of science, and which is now posed more
forcefully. The question is whether the philosophy of science should adopt the task of appraising scientific claims:
should it be content with the more modest aim of describing science, its methods and evolutionary patterns, or
should it have both descriptive and prescriptive functions? This is the normative-descriptive (N-D) dichotomy
which together with the D-J dichotomy evoke the cardinal issues related to the nature of the philosophy of science.
I will start by analyzing the N-D dichotomy. Traditionally, this dichotomy refers to the context of justification, i.e.
to the question whether the methodological rules of confirmation or refutation, acceptance or rejection of
hypotheses are normative or descriptive. However, since I attempt to develop a philosophy of science which deals
with both "contexts," I will discuss the implications of the N-D dichotomy for discovery as well. We will arrive at
the conclusion that neither purely prescriptive nor purely descriptive philosophy of science are possible. The
nonaprioristic scheme which will be expounded in this chapter can be viewed as a scheme of justification or as a
scheme of explanation. As a scheme of explanation it will function as a naturalized philosophy of science, or as a
science of science. Explanatory philosophy of science does
< previous page
page_117
next page >
page_118
< previous page
page_118
next page >

Page 118
not undertake the task of telling scientist how to do science. I will argue that it still has a normative role in a
specific sense which resembles the role of psychological advice or therapy. This kind of philosophy of science is
suitable for treating discovery and in particular involuntary processes of discovery.
4.1 Normative Philosophy of Science: Justification Relativized
4.1.1 Instrumental Rationality: Science as a Goal-Directed Activity
The mainstream of traditional philosophy of science starts off with a normative or prescriptive attitude. It seeks
rationality in science, i.e. it looks for the logic or reasoning behind scientific acts. Scientific rationality depends on
the goals of science. If the goal is different from the goal of truth, we might call this conception of rationality
instrumental rationality, i.e. rationality as an effective instrument for achieving the goal. It is therefore the first task
of the philosopher of science to uncover these goals. If the goals are found, the philosopher can try to answer the
question as to whether or not the proposed means for achieving them are appropriatei.e. rational. Most traditional
philosophers of science have taken for granted the assumptions that science is goal directed and that the main goal
of science is reaching comprehensive truth about the world. Thus, they were commited to categorical rationality.
Truth is a property of beliefs or statements. Therefore, assigning to science the goal of truth means that the task of
science is to generate true beliefs expressed by true statements about the world. Hence, the rules of deductive logic
are the natural candidates for showing us how to do good science. If we see the task of science as generating
statements which are highly probable, then some theory of probabilistic inference will guide us in doing science
rationally. Thus, deductive, inductive or probabilistic inference schemes will be the basis for rational acts in
science. Adopting this approach, which may be called "logicism," the philosopher of science views science as an
"inference machine." This view has implications for both discovery and justification.
Logicism has faced insurmountable difficulties. Some of these difficulties are related to the impossibility of
reconciling many of the actual acts of and decisions of scientists throughout the history of science with those
recommended by the logicist methodologies. One could assume that this might lead philosophers of science to
review their fundamental presuppositions. One such presupposition is that science is a truth-seeking system.
If we abandon this presupposition while still believing that science is a goal-directed activity, we can try to suggest
alternative goals. We may do this by examining the declarations of the scientists themselves throughout the history
of science. We will find out, indeed, that there are other declared goals besides the goal of truth. For example, the
following goals are very often cited:
< previous page
page_118
next page >
page_119
< previous page
page_119
next page >

Page 119
The goal of explaining and predicting natural events and phenomena and the goal of advancing technology and
mastering nature.
A more radical approach is to look for psychological or social motives which scientists are not aware of. Scientists
declare that they seek comprehensive truth, objectivity, etc., but their real, hidden, motives may be different. For
example, John Ziman suggests (1968) that the goal of scientific research is to arrive at a consensus, rather than
truth; this goal of consensus is what distinguishes science from other human activities. This kind of approach treats
science as a social phenomenon. Scientists are not free to choose their goals; they can only choose to participate in
the process and obey the rules of the game, without being fully aware of its significance. This being the case, the
activity of the individual scientist cannot be judged by a standard of rationality, as if it were a goal-directed
activity. The individual soldier is not always aware of the goals of the army as a whole; for example, he may get an
order to retreat while the army advances.
It is not clear, however, what is the origin of these hypothetical hidden goals. If we refer to a hidden, or
subconscious, motive which is shared by all scientists, we may treat it as a general phenomena and seek a
psychological explanation for it. If we refer to a collective goal of the scientific community, we may seek a
sociological, or socio-psychological, explanation. However, we should perhaps dispense with the notion of goal
altogether. According to the fundamental conception underlying modern natural science, natural phenomena are not
regulated by goals but governed by laws of nature. Hence, if our conception of rationality is based on a naturalistic
view, we would not expect natural science, as a natural phenomenon, to be regulated by goals. In particular,
teleologic explanation is absent from modern evolutionary explanation, which will occupy a central role in the
naturalistic theory which will be presented in the following chapters. The kind of rationality which we would seek
in science in case it is a natural (e.g. social or psychological) or an unintentional phenomenon is naturalistic
rationality. I shall explain the significance of this conception of rationality in section 4.4.
4.1.2 The Dilemma of the Normative Methodologist and Goodman's Solution: Rationality without Goals
The paradigmatic model which guides the normative philosopher of science with the goal of truth in mind, is
deductive logic. Logic yields prescriptive criteria for deductive validity of reasoning. These criteria do not attempt
to describe how people actually reason but how they should reason. When people violate the rules of logic the
logican would say that they are in error. Similarly, criteria of scientific rationality are not intended to be
descriptions of how scientists actually reason, but rather how they should reason.
There arises the question from where does the normative philosopher of
< previous page
page_119
next page >
page_120
< previous page
page_120
next page >

Page 120
science take his rules of rational reasoning, and how does he justify them. If his theory of rationality is derived
from a first philosophy or a priori principles, such as the principles of logical reasoning, then the problem of
possible violations of his recommendations in actual science will arise. He would face serious difficulty if, in the
light of his first philosophy, he recommended abandoning some of the central ingredients of scientific practice.
Such science could turn into philosophically fabricated science; the most celebrated successes of actual science
might not have been achieved had scientists adopted his methodology. Thus, the philosopher who attempts to deal
with real science and not with an ideal system of reasoning must keep an eye on the history of science. On the
other hand, as a normative philosopher, he should justify the methodological rules he prescribes, in the light of the
goals of science. Thus, the normative philosopher of science faces a dilemma: on the one hand, he wishes to
maintain the notion of justification; and on the other hand, to avoid a situation where justified rules of inference are
systematically violated by most scientists most of the time. He has to choose between abandoning his first
philosophy or rejecting most of the celebrated chapters of science as irrational or ''nonscientific." For example, an
empiricist philosopher may adopt the epistemological view that only observation sentences are justifiable. He may
draw from this the methodological rule that only theories which are wholly reducible to observation sentences are
scientific. If he then finds out the Newtonian mechanics or quantum mechanics cannot be wholly reduced to
observation sentences, he must either conclude that modern physics, which is erected upon these theories, is
nonscientific or nonrational, or abandon his first philosophy as a theory of rationality. The first possibility is bad
since it means killing science altogether. The second is worse since a first philosophy is, by definition, irrefutable
by facts.
Nelson Goodman provides us with an escape from this dilemma. With his approach we also avoid the task of
finding out what the goals of science are; the justification of methodological rules is not dependent on any possible
goals of science. This approach would be therefore appropriate for describing science in naturalistic terms. Indeed,
as we shall see, one way to interpret, or to make sense of, his approach is to view the human activity which is
governed by these rules as a natural, or involuntary, phenomenon. Goodman starts with an analysis of justification
of deductive rules:
Principles of deductive inference are justified by their conformity with accepted deductive practice. Their
validity depends on accordance with the particular deductive inferences we actually make and sanction. If a
rule yields inacceptable inferences, we drop it as invalid. Justification of general rules thus derives from
judgements rejecting or accepting particular inferences. ... A rule is amended if it yields an inference we are
unwilling to accept; an
< previous page
page_120
next page >
page_121
< previous page
page_121
next page >

Page 121
inference is rejected if it violates a rule we are unwilling to amend. (Goodman 1965, 6364)
In other words, the justification of rules of inference is not based on a priori principles, but on their accord with
inferential practice. This scheme contains a prescriptive element since an inference may be rejected "if it violates a
rule we are unwilling to amend." However, the reason for this unwillingness is not specified; it is treated as a given
fact. We may describe this sort of accord by saying that a justified rule is in reflective equilibrium with inferential
practice (see Stephen Stich and Richard Nisbett 1980, 190). This term is borrowed from John Rawls (1971, 20). I
may add that this should be a dynamic equilibrium, if we wish to entertain the possibility of changing our attitude
towards rules and particular inferences as our reasoning experience evolves.
Goodman generalizes this view of justification to include inductive reasoning: rules of induction are justified by
their being in reflective equilibrium with inductive practice. We may further generalize this analysis to include all
methodological rules in science; for example, rules of confirmation, refutation, acceptance, rejection and
generation of scientific theories. Methodological rules are rules of inference or rules which guide decisions, e.g.
decisions to accept or reject theories, decisions to perform certain observations or experiments, etc.
This approach to methodological rules does not presuppose any particular goals for science, or that science is a
goal-directed activity at all. Its validity merely derives from its accord with scientific practice. If, however, we
assume that science is goal-directed, then the line of reasoning with respect to the goals of science is reversed here:
when a stable set of rules is found to be in reflective equilibrium with the inferences and decisions of a given
community (e.g. the whole scientific community or a community of scientists engaged in a specific branch of
science) we assume that rational behavior in that community means obeying these rules. This means that rationality
is not universal but is community dependent (a notion which reminds us of the notion of community-specific
logic). Given the rules, we can infer or hypothesize what possible goals the community attributes to science. For
example, if we find that physicists are guided by a methodological rule which requires conducting active
experimentation rather than making only passive observations we may come to the conclusion that one of their
goals is to reproduce and control natural phenomena, or to advance technology. The goal of attaining true
descriptions and explanations of natural phenomena cannot by itself explain why, for example, particle physicists
produce more and more new, short-lived particles at higher and higher energies. As we saw in section 1.4, it seems
that by using this method of active research, physicists create artificial phenomena rather than discover natural
phenomena. We can imagine a science which would not intervene in the natural course of events and would only
be engaged with
< previous page
page_121
next page >
page_122
< previous page
page_122
next page >

Page 122
recording natural phenomena. Indeed, such a passive approach was actually adopted before the emergence of
modern experimental science. Thus, the goal of controlling natural phenomena and advancing technology can be
seen as one of the goals which distinguishes modern natural science from its predecessors. However, as we have
seen, if scientific discovery is a natural or an involuntary phenomenon, we would not treat it as a goal-directed
activity. If, for example, we view science as an evolutionary process, the same pattern of active experimentation
may be explained by an evolutionary mechanism which I call "growth by expansion" (see section 7.3). This
explanation does not attribute any goals to science.
We must remember that we are dealing here with a "second order" justification, i.e. with the justification of
methodological rules which are employed in the procedures of justifying knowledge claims. If we attribute a
normative role to these rules, we have to justify them. However, we are not interested only in rules of justification
or evaluation, we are in particular interested in rules or methods of generating theories or other kinds of knowledge
claims. These rules are not normative in the sense of warranting success or acceptance. They may be normative in
the sense that if we follow them, we have a high chance to generate successful, or acceptable, theories. This is the
sense in which rules of generative induction, for example, are normative. If these normative rules of generational
discovery are not derived from first principles, they may be justified by a procedure such as Goodman's.
In the following I will propose a scheme in which the rules are derived from an established theory of scientific
rationality. This would explain why we sometimes refuse to amend a rule following its clash with inferential
practice. The theory from which the rule is derived makes it immune to refutation by particular inferences which
clash with the theory, and as a result these inferences would be rejected as "invalid."
4.1.3 From Justification to Explication
Goodman's approach to justifcation is closely related to the notion of explication of intuitive rules. However, the
explication of methodological rules is based not only on the empirical data of intuitions and practice. It constitutes
one step towards a theoretical support to the rules. Carnap introduced this notion while attempting to explicate
various concepts of probability and induction (Carnap 1950). By "explication" he meant formalization or
axiomatization. Explication transforms vague concepts used in ordinary or scientific discourse into clearer
concepts. A good explication should achieve a good agreement between the formal system and the intuitive
concepts. Such a process can expose and remove inconsistencies in the use of the intuitive concepts.
Mary Hesse advocates reducing the problem of justification of induction to the problem of explicating intuitive
inductive rules. According to Hesse the
< previous page
page_122
next page >
page_123
< previous page
page_123
next page >

Page 123
process of explication is divided into two tasks: "(i) To formulate a set of rules which capture as far as possible the
implicit rules which govern our inductive behavior. (ii) To formalize these in an economical postulate system"
(Hesse 1974, 97). An example for a rule of induction generated at stage (i) is the rule of enumerative induction,
which can be formulated as follows: "An empirical generalization should be increasingly confirmed by observation
of an increasing number of its positive instances and no negative instance." Another example for a presystematized
rule is the following rule which Hesse subsumes under the category of induction, but which can just as well be
classified into the hypothetico-deductive method: "A hypothesis should be strongly confirmed by the observation
of the truth of one of its logical consequences which was not expected before the hypothesis was proposed." A
formal system most appropriate for explicating such rules of scientific method is probability theory. As was
mentioned in section 2.2.7, in such a system the probability of a theory, for example, explicates its degree of
confirmation.
Let us now compare Hesse's two-stage scheme with Goodman's notion of justification. At first sight we might be
tempted to identify the process of reaching reflective equilibrium with stage (i); that is, the presystematized rules of
inductive inference are perhaps formulated via an interaction with inductive practice, and thus justified. However,
Hesse seeks justification specifically for the postulates of the formal system generated at stage (ii) and not just for
the inductive rules formulated at stage (i): "a sufficient, and perhaps the only possible, justification of a set of
postulates of inductive inference would be that they form a 'good explication' of the intuitive inductive rules.
Justification in this sense resides in the interaction of postulates and rules and not in any external support for the
postulates independently of the rules" (ibid., 98). Goodman refers to two methodological levels: (a) particular
inferences, or inferential practice, and (b) principles or rules of inference, where the latter need justification. Hesse
refers to three levels: (a') implicit inductive rules; (b') an explicit set of rules; and (c') a formalized system, where
the postulates of the latter need justification. The implicit rules of level (a') govern the inferential practice of level
(a). Furthermore, levels (b) and (b') are identical. We can therefore identify stage (i) of the explication process with
the process of reaching reflective equilibrium between rules [level (b)] and practice [level (a)]. This process,
according to Goodman, provides justification to the rules. According to Hesse, however, the justification is shifted
"upward" to the formal system [level (c')] and it is attained through the interaction of the system with the rules.
4.1.4 From Explication to Explanation: Paradigms of Rationality
I shall now propose a scheme of justification which will generalize Hesse's view and will employ some ingredients
from Goodman's approach. The cen-
< previous page
page_123
next page >
page_124
< previous page
page_124
next page >

Page 124
tral role in this scheme will be played by a notion which I call "paradigm of rationality." This notion will
contribute to the normative dimension of the scheme. As we will see, it is akin to one of the uses Thomas Kuhn
makes of his notion of paradigm. The paradigm of rationality replaces the first principles of rationality of the
aprioristic philosophy of science. However, unlike the latter the paradigm of rationality is not eternal, and may
undergo changes with the evolution of science.
The structure of my scheme of justification can be represented by analogy with the structure of theoretical
explanation peculiar to modern natural science. I can describe such a process of explanation as an interplay
between the following layers of scientific knowledge: (a) observational data, (b) empirical generalizations, (c) an
explanatory theory and (d) the general world picture prevailing in science, which is the scientist's general view
about the structure of the world and its fundamental building blocks (e.g. epicycles, particles, forces, fields, etc.).
For example, the kinetic theory of gases [layer (c)] explains the empirical laws describing gas behavior, such as
Boyle's law, Gay-Lussac's law and the ideal gas law [layer (b)], which are the regularities found in the
observational data [layer (a)]. According to the view of explanation which I will adopt here, not every theory
which entails the empirical generalizations explains them; a necessary condition for a theory to be explanatory is
that the theory comply with the world picture. For example, an explanatory theory in nineteenth-century physics
had to comply with the mechanistic-corpuscularian world picture [layer (d)]. The kinetic theory not only entailed
the gas laws, but also explained them, since it complied with that world picture. Furthermore, only empirical
generalizations which are faithful to the observational data are candidates for explanation. In the context of of
scientific explanation this requirement seems trivial; no one would suggest explaining "laws" or "generalizations"
which contradict most of the data. By definition, we explain something which we believe to be true, or
approximately true. An explanatory theory, however, may somewhat correct the original generalizations. For
example, Newtonian theory entailed a corrected version of Kepler's original laws of planetary motion. It therefore
explains a modified version of the laws. Thus, the process of finding an explanation may change the original
explanandum. This would have important implications when we carry over the analogy from the process of
explanation in science to the process of justification in the philosophy of science. Furthermore, according to the
standard methodological practice, the explanatory theory may, or should, predict new laws. In scientific practice if
the modified generalizations or the predicted laws agree with the data, the theory is strongly confirmed.
Our scheme of justification can likewise be divided into four layers respectively: (a) particular scientific inferences
and decisions ("scientific practice"), (b) methodological rules, (c) methodological theory or theory of science,
which may be a formal system (such as Hesse's) but not necessarily, and (d) the
< previous page
page_124
next page >
page_125
< previous page
page_125
next page >

Page 125
paradigm of rationality (POR), which is our general view about the nature of science and scientific rationality.
According to the conception of justification I propose here, a methodological theory, or a theory of science,
justifies the methdological rules which it entails only if it complies with the paradigm of rationality. This is the first
condition for justification. Thus, justification is relativized with respect to the POR. There is no absolute or
universal justification, as there is no final or eternal explanation.
Before I try to characterize in general the notion of POR, I would like to illustrate its meaning by means of some
examples. First, I will consider logicism. This POR dominated the twentieth-century philosophy of science. Its
proponents did not consider it as one possible POR. They treated it as a priori valid. This POR views science as
proceeding by inferences. The goal of science is generating true, or approximately true, statements or to eliminate
false statements. In order to achieve this goal, scientists should obey prescribed rules of inference. This view of
science would require the construction of a formal system analogous to, or an extension of, deductive logic.
Possible methodological theories which comply with this paraidgm are inductive logics, probabilistic confirmation
theories, such as Bayesian theory, or falsificationist methodology. Hesse's scheme may be viewed as such a
methodological theory.
Another example of POR is sociologism, which views science as a social phenomenon. This view may have a
number of versions: e.g. science as a tool for the advancement of society, or of technology. If the goal is advancing
technology, a methodological theory should imply rules of preference for research topics or rules for choosing
between theories according to their technological utility. Sociologism may have a stronger version, referring to the
internal social characteristics of the scientific community as essential to the nature of science.
Finally, I would like to mention evolutionism, i.e science as an evolutionary phenomenon analogous to, or
constituting a continuation of, organic evolution. One possible methodological rule which might be advocated by
this POR is the need for proliferation of hypotheses such that there will be great variability, this being the source of
evolutionary progress. Another possible methodological recommendation derived from this POR pertains to the
manner by which hypotheses should be generated, i.e. independently of the phenomena to be explained or the
problems to be solved ("blind" variation). Finally, rules of elimination and falsification (selection) are indispensible
to this paradigm. Thus, Popperian falsificationism may be viewed as inspired by a POR which combines logicism
and evolutionism. In Chapter 7 I will expound a POR which couples sociologism with evolutionism.
How can we generalize from the above examples in order to characterize a paradigm of rationality? First, it should
be noted that in each of the above cases there is more than one methodological theory, or theory of science, which
complies with, or is implied by, a given POR. Hence, a POR cannot be
< previous page
page_125
next page >
page_126
< previous page
page_126
next page >

Page 126
equated with a methodological theory. Second, each of the above paradigms says something about the nature of
science as a phenomenon. Each sets a general theme for science: "science as an inference machine," "science as a
social phenomenon" or "science as an evolutionary process." Third, since the POR specifies the general nature of
science, it implies general requirements to be met by a methodological theory, i.e. it determines which theories of
science will not be acceptable on first inspection.
The second condition for justification is that the justified methodological rules be faithful to scientific practice, e.g.
that they be in reflective equilibrium with scientific practice. In other words, the notion of justification can be
applied only to rules which more or less accord with scientific practice. This is analogous to the notion of
explanation which can be applied only to empirical laws or generalizations which accord with the observational
data. This condition guarantees that we will not generate justified methodological rules which are not adhered by
most practicing scientists. Philosophers of science who base their methodology on a first philosophy do generate
justified rules which are not obeyed in actual science, hence the above condition is not trivial as is its counterpart
in the scheme of scientific explanation.
The above scheme, however, is normative, although not in an absolute sense. As in Goodman's approach, we can
sometimes refuse to amend a rule which does not accord with a particular scientific inference or decision. Unlike
in Goodman's account, our paradigm-guided scheme gives us a definite reason for refusing to amend a rule when it
is entailed by an established methodological theory which complies with our POR. Therefore, when a particular
inference or decision violates such an entrenched rule, we will reject such an inference or decision. However, a
POR which leads us to rule out some important elements of scientific practice should in the long run be amended
or be replaced by a new one.
Thus, the normativity of this scheme is expressed by the fact that it may lead to the rejection of particular
inferences or decisions, and by the fact that it lends justification to particular inferences or decisions which comply
with the methodological rules. However, this is not an aprioristic justification, since the methodological rules have
themselves been justified by being in reflective equilibrium with scientific practice and by complying with the
POR. This is not a universal justification; it depends on the POR. Furthermore, the system of methodological rules
is dynamic; it may incorporate new rules or drop old ones, according to the changing practice of science, on the
one hand, and the changing image of science and scientific rationality, on the other.
Hence, the standards of rationality are not dictated from outside science by some first philosophy. Rather they are
drawn from both the practice of science and the POR. The POR itself is determined by external and internal
sources. The external sources include epistemological and metaphysical views, the general conception of what
science is and the attitudes of society towards
< previous page
page_126
next page >
page_127
< previous page
page_127
next page >

Page 127
science. The internal sources include our knowledge of the history of science and scientific practice. Thus, the
POR interacts with scientific practice through the methodological rules.
The above scheme applies to justificatory or evaluative methodological rules, such as rules of confirmation,
falsification, acceptance or rejection of hypotheses. However, we may extend the scheme to apply to all kinds of
methodological rules including methods of theory-generation and discovery. As in the case of justification, the
scheme will have both normative and explanatory roles.
4.2 From Description to Explanation
In the previous section I described a philosophy of science the starting point of which is normative or prescriptive
and the impossibility of this approach. I proposed to replace the traditional notion of justification by a relativized
notion which is analogous to the notion of theoretical explanation. I will now consider the implications when the
point of departure is descriptive, and I will arrive at the conclusion that a purely descriptive approach is also
impossible and should be replaced by an explanatory approach. Thus, from both points of viewthe normative, or
prescriptive, and the descriptivewe would be led to the conception of an explanatory philosophy of science which
contains some elements of both approaches.
In dealing with the descriptive approach, we first have to see how it differs from a history of science. A historian
of science cannot be a neutral observer and describe "mere" facts since he has prior expectations and attitudes
towards science and he has initial concepts by which he comprehends the phenomena of science. As Lakatos puts
it: "history of science without philosophy of science is blind" (Lakatos 1971, 91). What is then the difference
between the two, if any? A minimalist historian of science might be distinguished by his intention to be as
"neutral" as possible, i.e. to describe the empirical facts and to avoid using generalizations or theories as far as
possible. On the other hand, the descriptive philosopher of science, being a philosopher or a methodologist, expects
to find in science methods and general characteristics. He expects science to be a rule-governed phenomenon. Such
a position involves an intention to analyze, to generalize or to theorize and not just to remain on the level of
reporting what scientists do. However, since he does not intend to be prescriptive, he would not employ the notion
of justification.
The descriptive philosopher of science will not find any list of explicit and clear methodological rules which guide
scientists in their work. I do not refer here to specific research methods, such as how to prepare a chemical solution
or to solve an equation, but to universal rules, such as the rules of the hypothetico-deductive method, which
constitute the so-called scientific
< previous page
page_127
next page >
page_128
< previous page
page_128
next page >

Page 128
method. Although some general methods of science are occasionally discussed in scientific literature, there is no
general agreement with respect to their clear formulation. Furthermore, scientists do not learn their profession by
studying a methodology. One of the lessons which a graduate student learns when he turns to actual research is
that he has to ignore many of the nice and neat principles and slogans he has learnt during his undergraduate
studies, in particular some of those principles which are supposed to constitute the scientific method.
Even when great scientists mention certain methodological principles, the philosopher of science may find that the
scientists do not actually adhere to them. Perhaps the most conspicuous example of this appears at the very
beginning of modern science. Issac Newton, who declared "hypotheses non fingo" or "I feign no hypotheses,"
created one of the most celebrated hypotheses in the history of science. Newton's theory of universal gravitation
goes far beyond commonsense experience and intuition, and has far-reaching predictions. Newton's use of the term
hypothesis, however, is somewhat different from that of twentieth-century physicists or philosophers of science. In
one case, for example, he uses this term to mean a proposition which refers to "occult qualities" which are not
observable and measurable. This indicates another problem facing the descriptive philosopher of science.
Contemporary examples are abundant. A typical example is that of the theoretical physicist who emphatically
declares that his theories are nothing but an economical means of organizing observational data. The philosopher of
science might point out that such a physicist employs the hypothetico-deductive method, where the theory goes
beyond a mere summary of observed data. Another example is that of the scientist who claims that he is making
observations in order to confirm a theory, but a philosopher (such as Karl Popper) might tell him that his
experiments are actually attempts to refute the theory.
Thus, philosophers of science who view their task as descriptive, face the problem that they cannot take at face
value the declarations of scientists about the scientific method in general, and even about the principles which they
employ in their own research. So perhaps a descriptive philosophy of science should not take very seriously what
scientists say, but rather study how scientists actually do science. For example, the descriptive philosopher of
science should study how scientists construct theories and check them against experimental results, and then the
philosopher might try to generalize from these findings. However, here arises a problem which faces the historian
of science as well: the philosopher of science does not encounter neutral facts when he studies science; already
when he starts his studies he has to choose where to look and how to interpret and categorize what he sees. By the
metascientific terms he employs he tries to capture the data of science. The raw data may include scientific papers
and reports, conference proceedings, letters (scientific products), or perhaps more abstract entities such as theories
and experi-
< previous page
page_128
next page >
page_129
< previous page
page_129
next page >

Page 129
ments. Here there is a parallel between a descriptive philosophy of science and science itself. It is a widely
accepted view that there are no pure observational terms and statements in science; every descriptive statement
employs terms which are loaded with theoretical pressupositions. The same applies to descriptive statements about
science.
The descriptive philosopher of science will naturally try to first use the metascientific terminology employed by the
scientists themselves. As in the case of the scientists' declarations about the scientific method, however, he will
very soon find out that there exists no such unified and consistent terminology. As we have seen from the example
of Newton's use of the term hypothesis, metascientific terms may be interpreted in different ways at different times.
Even in contermporary scientific writings, we do not find any systematic metascientific terminology. Metascientific
terms as used by scientists are frequently ambiguous. The term science itself conveys different meanings to
different scientists. Terms such as "theory" or "model" are used with a variety of meanings. As was indicated in
section 1.2, the term theory, for example, which is central to modern science and which is extensively used by
scientists and philosophers, has a number of possible meanings, some of which are interrelated. A theory might be:
(1) a conjecture, as opposed to a solid factual statement; (2) a system of statements which employs so-called
theoretical terms, i.e. terms which do not appear in the observational vocabulary; (3) an explanatory system, as
opposed to an empirical generalization which does not explain but only describes and summarizes observational
data; (4) a system of laws of nature; (5) an uninterpreted deductive system which is related to observational data
through correspondence rules; or (6) a dynamic system in the sense described in section 1.2. Thus, when we refer
to "Newtonian theory" we might refer to a system of laws or to a system of statements which express a particular
version of the historical entity stretching from the the seventeenth century till the end of the nineteenth century, or
we might refer to the whole historical entity. Furthermore, terms such as "theory" and "model'' are sometimes used
interchangeably to refer to the same entity, e.g. the Bohr atomic model or theory. Methodological terms such as
"proof" and "refutation" are frequently used misleadingly: scientists often claim that a certain theory was proved or
refuted by experiment, whereas it is well known that even if theories are universal statements, they cannot be
logically proved by any finite quantity of observational data. It is further known that logical refutation can be
avoided by making ad hoc modifications of the theory, by reinterpreting or by ignoring the anomalous data, or by
introducing some auxiliary assumptions. Moreover, metascientific terms convey different meanings in different
sciences.
Hence, the descriptive philosopher of science must choose for himself a proper metascientific terminology and a
proper categorization of scientific activity and products. He may use in a more refined manner some of the terms
employed by scientists. He may reintepret other terms and add new
< previous page
page_129
next page >
page_130
< previous page
page_130
next page >

Page 130
ones; the terms research program and paradigm are examples of the latter. The choice of terminology and
categorization will be made in compliance with criteria which guide the philosopher of science, such as fruitfulness
or explanatory power.
In other words, the descriptive philosopher of science is led to play the role of a scientist; he will invent a scientific
theory of scientific method or a scientific theory of science. If the term theory, for example, appears in such a
metascientific theory, it will be a theoretical term.
Thus, descriptive philosophy of science becomes a science of science, or rather a theoretical science of science,
since the philosopher of science is not engaged in performing experiments or making observations; he just uses the
data supplied to him by the historian of science, who therefore plays the role of the data-collector or the
"experimental" scientist of science who makes the observations.
4.3 Explanatory Philosophy of Science
Following the above arguments, we are led from the descriptive view of the philosophy of science to the
explanatory view. However, the explanatory view is inherent also in the scheme of justification I proposed. Indeed,
the scheme was constructed analogously with a scheme of theoretical explanation which is widely practised in
science, especially in the "hard" sciences such as physics. Now that we have arrived at the explanatory view from
the descriptive starting point, it would be natural to consider the possibility of treating the scheme of justification as
a scheme of explanation and thus, converting the structural analogy between justification and theoretical
explanation into a deeper similarity. As a scheme of justification it suffers from the disadvantage that it is
relativized on the POR, whereas the latter should in turn be justified. However, if we treat it, instead, as an
explanatory scheme we would avoid this problem; we would deal with a science of science and in scientific
explanation there is no problem of relativizing the explanation on our world picture, since we do not expect
explanation to be absolute. Moreover, in the next section I will argue that as an explanatory scheme the philosophy
of science still has a normative role, although not in the traditional sense of normativity.
Thus, we may employ the same four-layered scheme which was employed for justification to describe the structure
of the explanatory philosophy of science. We would treat the layer of scientific practice as the layer of
observational data. The layer of methodological rules, which originally were subject to justification by the
methodological theory, will now be treated as the empirical laws to be explained by the theory of science. The
empirical laws, which include the methdological rules, summarize and generalize scientific practice.
< previous page
page_130
next page >
page_131
< previous page
page_131
next page >

Page 131
Notice that I am talking here about the justification of the methodological rules rather than of the scientific practice
and its products, such as laws and theories; the latter are traditionally justified by applying the rules. Similarly, I
am talking about the explanation of empirical generalizations and laws (by deriving them from an explanatory
theory), rather than about the explanation of the observational data, which is a different matter. Thus, the
methodological theory in its new role will become an explanatory theory. Finally, the paradigm of rationality will
turn into the philosopher of science's general outlook on science, i.e. his paradigm of science through which he
views science as phenomenon in the world.
The same four-layered structure can be viewed as a scheme of justification based on a POR or a scheme of
explanation, depending on the meaning attached to the paradigm. If the paradigm is a POR, we are in the realm of
justification; if the paradigm is our general view of science as a phenomenon, we are in the realm of explanation.
The two views are not necessarily contradictory. In fact, these are the two faces of the same view. The dual
structure is exhibited in the following diagram:
In both schemes, every two adjacent layers are adjusted to each other. The methodological rules are adjusted to the
practice and the latter may be modified in view of the former. The methodological theory is tested through its
predictions regarding methodological rules and practice. The POR may change, or even be replaced, in the long
run in view of the difficulties facing the methodological theories inspired by it. A similar interaction takes place in
the explanatory scheme.
I shall illustrate the duality between justification and explanation with respect to logicism. If we treat logicism as a
POR we mean that particular scientific inferences and decisions are justified only if they obey the rules prescribed
by the methodological theory, which is derived from, or modeled on, a logical theory. If we change our attitude
towards logicism and treat it as a scientific paradigm, rather than as a scheme of justification, this means that it is
used as a guide for constructing explanatory theories for the phenomenon
< previous page
page_131
next page >
page_132
< previous page
page_132
next page >

Page 132
called science and for scientific practice. In this capacity, logicism may be a general psychological or
psychosociological paradigm on the nature of scientific knowledge, and a general view of how scientists reason in
fact, rather than a normative view of how they should reason or act. Thus, explanatory epistemology becomes part
of psychology. The task of the epistemologist qua scientist is to propose hypotheses as to exactly what the rules of
inference are. A theory which explains and describes the reasoning and action of scientists is an empirical theory
which is testable and refutable. Violations of the empirical rules of inference will be treated as problems to be
solved or anomalies to be explained.
Hence, logicism as a POR leads to prescriptions, whereas logicism as a scientific paradigm guides explanations.
However, the distinction between the two attitudes in not as sharp as in the case where logicism is an aprioristic
scheme. In my approach the methodological rules draw their justification in part from their interaction with actual
inferential practice. Moreover, logicism as a POR is fed back indirectly by inferential practice through its
interaction with the methodological rule.
The switch from one attitude to the other is therefore not so drastic. When we are in the justificatory mode of the
system and we face a situation where the methodological rules are violated in many cases, we have the option to
modify or replace our methodological theory. For instance, we may replace an inductive theory by a Bayesian
theory of confirmation, or a naive falsificationism by a sophisticated one (Lakatos 1970). If we cannot find a
satisfactory logicist theory, however, we may look for another POR, but this weakens our normative stand
significantly. Indeed, in our search for a new POR we aim at adopting a POR which will not clash too much with
scientific practice. In other words, if we are willing to replace our POR in order to avoid the dilemma of the
normative methodologist, we will be dragged by inferential practice of scientists, rather than impose our
prescriptions on the latter. By this we switch from the justificatory mode to the explanatory mode. In order to find
out that a certain POR leads to a successful theory of sciencei.e. a theory which does not clash too much with
scientific practicewe must act as scientists rather than as normative methodologists. When such a POR is highly
established through the theories it inspires, we might switch back to the justificatory mode and recommend that
scientists employ the rules which are derived from the theory, knowing the limitations imposed on the normative
strength of such recommendations. In the next section I will attempt to clarify this sense of normativity.
If we treat our theory of science as an explanatory theory, it should explain scientists' inferences, decisions and
acts, concerning the generation and evaluation of hypotheses. It may explain, for example, why scientists generate,
accept or reject a certain theory in a given situationa situation which might be characterized epistemologically,
sociologically or psychologically.
< previous page
page_132
next page >
page_133
< previous page
page_133
next page >

Page 133
The intermingling of epistemology with psychology in the last statement is natural, since epistemology is regarded
here as naturalized. The explanatory approach treats science as a natural phenomenon to be explained, rather than
justified in the traditional manner. This is an extension of Quine's naturalized epistemology (Quine 1969).
Quine did not distinguish between epistemology as a theory of ordinary human knowledge and the epistemology
of science. He therefore treated epistemology, including the epistemology of science, as part of psychology.
According to my approach, the naturalistic philosophy or epistemology of science is not exclusively a psychology
of science. Depending on our paradigm of rationality, the theory of science might attribute to the phenomenon of
science social or evolutionary dimensions which have an essential epistemic role. Moreover, not every possible
POR would view science as a purely epistemic phenomenon. For example, a technologically oriented POR might
view science as a tool for advancing practical human needs etc. Another possible POR might view science as a
broadly sociocultural phenomenon, which has non-epistemic sociocultrual dimensions.
What are the criteria for chosing or for evaluating a POR? We start with some initial criteria which we do not
attempt to justify, as scientists do. These are intuitive criteria of commonsense rationality, i.e. the minimal criteria
which we use in everyday reasoning, and which are shared by all people with whom we can communicate. These
criteria will be included in the common hard core of all possible PORs. Yet these criteria will not be immune to
non-radical revision, in the light of the developing POR. We set out to investigate scientific rationality, taking for
granted this commonsense rationality. For example, one of these criteria will be empirical adequacy; other things
being equal, we would chose an initial POR which generates theories of science which are more faithful to the
facts than their rivals.
In summary, the above justificatory-explanatory scheme draws its normative strength from the fact that it is not
derived exclusively from scientific practice and the intuitions of scientists, rather it is derived from an additional
source: the POR and the theory of science. The POR is partially independent of our knowledge of scientific
practice. It is fed by our conception of rationality and by our general beliefs on human nature and on the epistemic
relations between us and the world. Of course, these beliefs are partially influenced by scientific conceptions and
theories. And these are produced by the very scientific practice which is under evaluation. To this extent, the
normative strength of this scheme is weaker than that of an aprioristic scheme. But the above scheme is nonfoundational. It is unavoidably coherentist. Yet, as a coherentist scheme of justification it is stronger than a semidescriptive scheme which
< previous page
page_133
next page >
page_134
< previous page
page_134
next page >

Page 134
relies on scientific practice only. Thus, justification is not absolute, but it draws it strength from all relevant sources
of human knowledge and beliefs. As James Cushing expresess it: "Scientific knowledge, as well as knowledge
about science, must be bootstrapped, starting from what may seem a plausible position, but always keeping the
open possibility (often the likelihood) of fundamental revisions" (Cushing 1990, 240).
4.4 Normative Naturalism: Shallow vs. Deep Theories of Scientific Rationality
4.4.1 Phenomenological Theories of Rationality
In this section I would like to shed some more light on the difference between a semi-descriptive theory of
scientific rationality, which relies on, or is adjusted to, the practice of science, and a theory of the kind discussed in
the last section, which seeks a theoretical explanation for the scientific practice. This will help us in understanding
the role of a naturalistic theory of science in explaining the processes or the phenomena of discovery.
Descriptive or intuitionistic meta-theories of scientific rationality do not rise above the phenomenological level of
methodological practice or normative intuitions about particular instances of reasoning. According to the
descriptive or empirically oriented approaches the task of the philosopher of science is restricted to recording,
describing, and at best systematizing the inferential or methodological practice of scientists. The intuitionistic
approaches derive their theories from scientists' or philosophers' intuitions about scientific reasoning. If on this
basis alone one attempts to draw normative or prescriptive conclusions, it seems that the is-ought fallacy is
commited. However, even if such a theory does not attempt to be normative, it does not add much to our deep
understanding of scientific rationality and the nature of science in general. In this section I will show that my
explanatory approach avoids the is-ought fallacy and attempts to add a deeper level to our understanding of
scientific rationality, although it retains the empirical or naturalistic character of the theory of rationality. It is the
deeper level which makes the theory normative, to the extent that a non-foundationist theory may be normative.
I find it instructive to present my approach here as an alternative to the intuitionistic approach which was proposed
by Laudan (1986), although it has been already abandoned by its author. I also find it very illuminating to compare
my scheme with L. Jonathan Cohen's account (1981) of normative theories of reasoning which are based on the
data of human intuition.
"Meta-methodological intuitionism," according to Laudan (1986), is the view that "we or scientists usually make
reliable and trustworthy judg-
< previous page
page_134
next page >
page_135
< previous page
page_135
next page >

Page 135
ments about methodological matters (such judgments make our shared 'intuitions'), but that our explicit theories
about such matters are not usually so reliable, presumably because we have yet to develop a methodological theory
which does justice to our presumably sound intuitions. If one is an intuitionist, one believes that one should choose
between rival theories of methodology by asking how well they square with (at list some of) our shared intuitions"
(Laudan 1986, 120). A methodological theory, according to this view, is normative, since it is based only on
exemplary cases; the methodology that explicates the norms behind these cases is applied prescriptively to cases
about which we do not have shared pre-analytic intuitions. The "bedrock" or the "database" on which the
intuitionist bases his judgments consists of paradigmatic cases taken from the history of science. The paradigmatic
cases are distinguished by the fact that we possess shared intuitions about their rationality or irrationality.
Laudan places under the intuitionist roof Reichenbach's method of rational reconstruction and Carnap's theory of
explication. As we saw, the later conception has been utilized in constructing inductive logic or probabilistic
theories that explicate our pre-analytic notions of probability, evidential support or confirmation. Goodman's
approach can also be subsumed under this category. All these approaches deal with the context of justification.
However, since in the naturalistic approach there is no clear distinction between theories of justification and
theories of discovery and generation, we may include the latter in the intuitionist approach to methodology. An
intuitionist theory of discovery will rely on the scientists' or the philosophers' intuitions about methods and
procedures of arriving at theories. It will seek, for example, to describe, explicate or reconstruct the implicit rules
which guide scientists in arriving at reasonable theories which have chances to be successful.
The question which immediately arises is whose intuitions are we talking about? Laudan addresses this question
and his answer is that we should rely on our current intuitions, rather than on the intuitions or judgments of the
scientists involved in the exemplary historical cases, e.g. the judgments of the scientific elite. He refers to "our
intuitive and implicit modes of ampliative reasoning," i.e. to the intuitions of philosophers or methodologists or
contemporary scientists; actually to the intuitions of all "sensitive readers of the historical record." Thus, according
to this view, anyone can have intuitions about scientific reasoning, just as one has intuitions about ordinary
inductive reasoning, provided he is familiar with the subject matter (i.e., in the case of scientific reasoningwith the
history of science). As Laudan himself concedes, this metamethodological theory, as it stands, cannot assign to
methodology any significant critical role. The criterion for choosing those people on whose intuitions we would
base our theory is not justified, and cannot be justified in a non-circular manner.
< previous page
page_135
next page >
page_136
< previous page
page_136
next page >

Page 136
However, the requirement of familiarity with the subject matter may lead us to some useful insights. Familiarity
with the subject matter is required for every ampliative reasoning or judgment which cannot be reduced to the
application of a content-neutral method, in contradistinction with intuitions about deductive reasoning which do
not require familiarity with any particular subject matter. Indeed, in some cases of inductive reasoning, one has to
be familiar with the subject matter in order to be able to identify the "natural kinds" in the field in question, on
which inductive projections can be made. In ordinary experience the ability of identifying the natural kinds is in
most cases genetically and culturally based. However, in science the "natural kinds" or the theoretical categories
are frequently remote from those of ordinary experience so that scientific inferential practice, in particular, requires
familiarity with the subject matter. Actually, only active scientists in the field may have the intuitive grasp of the
modes of inference and the ways of learning from experience in the field. We are therefore led to the view that
discerning power is essential for scientific reasoning. This is another feature which is common to the generation
and the evaluation of hypotheses.
Indeed, if we entertain the view that scientific inferential practice exceeds ordinary ampliative reasoning and that
there are methodological and inferential rules and norms which are peculiar to science or to a particular field of
science, we must refer to the shared intuitions of the active scientists, or the leading scientists, in the relevant
scientific community. Such a view of science may follow from the evolutionary view of science which is presented
in the following chapters. One of the main conclusions which will be reached there is that our inborn and culturally
determined intuitions are not appropriate for comprehending the phenomena investigated by modern science, since
these phenomena are radically different from those prevailing in our natural habitat. We cannot therefore expect
non-scientists to have reliable intuitions about the rationality of genuine scientific practice; at most some of these
intuitions might infiltrate society at large. Thus, scientific practice is guided by extra-logical methodological rules
and heuristics, some of which belong to the tacit knowledge shared only by the members of the scientific
community. This is a community-specific world view and logic. The proper way to gain the shared intuitions about
scientific inferential or methodological practice is to undergo the lengthy process of scientific training. For
example, we cannot expect that the sense of mathematical elegance or simplicity which leads physicists to accept
certain theories and to reject others would be shared by all philosophers or historians of science.
Stich and Nisbett (1980) recommend to amend the intuitionist method of reflective equilibrium, such that the
intuitions of experts in logic should be consulted, rather than the intuitions of laymen, since non-experts sometimes
make logical errors. In the case of deductive inference this means consulting the intuitions of professional
logicians. However, the logician does not have a
< previous page
page_136
next page >
page_137
< previous page
page_137
next page >

Page 137
privileged access to ordinary human reasoning in the manner an active scientist has to the inferential practice of
science. The logician studies human reasoning but does not have prefered intuitions about human reasoning. With
respect to science, it is the philosopher of science who plays the analogous role to that of the logician. The
professional scientist is not a professional with respect to scientific inference or method. He carries out his
"inferences" quite intuitively, being in many cases unaware of any underlying rules of inference, or that he is
making an inference at all. As I have noted, training in scientific method or inference (as opposed to specific
methods of scientific research) is only optional for scientists and in general it does not help them much in their
scientific practice; an explicit knowledge of scientific method is neither a necessary nor a sufficient condition for
being a good scientist. Furthermore, intuitions on which we should rely in the process of arriving at a reflective
equilibrium, as L. Jonathan Cohen rightly claims (Cohen, L. J. 1981), are "immediate and untutored inclination[s]"
to make some judgments (my emphasis). And "in order to avoid an obvious risk of bias, these must always be the
intuitions of those who are not theorists themselves.'' When we refer to science, we can learn about the "immediate
and untutored" inclinations of scientists by watching them at work, when they do things naturally without
necessarilly being fully aware of the methods they follow.
If the intuitions and the discerning power which guide scientists in evaluating hypotheses are field-dependent and
community-dependent then, a fortiori, the intuitions and the discerning power which guide scientists is discovering
and generating hypotheses and theories that are so dependent. If we do not believe in a universal recipe for
generating and discovering theories, the philosopher cannot rely on his own intuitions; he should consult the
intuitions of the scientists in seeking the rules or the methods of discovery and generation.
Thus, the task of the intuitionist philosopher or methodologist of science is to try and find and explicate the
intuitive rules and heuristics which guide scientists, rather than to rely on his own judgment. The fact that all or
most contemporary philosophers and methodologists share judgments such as the one which maintains that "it was
rational to accept Newtonian mechanics and to reject Aristotelian mechanics by, say, 1800," or that "it was
irrational after 1920 to believe that the chemical atom had no parts" (Laudan 1977, 160) does not reflect an
"innate" intuition about scientific rationality. Rather it reflects a view which is entrenched in our society. Certainly,
judgments of this origin cannot be treated as a "bedrock" of any sound meta-methodology. Hence, an intuitionist
who believes that the intuitions about scientific rationality are not inborn or shared by all humans brought up in our
general cultural settingin the sense that our pre-analytic logical intuitions aremust attend to the intuitions shared by
active scientists. But then it seems that we accept scientific rationality as a matter of fact in science and we
< previous page
page_137
next page >
page_138
< previous page
page_138
next page >

Page 138
license all scientists' intuitions and acts as rational. A theory of rationality such as this will be vacuous, devoid of
any critical or normative force. In fact, any naturalistic-descriptive theory seems to suffer from this sort of vacuity.
However, as I will argue, the explanatory scheme of scientific rationality that I proposed in the last section is a
prescriptive, non-vacuous naturalistic theory of rationality.
4.4.2 Explanatory Theories of Rationality: How the Is-Ought Fallacy is Avoided
The main difference between the explanatory approach and the intuitionistic approaches is that the latter give an
account of scientific rationality and scientific method only on the phenomenological level of intuitions and
practice, whereas my approach seeks a deeper theory which will explain the intuitive rules. Only when we abandon
the attitude of relying on our own intuitions and go to the field and look for the intuitions and practice of scientists
as our empirical data, is the way open for us to "switch" to theoretical explanation of science and its method. We
may arrive at the conclusion that the theory of science or the theory of scientific rationality does not necessarily
consist only of a systematized list of rules. Such a theory may refer to deeper ontology, whereas the
phenomenological level of methodological rules is derived from the theory as the empirical laws of gas behavior
are derived from and explained by the kinetic or molecular theory of gases. I would like to shed some more light on
the question of normativity of the explanatory scheme, i.e. in what sense it can be viewed as a normative or
justificatory scheme. As a first step we will see that the scheme satisfies a necessary condition for a normative
scheme, i.e. that the is-ought fallacy is not commited.
When we view the theory of science as an explanatory scientific theory, we can give four reasons for why we do
not commit the is-ought fallacy in our approach.
(a) Already when we describe the facts of scientific practice, we do not give an unbiased or objective description
of facts, since we describe the facts within our conceptual and presuppositional systems. Indeed, one may observe
scientists dealing with theories or research programs or just recording and organizing observational data, while
referring to the same activity. One may observe scientists trying to confirm their theories, whereas another observer
may see the same scientists trying only to refute their theories. All depends on the fundamental concepts we
employ and on our initial point of view and presuppositions. Our point of view will determine the kind of rules we
will find in the methodological practice of science. It will not help us attend to what the scientists themselves say;
they may say they are doing one thing but we may interpret their actions as doing something else. Consequently,
some of our norma-
< previous page
page_138
next page >
page_139
< previous page
page_139
next page >

Page 139
tive intuitions are already woven into our description of the "is"; we do not infer them from the "is."
(b) Even when we have one uncontestable and stable "observational vocabulary" for describing the facts of
scientific practice, our explanatory theories about science employ theoretical concepts not appearing in the
description of the facts. For example, if we adopt a logicist theory, it may employ concepts such as "inference," or
"prior probability," which does not necessarily appear in the observational vocabulary. If we follow an
evolutionary theory of science, we might employ concepts such as "adaptation,'' "selection," "environment" and
"variation" and so on. Hence, the "ought" is cauched in entirely different vocabulary than the "is."
(c) The way we arrive at our theory of science is not necessarily "from the facts," whatever this might mean. We
may guess the theory, we may construct it by making an analogy with another theory from another field, we may
be guided in constructing the theory by our general world picture or we may arrive at it unintentionally.
(d) Even if we construct the theory of science "from" certain observed facts, the process of confirming the theory
(provided our methodology entertains the notion of confirmation) may require independent confirming facts; we
may require that the theory will successfully predict new (perhaps unexpected) facts and phenomena or previously
unknown methodological patterns or rules.
These four points, and in particular the last three which are not available to the shallow intuitionist approach,
would guarantee that in constructing the theory of science and the methodological rules, we will not possibly
commit the is-ought fallacy. This is a necessary condition for a prescriptive theory; we do not prescribe to
scientists what we observe them doing anyway.
4.4.3 Ideal Theories of Rationality and the Competence-Performance Distinction
If the task of our theory of rationality is to explain the intuitions and acts of scientists, it means that we do not
impose norms of rationality from outside science, rather we accept scientific rationality as a matter of fact. This
does not mean that we license as rational all that scientists do. When we introduce the deeper level of theoretical
explanation based on our POR and theory of science, not everything every scientist is doing is licensed. Scientific
practice supplies us with disconfirming or confirming instances for our theory of science. However, whenever the
theory becomes highly confirmed or established, it is the theory rather than scientific practice which serves as our
(provisional) court of appeal for judging whether or not a decision or act of a scientist can
< previous page
page_139
next page >
page_140
< previous page
page_140
next page >

Page 140
be licensed or not. The criterion is whether or not what the scientist is doing complies with what the theory would
entail. Hence, although we accept scientific rationality as a matter of fact, our theory becomes a prescriptive theory.
In order to comprehend the sort of "normativity" implied by such a view, it would be very illuminating to compare
this view with L. Jonathan Cohen's account of normative theories of reasoning. Cohen treats a normative theory of
rationality analogously to an ideal theory in physics, such as the kinetic theory of gases, which applies only to
perfect gases in isolated systems. Real gases may be treated by such a theory only when additional factors are taken
into account. Another example is the law of free fall in classical mechanics, which applies only to the ideal,
frictionless case. This is the price paid for having a theory with a high degree of generality and comprehensiveness.
So is the case with normative hypotheses. Logical theories are abstract or ideal theories since they ignore spatial,
temporal and causal effects. Actual human reasonings can be accounted for by such a theory only if the extralogical factors are taken into account. Thus, the normative theory is an ideal theory which serves as an essential
premise for explaining and predicting actual human judgments and reasonings. But it is by no means the sole
determinant factor in human reasoning.
According to our dual scheme of scientific rationality, the theory of science serves both as a prescriptive theory of
rationality and as an explanatory or empirical theory. Cohen, too, ascribes a similar dual role to the ideal theory of
human reasoning. As an empirical theory, it describes a cognitive competence of human beings to form intuitive
judgments about deductive or probabilistic inferences. This cognitive competence is a "culturally or genetically
inherent ability which, under ideal conditions, every member of the community would exercise in appropriate
circumstances. It states what people can do, rather than what they will do..." (Cohen 1981, 321). This allows for
people to make inferential errors, but these will be performance errors. Thus, the explanatory approach to
justification of human reasoning can be comprehended in terms of the competence-performance distinction. The
theory of rationality postulates the existence of a genetically and culturally based cognitive capacity in human
beings which in ideal cases will guarantee rational inferential behavior. In real situations, which are not isolated
from the causal influence of other factors, this cognitive capacity will contribute, together with the other factors, to
the inferential behavior. If we can satisfactorily explain the data of inferential practice by the (ideal) theory of
rationality together with theories about the other factors, the theory of rationality will be confirmed. A similar
argument was proposed by Jerry Fodor (1981, 1201).
Analogously, in my scheme of justification the theory of science describes something like a "competence" of
scientists for generating successful ideas and for making rational judgments. This competence is not inborn or
innate, rather it is acquired. In the following chapters I will propose a theory of science
< previous page
page_140
next page >
page_141
< previous page
page_141
next page >

Page 141
according to which judgments in science are not related only to a competence of an individual. Judgments are
related also to the collective wisdom of the scientific community which has evolved with science. This theory
attributes an essential epistemic significance to this collective wisdom. Another possible theory of science, such as
a psychologically oriented theory, might look for the competence in the individual scientist. In any case, the task of
the philosopher of science is to put forward hypotheses about this ideal rationality and about the "external" factors
which motivate scientists' behavior and judgments and which may cause performance "errors." If the philosopher
of science finds a theory of science which together with the external factors successfully explains and predicts the
intuitions and behavior of scientists, his theory is confirmed and he is in a position to criticize improper judgments
or behavior. He may further identify in specific cases disturbing factors which, when removed or avoided, will
clear the way to rational behavior. The disturbing factors may be, for example of psychological, social, economic
or political origin.
4.4.4 The Therapist Model of Rationality and Its Implications for Involuntary Processes of Discovery
The best way to clarify the prescriptive function of the philosopher of science according to the above approach is
to compare it with the function of a therapist in guiding people in overcoming mental or physical problems. Let us
take the example of breathing. A normal human being has the inborn "competence" for breathing correctly.
Breathing is a physiological function regulated by the parasympathetic nervous system. Thus, correct breathing is
an involuntary activity. When there are no breathing problems, a person is in general not at all aware of this
activity. Only when problems arise, might one be helped by the guidance of a therapist. The physiologist or the
medical researcher does not learn the chracteristics of correct breathing only by watching how people breathe. His
knowledge of the functions of breathing derives from his theory of human physiology. In constructing the theory,
actual breathing practice will be only part of the data at his disposal. He will not prescribe to us how to breathe
correctly merely on the basis of his observations of how people breathe. Thus, he does not commit the is-ought
fallacy. If he would stay only on the phenomenological level and draw his knowledge about breathing only from
observing how humans breathe in fact, he would not be able to distinguish between "correct" and "incorrect''
breathing. The medical researcher who is equipped with a physiological theory will be able to characterize correct
breathing and to identify factors which cause people to breathe incorrectly. He may recommend them, for example,
to improve their posture, to perform proper exercises, to change their life style and so on. He may also teach them
how to breathe correctly. The correctness of breathing is thus a functional attribute, related to the function of
breathing in the organism.
< previous page
page_141
next page >
page_142
< previous page
page_142
next page >

Page 142
The above metaphor refers to involuntary processes. The competence to reason correctly is treated as an
involuntary faculty. The competence to make discoveries, a fortiori, can be treated so. As I have indicated, some of
the major kinds of discovery processes are involuntary: e.g., the process of incubation and the process of
generating a discovery by cooperation. So, the above conception of rationality is naturally applied to these kinds of
processes.
Thus, the philosopher of science who bases his theory of rationality only on the data of intuitions and the
methodological practice of scientists can be likened to the therapist who bases his knowledge and
recommendations only on the data of actual breathing. The moral of the above story is that the competence for
making proper judgments in science is unintentional or "innate" in the sense that the scientist is unaware of it.
The exact manner by which scientists acquire this competence and how it is expressed in them depends on our
theory of science or scientific rationality. For example, if our POR is "sociologism," our theory of science might
claim that the individual scientist is compelled to make the right judgments whenever he is integrated with the
scientific community and obeys its norms of behavior. The institutions and modes of behavior of the scientific
community, such as education, imitation, criticism, cooperation, communication and publication systems, would
keep the individual scientist on the right track. The competence for making right judgments is a collective
competence, built in the social system of science. The social structure and dynamics of the scientific community
regulates processes such as theory generation, acceptance or rejection, analogously to the manner by which the
parasympathetic nervous system regulates breathing in the human organism. If our POR is "evolutionism," we
might hypothesize, for example, that scientists are (unknowingly) engaged in the selection of "blindly'' generated
ideas and theories.
The above kinds of theories of scientific rationality do not yield a method of discovery. Instead, the discoverer may
derive from them suggestions and recommedations for "preparing" his mind. The notion of rationality refers here to
involuntary and unintentional acts. Alternatively, we may dispense with the notion of scientific rationality
altogether. In this case, we would maintain that the notion of rationality is not applicable to involuntary or natural
processes, including science.
<><><><><><><><><><><><>
To conclude our general discussion of the naturalistic approach, we may confront the two movements: naturalism
and mechanism. They seem to be diametrically opposed. Mechanical lung is no real replacement for natural lung.
The former would always be inferior in performing the physiological function of breathing. By the same token, the
naturalistic and involuntary aspects of discovery are not likely to be fully mechanized. There is a shortcom-
< previous page
page_142
next page >
page_143
< previous page
page_143
next page >

Page 143
ing inherent in mechanized discovery in virtue of its technological character. Mechanized discovery is a
technology which purports to solve problems, to discover regularities and perhaps also to generate theories. If a
process of discovery is an involuntary phenomenon, or, in a very broad sense, a natural phenomenon, a
technological device might, at most, imitate or simulate it. However, the replacement of natural phenomena by
technology would produce discoveries which will suffer from all the shortcomings of synthetic products. It is very
doubtful whether synthetic science would have produced Newtonian and quantum mechanics and the theories of
evolution and relativity, as it is very doubtful that technology would produce great symphonies. If science had
goals and if we knew what these goals were, we could replace it by technology. And technology might do a better
job than human scientists. However, in treating science and scientific discovery as an involuntary or a natural
phenomenon, we rule out the possibility of the existence of any such goals. To pursue the analogy with art,
technology plays an essential role, in paintings or in sculpture and in particular in music. But technology only
supplies the tools and the materials. It might open new possibilities for creativity. But it cannot replace the creative
artist. In the same fashion, the computer cannot replace the creative scientist or creative processes of discovery,
although it might serve as a very useful tool in the hands of the discoverer and might open new directions for the
progress of science.
< previous page
page_143
next page >
page_145
< previous page
page_145
next page >

Page 145
Chapter 5
An Evolutionary Theory Of Discovery: In Search For The Unexpected
5.1 Evolutionary Epistemology: Taking Natural Selection Seriously
Evolutionary epistemology (EE) is perhaps the best known example of a naturalistic philosophy of science, where
the theory of evolution is brought to bear upon the philosophy of science. Originally, EE started as a theory about
the nature of our cognitive apparatus. Konrad Lorenz (1941) was one of the pioneers in this approach. Only later,
new approaches by Stephen Toulmin (1972), Karl Popper (1972), Donald Campbell (1974b) and others referred
also to the evolution of scientific knowledge.
The basic idea behind this approach is that all biological evolution is an evolution of knowledge. Here knowledge
is not construed as "justified true belief," as traditional justificationist epistemologies would have it. Rather, it is
conjectural knowledge in a Popperian sense, according to which knowledge is always conjectural. Information
about the environment is reflected in the anatomy, physiology and behavior of organisms which have to adapt to
that environment. This is an endosomatic knowledge which is encoded in the form of genetic information and
which is a product of the phylogenetic history of the species. An evolutionary change which brings about a better
adaptation to a given environment means, therefore, growth of endosomatic knowledge about that environment.
The basic assumption behind this view is that the growth of endosomatic and exosomatic (non-organic) forms of
knowledge is governed by the same basic rules. A stronger claim is that the two processes are different phases of
the same phenomenon. This view is summarized by Kai Hahlweg and C. A. Hooker (1989, 23) as follows:
"Knowledge development is a direct extension of evolutionary development, and the dynamics of the two
processes are identical." If the rules of natural selection apply to organic evolution, i.e. to the
< previous page
page_145
next page >
page_146
< previous page
page_146
next page >

Page 146
growth of endosomatic knowledge, they should, therefore, apply also to exosomatic knowledge, such as science
and technology. New mutant organisms, which are to be selected by the environment, fulfill the same function as
new hypotheses or new conjectures to be tested and selected as providing the best explanations for the
observational data. The model of natural selection requires that the new variations (mutations and recombinations)
will be quasi-random, i.e. that they will occur independently of their eventual contribution to the needs of the
organism and their survival value, or without any correlation to environmental pressures. Hence, the growth of
exosomatic knowledge should take place via so-called blind conjectures. This means that the generation of new
conjectures or hypotheses will not be influenced by the "pressure" of the problems they ought to solve or the data
they are supposed to explain. This is an epistemological claim.
Thus, here we draw epistemological conclusions from a scientific theory. This intermingling of epistemology,
traditionally a "pure" philosophical discipline, and science is characteristic of the naturalistic movement in the
philosophy of science, which allows epistemology and the philosophy of science to benefit from the results of
science, or be treated as part of natural science.
Popper reached the same epistemological conclusion, about the "blindness" of scientific conjectures, from the
vantage point of his theory of knowledge. According to his theory, scientific knowledge grows by chains of
conjectures and refutations. Scientists put forward conjectures to tests. Those conjectures which withstand severe
tests temporarily survive, whereas the others are rejected. Scientists do not arrive at their hypotheses by collecting
facts and generalizing from them, or by inferring their hypotheses from the data in some other ways. There is no
logically valid way to do this. For example, any number of observed white swans would not give us logical proof
for the conjecture "all swans are white." The only thing we can infer from observational data, with logical validity,
is that our conjecture is false. This happens when we observe a counterexample, such as a black swan, which
would refute the above generalization. Thus, in science and in everyday life, we arrive at our conjectures blindly
since we have no other choice; no conjecture can be validly inferred from the data.
I would like to add a hypothetical remark. The line of reasoning presented before can be reversed. We have started
by drawing epistemological conclusions from a scientific theory. Now an epistemological theory may have
implications for science. Starting with the Popperian theory of knowledge and assuming that biological evolution is
an evolution of knowledge, we may conclude that the method of learning from experience through blind
conjectures and refutations, which applies to exosomatic knowledge, should apply also to endosomatic knowledge.
In other words, as it turns out, an epistemological theory has an implication for empirical science, biology: organic
evolution "should" obey the rules of natural selection. If the first approach is labeled
< previous page
page_146
next page >
page_147
< previous page
page_147
next page >

Page 147
"naturalizing epistemology," the second, hypothetical, argument could be termed "epistemologizing biology."
Evolutionary epistemology deals with both the evolution of our inborn cognitive capacitiesthe "hardwired" basis of
human knowledgeand with the evolution of human knowledgethe "software." When dealing with the genetically
based ''hardwired" capacities, the evolutionary model is taken seriously, since we assume that the genetically based
capacities are functions of the genetic "hardware," which is an evolutionary product. When we deal with the
implications of EE for the evolution of human knowledge, and if we treat EE as a descriptive or as an explanatory
scheme, it becomes a scientific theory of sociocultural evolution. However, there are approaches which treat the
natural selection model just as a metaphor, or as an analogy, for the growth of knowledge. If we remain on this
level, there is neither a prescriptive nor an explanatory value to the model; we would hardly gain any new insight
when we state, for example, that ideas or theories compete just as organisms struggle for survival or as different
species compete for the resources of some niche. However, if we treat the evolutionary theory of knowledge as a
scientific theory, the metaphor will eventually become a realistic scientific model, transcending its literary value.
When we take evolution more seriously and treat EE as an explanatory theory, we may adopt either of the
following two views. We may treat the growth of scientific knowledge as literally a continuation of the
evolutionary process of which organic and cultural evolution are parts. Or we may treat science as a vicarious
evolutionary process which serves as a tool for the survival of humankind. Of course, the two views need not be
totally exclusive; they may partially overlap. Thus, taking the natural selection model seriously means that we are
ontologically committed to this model, namely we do not treat the natural selection model just as a useful tool for
describing the growth of knowledge; rather we consider natural selection to be a true feature of the phenomenon of
human knowledge.
In adopting the evolutionary, or the natural-selection paradigm of rationality, and an evolutionary theory of
science, we do not "accept" evolutionary theory, or appraise it as better than its rivals. Our general beliefs and
background knowledge, including scientific knowledge, gives the evolutionist POR some initial plausibility. But its
merits as a POR will also be judged according to its success in explaining science. The criteria of explanatory
success are those hard-core criteria for theory appraisal mentioned in the last chapter. It is a combination of both
the initial plausibility and the consequentialist criteria of confirmation, fruitfulness, etc., which will determine its
fate.
Natural selection can be viewed as a universal explanatory paradigm referring to a general pattern of change
prevailing in complex systems. It applies to processes in which a higher order is created, new forms emerge and
new information is gained. The growth of knowledge and the accompanying
< previous page
page_147
next page >
page_148
< previous page
page_148
next page >

Page 148
technological progress and social changes indeed yield increments of information and the emergence of a new
conceptual order and new social forms.
A process which proceeds according to the model of natural selection must exhibit three main features: (1) a
mechanism for generating blind or quasi-random variations, (2) a mechanism for eliminating unfit variations, (3) a
mechanism for propagating the surviving variations. In science, variations are ideas or hypotheses.
5.2 Blind Variation: The Principle of Serendipity
Point (1), which has direct implications for discovery, has evoked the most serious objections to the natural
selection model for the growth of scientific knowledge (see, for example, Thagard 1980). Although, as Popper
argues, scientific theories cannot be logically derived from observations, the generation of theories seems to be by
no means blind to the observational data. When scientists generate theories for explaining some phenomena, they
seem to be guided by methods and beliefs, although the latter do not yield logical validity. Yet, as will be argued in
this chapter, although scientists are guided by methods and established theories, many creative leaps and
breakthroughs in science result from serendipitous processes which represent the blind discoveries in science
(Kantorovich and Ne'eman 1989). And, according to the theory which will be discussed in the next chapter, even
those discoveries which do not seem to be serendipitous include subconscious stages of blind variation and
selection. Thus, Einstein's discovery of special relativity, which seems to be a typical intentional process, probably
included such stages. Einstein himself, drawing upon his own experience, points at infraconscious stages in the
creative process.
5.2.1 Are Scientific Discoveries Analogous to Blind Mutations?
Karl Popper and Donald Campbell relate the requirement of blind variation to the fact that it is logically impossible
to validly infer empirical generalizations or theories from observational data. If, indeed, there is no rational theorygenerating method and no logical way of assessing a theory's validity, there is no justification for the theory in the
traditional sense. This is why Campbell qualifies the generation of new scientific ideas or theories as "unjustified
variation" (Campbell 1974a). However, do these "unjustified" variations truely have the same significance as quasirandom mutations or recombinations in biology?
At first sight, the answer to this question would seem to be in the negative, considering that in conceiving new
theories, we are already equipped with a guiding apparatus even though this apparatus itself is an evolutionary
product. One of the major programs of EE expounds the view that our cogni-
< previous page
page_148
next page >
page_149
< previous page
page_149
next page >

Page 149
tive apparatus is itself an evolutionary productboth on the organic and on the cultural levels. Our cognitive
apparatus determines our ways of concept formation and determines what kinds of predicates will appear in our
natural languages, and thusour standards of similarity. This is a Kantian-like outlook: we impose our notion of
similarity and our categories upon the world. However, our system of natural kinds and our conceptual systems are
not eternal and are not immune to revisions. In science, the system of natural kinds is a product of the selection of
ideas and theories. Electrons, which have both particle and wave aspects, are natural kinds in science, although
natural languages do not accomodate this kind of predicate. Scientific language is thus the successor of natural
language in this respect: it is a product of an extended cognitive apparatus which has evolved with science and
which consists of the widely accepted world-view of the scientific community.
Our cognitive apparatus, our general expectations and our system of natural kinds are genetically and culturally
based. However, they are not infallible and are not rationally justified. Since they are the results of a natural
selection process, they reflect both the nature of our species (our interests and needs) and the structure of the
environments in which they have evolved. Thus, we are imprisoned within our conceptual system. But, unlike what
is claimed by "pessimistic" Kantianism, we transcend this system in the process of scientific evolution.
The system of natural kinds and the related general world picture thus guide us in the construction of a definite
hypothesis (among the many logically possible ones) when explaining a given set of data or when solving a given
problem; i.e. they fulfill the task of narrowing down the range of possible explanations or solutions and in many
cases we are left with a unique possibility. Let us look first at an example from everyday life. Suppose we are
reading a book and it becomes too dark to read; if there is a lamp in front of us, we do not have to guess blindly in
order to generate the hypothesis which will almost always solve our problem. We simply hypothesize that the
solution is to switch on the lamp. Although there is no logical justification for this hypothesis, we do not discover it
blindly or generate it randomly. In arriving at our fallible solution, we are guided by our implicit rules of induction
and by our system of natural kinds (the latter determines what sort of predicates are amenable to inductive
projection). This sort of solution is fallible, but in most cases successful. Thus, although we go beyond what is
strictly known (we do not really know that will happen when we press the button), and although our hypothesis is
logically unjustified, we do not arrive at it blindly; our prior expectations and our general world picture yield
algorithms or heuristics which guide us in explaining given data or in solving given problems. It is in this sense
that the process of discovery of a solution to a problem would appear to contradict the analogy with a blind
mutation, which is generated independently of its contribution to the survival or the needs of the organism.
< previous page
page_149
next page >
page_150
< previous page
page_150
next page >

Page 150
In the development of science, we thus indeed have a very elaborate guiding apparatus. "Normal" science,
especially, is based on a relatively stable world picture and conceptual system, on established general theories and
on methods of theory-construction and concept-formation. Thus, problem-solving in normal science is guided. The
prevailing tradition or heuristic partially guides scientists in the choice or in the construction of a hypothesis. In
Chapter 2 I described a variety of methods and guiding tools for exposing and generating discoveries. For example,
in constructing an empirical law scientists are guided either by inductive rules or by statistical methods.
Alternatively, there are heuristic principles for constructing explanatory theories which are sometimes derived from
a general metaphysical outlook or world picture.
Thus, in normal science, even more than in everyday life, the typical case is that scientists appear to solve problems
by intentionally trying to solve them, unlike the case of blind mutations, which do not arise as a response to
selective environmental pressures. The fact that problem-solving in normal science is not blind would therefore
seem to contradict the paradigm of natural selection. Campbell, however, gives two arguments showing that natural
selection still operates in the evolution of science (Campbell 1974b). The first argument maintains, as I mentioned
above, that the guiding scientific world picture or tradition is itself a product of preadaptation, i.e. it is a product of
selection on the scientific, cultural and organic levels. The range of variation on the scientific level is thereby
reduced by selective processes at the cultural and organic levels; our cognitive apparatus, which is a product of
organic and cultural evolution, reduces the range of possible ideas and theories. The scientific world picture, which
is a product of selection on the scientific level, then further reduces the range of possible variation on the level of
normal science. We observe, therefore, the following rule in this hierarchy: the range of possible variation at any
given level is reduced by selective processes at the underlying level (see Amundsen 1989). Thus, normal science is
preadapted by the process of selection to its domain of investigation.
If we look for the parallel situation on the organic level, we notice that the range of possible variation is limited by
the constitution of the organism's genotype and by the laws of molecular biology. More specifically, the range of
mutations which a given gene can undergo is restricted by the gene's structure. Hence the mutational repertoire of a
gene pool is restricted or determined by the evolutionary history of the species, just as the repertoire of new ideas
in science is restricted by tradition or by the world picture. Futhermore, as was argued by Francisco Ayala
(Dobzhansky et al. 1977, 6566) when a species is in a state of stability (which is the parallel of normal science),
and when the environmental conditions are stable and the species does not move into a new habitat, it is more
probable that new mutations will be detrimental rather than advantageous. The reason for this is that if a certain
variation does not appear with a high percentage in the gene pool, it is most probable that it
< previous page
page_150
next page >
page_151
< previous page
page_151
next page >

Page 151
has been tried in the phylogenetic history of the species and has already been repressed. Only radical
environmental changes might still expose an advantageous variation. The parallel statement with respect to science
is that radically new ideas are repressed in a mature, or an established, stage of normal science, when unexpected
experimental data are not produced, no radically new experimental technology is introduced and no new theoretical
results are imported from other areas. Thus, the presence of a restricting and guiding tradition does not contradict
the model of natural selection.
Campbell's second and main argument, in defending natural selection in science, is that even in the presence of a
tradition and a background knowledge there still must remain an element of blindness in scientific problemsolving, since the tradition or heuristic and the background knowledge do not uniquely determine the solution for a
given scientific problem. Thus, Campbell maintains that in general, there remains a range of possible solutions
among which the scientist can only choose blindly, i.e. independently of the data or of the problem to be solved.
His argument is analytic or logical: "In going beyond what is already known, one cannot but go blindly. If one can
go wisely, this indicates already achieved wisdom of some general sort...which limits the range of trials" (ibid.,
422). Note that to "go wisely" still means here to go blindly, though within a narrower range of trials.
However, in contrast with the above argument, in many cases scientists do arrive at new pieces of knowledge,
including new laws and theories, in a non-blind manner. There are two major ways of doing this: by deductive
inference and mathematical derivation, or through experimentation and observation. In normal science, scientists
aspire to gain new knowledge or to solve problems in a methodical manner through these means. Thus blind search
is by no means typical to normal science; it is the exception rather than the rule.
The first way of generating variations in a guided manner is through deductive inference or mathematical
derivation, whereby the scientist exposes a new theoretical result or prediction hidden in a given theory. For
example, Maxwell derived the existence of radio waves from his equations. However, as I have already mentioned,
although the results of a deductive inference or a mathematical derivation are logically contained in the premises, it
is not always true that the results are known to the scientist before he makes the derivation or the inference. Thus, it
is not true that Maxwell knew about radio waves as soon as he arrived at his equations. Yet, deduction and
mathematical derivation are by no means blind activities. Deductive inference is the prototypical way of gaining
new knowledge in a non-blind manner.
The most common way of gaining new knowledge by deduction or by mathematical calculation is when scientists
derive a prediction from a law or a theory in conjunction with statements describing observational data or initial
conditions. Thus, if the data or the initial conditions are accepted by the scien-
< previous page
page_151
next page >
page_152
< previous page
page_152
next page >

Page 152
tific community as true or reliable and if the theory is highly established, then the prediction may constitute a new
piece of knowledge. In any case, the prediction can be tested experimentally and if the results agree with the
prediction, we must conclude that we have here a new piece of knowledge which was gained non-blindly, since the
prediction guided us in deciding where to look. In this manner scientists predict not only singular events, such as
an eclipse, but also empirical laws, such as the laws which govern the behavior of a given configuration of bodies
(e.g. a planetary system) or particles (e.g. an atom of a given element or a nuclear system), which can be derived
from Newtonian mechanics or quantum mechanics, respectively, in conjunction with the physical properties of the
system.
However, there is a wider range of scientific activity which is modeled on deductive inference and which is
therefore non-blind. It includes research programs or processes of problem-solving which are guided by some
comprehensive heuristic. In section 2.2, I discussed the approach which converts ampliative inference into a
deductive inference. I also discussed heuristic-guided theory-construction. We may view these processes, too, as
generalized inferences, as if the heuristic fills the gaps in the deductive inference, playing the role of an "inferencelicense" or of a missing premise. The accepted heuristic for problem-solving and for theory-construction in a given
field is derived from the guiding apparatus of the scientific community. As we have seen, the heuristic, which
narrows the range of possible explanations or solutions, may leave room for creativity. Yet in some cases it may
reduce scientific inference to a deductive inference.
The second way of generating new variations in science in a guided manner is by making observations and
experiments. Empirical generalizations and laws of nature can then be derived from the observational data by
applying inductive rules or statistical inference, without a resort to blind groping. The available conceptual system
and the general world picture narrow down the range of variation such that in a given situation controlled
experiments may lead to a unique generalization. Furthermore, if there is a comprehensive theoretical framework
and elaborate methods or heuristics for theory-generation, new models or theories may be deduced from
experimental results. This might happen when the number of possible solutions of a given problem or the number
of possible explanatory theories is manageable. In order to obtain a unique solution the scientist will conduct
controlled experiments which may leave him with a unique solution, eliminating all other alternatives.
Thus, the general world picture and the heuristic may narrow down the range of possible variations so that the
arrival at the final discovery is a result of accumulated experimental data and deductive inference or mathematical
derivation. This is the reason why scientists often claim that they deduced or derived a certain theory or model
from the data. This deduction or derivation is valid only in view of the accepted wisdom of the scientific
community. Also,
< previous page
page_152
next page >
page_153
< previous page
page_153
next page >

Page 153
in this light we can understand how scientists who share a general theoretical framework, general beliefs and
heuristics do arrive at a consensus with respect to the acceptance of a certain theory, or sometimes independently
arrive at the same theoretical result, as if it were a result of a straightforward mathematical calculation.
The above description refers to the ideal case of scientific inference which is not always realized in practice. In
many cases an element of chance infiltrates into the process. But the point is that in principle blindness is not a
necessary condition for gaining new knowledge and for generating new solutions to problems or new explanatory
theories. On the contrary, in "normal" science scientists expect to solve problems and to discover explanatory
theories by inference or by heuristic-guided problem-solving. Thus, the model of blind variation would appear to
fail when we refer to normal science, which does not break with tradition; if scientists would adhere to this model
they would not waste so much time and energy in trying to solve problems in a methodical or a guided manner. It
is only radical or "revolutionary" changes in science that can clearly be described as typically involving blind
variation. However, even scientific revolutions have their roots in normal research or problem solving, i.e. in a
guided activity.
And yet, it will be shown in the next section that the paradigm of natural selection can indeed be retained over a
very wide range. I will suggest that science advances via a special class of blind discoveries, even though these
originate in the intentional or guided action of problem-solving. Although scientists do employ methods and
heuristics which guide their research in view of the data or the problems, these methods and heuristics frequently
lead them to unexpected discoveries. Furthermore, we will see that the "mechanism" which turns a problem-guided
activity towards an unforeseen direction may also be responsible for the revolutionary discoveries which destroy
the prevailing order and open new vistas for science.
5.2.2 The Evolutionary-Epistemic Significance of Serendipitous Discovery
In The Sleepwalkers (1964) Arthur Koestler writes: "...the manner in which some of the most important individual
discoveries were arrived at reminds one more of a sleepwalker's performance than an electronic brain's." The thesis
which will be offered here provides a specific interpretation for these words. It will be suggested that radical
scientific changes are very often triggered unintentionally by an innocent problem-solving activity within normal
science. At times, an activity intended to solve a given problem leads to unintended results. This pattern of
discovery might still keep us within normal science. However, it might initiate a process of scientific revolution. In
fact, blind discovery is a necessary condition for scientific revolution; since the scientist is in general "imprisoned"
within the prevailing paradigm or world picture, he
< previous page
page_153
next page >
page_154
< previous page
page_154
next page >

Page 154
would not intentionally try to go beyond the boundaries of what is considered true or plausible. And even if he is
aware of the limitations of the scientific world picture and desires to transcend it, he does not have a clue how to
do it; he is blind to any territory which lies outside the one governed by his world picture. As we have seen, there
are two major ways of gaining new knowledge in a non-blind manner, and both ways will always keep us within
the confines of our framework of knowledge. By deductive or inductive inference we cannot construct a radically
new conceptual system or world view. Also, the heuristic which helps us in constructing theories in view of the
observational data is part and parcel of our world picture and thus it cannot guide us in transcending the world
picture.
Thus, one of the major ways of transcending an established state of knowledge is to do it unintentionally while
trying to solve some problem within the confines of the prevailing paradigm. Indeed, it is well known to working
scientists that a large percentage of research programs in natural science deviate from the original path planned for
them, as if chance "drags" the research program in a new direction, a direction which sometimes leads to the
discovery of a new phenomenon or a new domain of reality. This is how serendipity is realized in natural science
(see Ne'eman 1980).
The term serendipity was coined by Horace Walpole. In a letter written to Horace Mann on the 28th of January
1754 he says that he formed this term following his reading of a "silly fairy tale" called "The Three Princes of
Serendip" (Serendip is an ancient name for Ceylon or Sri Lanka). The three heroes of this tale "were always
making discoveries by accidents and sagacity, of things they were not in quest of" (Lewis 1960). The Oxford
English Dictionary defines the term as "the faculty of making happy and unexpected discoveries by accident." The
Dictionary adds, however, the following sharper definition: "looking for one thing and finding another.'' The latter
definition refers to cases where one looks for A and finds B. Thus the scientist may act in a guided manner in order
to solve a problemwhile he discovers that the end result provides a solution for another problem, of which he was
not aware. The notion of serendipity implies that the discoverer is aware of the fact that he found B, or at least of
the fact that he found something unexpected or significant. Thus, science can benefit from a hint given by nature
only if there are open-minded scientists who grasp the significance of the hint. Sometimes the scientist who made
the discovery is not aware of the full significance of his discovery while other scientists complete the task. So in
many cases serendipity in science is a cooperative enterprise.
One of the best known cases of that nature is Fleming's discovery of penicillin. Taking a Petri dish containing a
bacterial culture he noticed that the loose cover had not been properly set, and a mold had grown over the exposed
area. The bacteria, on the other hand, were dead. This may have occurred to other researchers before him and their
conclusion must have been
< previous page
page_154
next page >
page_155
< previous page
page_155
next page >

Page 155
to see that lids should be properly clamped. ... Fleming realized that this implied that some molds could kill the
bacteria.
A very similar sequence ushered in the "superstring" hypothesis (see for example Schwarz 1985) in physics, which
was very popular in the eighties. "Dual Models," later shown to represent the quantum excitations of a string, were
developed in order to explain the strong nuclear force. One difficulty that appeared to plague the model was the
appearance of a spin-two massless state as the lowest physical state of the string's spectrum of excitations. Strong
interactions involve only massive states, and physicists tried unsuccessfully to give a mass to this spin-two state.
Yoneya (1974) and Scherk and Schwarz (1974) suggested that the quantum string be reinterpreted as a theory of
quantum gravity (since gravity is mediated by gravitons, massless spin-two ''particles") rather than of the strong
nuclear interaction. The string tension, the only free parameter in the theory, thus had to be changed by twenty
orders of magnitudes! Thus a theoretical construct invented to explain the strong nuclear force led to a deeper
understanding of quantum gravity. What is common to the above two events is that a difficulty was turned into a
discovery in an unexpected direction.
I will argue that serendipity in science in not a casual phenomenon. Understanding the role of serendipitous
discoveries will contribute to understanding the epistemic role of science and its evolutionary character. Serendipity
supplies science with its blind edge: The human mind makes plans which have a chance of yielding successful
results only in familiar territories of nature, while serendipity causes science to deviate from its planned course
towards unexplored domains of nature. Actually, serendipity enables the human mind to transcend established
frameworks of knowledge, established world pictures.
The requirement of blind discovery is realized in the phenomenon of serendipity in such a way that it does not
contradict the fact that scientists do act intentionally and that they direct their efforts towards solving given
problems. Indeed, when a scientist makes a serendipitous discovery he does not guess blindly. Rather he is
occupied with directed problem-solving in the framework of a research program, employing algorithms and
established methods. However, since he tries to solve problem A, being aware of problem A, while accidentally
solving (another) problem B, the solution of problem B is, indeed, generated blindly with respect to B. Thus, the
discoverer does act intentionally, being affected by the problem he intended to solve; and yet he ends up making a
"blind" discovery. By this we reconcile in a straightforward manner the fact that science appears to be a guided
enterprise with the evolutionary model of blind variation. Thus, variations are generated via the activity of
problem-solving and are selected by problems which they were not intended to solve. We might describe the
situation by saying that in his problem-solving activity the scientist generates "solutions in search of problems."
< previous page
page_155
next page >
page_156
< previous page
page_156
next page >

Page 156
Now, the more profound question is why does science advance in this manner? Why, for example, shouldn't
scientists who wish to find novel solutions to problems (or to free themselves from the prevailing world picture and
look for a new one) adopt the Feyerabendian slogan (1978) "anything goes" and start gambling with nature? Why
should they necessarily expect to find unexpected clues for the understanding of natural phenomena while
conducting "normal" research? Why shouldn't they instead draw their inspiration, for example, from works of art,
fairy-tales or wizards? The answer to this question may rest on the stepwise pattern of the growth of science.
When one plays with existing building blocks one may discover a new combination or a new configuration which
constitutes a solution to a problem. Perhaps every new tool is discovered in this way; the chimpanzee, for example,
may discover by innocent playing with sticks that two sticks can be combined to form a new tool which can be
used to knock down a banana from a tree. The scientist playing with theoretical concepts while trying to solve a
problem may find a new theoretical construct with which he can solve another problem. A novel form may appear
as an emergent property out of familiar constituents.
What is characteristic of the above pattern of discovery is that the discovered entity is constructed out of existing
building blocks. There can be no shortcuts in this process. A building cannot be constructed directly out of protons,
neutrons and electrons or even out of chemical compounds. First the bricks must be prepared. However, in our
context the most appropriate analogue can be drawn from organic evolution: mutations and recombinations are
superimposed on given genes and on given genetic structures. Following a process of natural selection, the system
may stabilize with a new genetic structure or a new gene pool. Thus, the new state of stability emerges out of the
old one. For example, Homo sapiens arose from some hominid ancestor. It could not have evolved directly from a
unicellular species, for example. Similarly, a new stratum of knowledge can be constructed out of the prevailing
stratum. The new conceptual system or the new world picture is constructed on the basis of some central concepts
and ideas of the old world picture. Quantum mechanics, for example, employs "mutated" concepts of classical
mechanics such as energy, momentum and the Hamiltonian formalism. It is improbable that quantum mechanics
would have been created on the basis of Aristotelian physics or even on the basis of the early version of Newtonian
physics, before classical mechanics was fully developed by Euler, d'Alembert, Lagrange and Hamilton into a
general dynamical theory. Thus, before a new layer of reality is exposed, and a new stage of knowledge is made to
emerge, the prevailing layer should be thoroughly explored and the prevailing stage should be fully developed.
This principle of "gradualism" or stratification in the growth of knowledge can be accounted for by the evolutinary
model of stratified stability which will be described in section 7.4.
Hence, discovery by serendipity is essential for the continuity of the advance of science. Variations at the organic
and at the scientific levels are not
< previous page
page_156
next page >
page_157
< previous page
page_157
next page >

Page 157
generated in vacuo, out of nothing; they are imposed on existing formsexisting genes or existing ideasrespectively.
Thus, serendipitous discovery guarantees both independence of problem-solving pressures ("blindness"), and
continuity.
It should be stressed that serendipity is needed for the advance of science because we conduct our scientific
investigations from within a given framework: a given conceptual system or a given world picture. Had we
adopted the naive empiricist view that we can acquire objective knowledge about the world just by making
unbiased observations, then serendipity would be unnecessary. Serendipity is needed in order to transcend an
established framework of knowledge.
5.3 Some Implications of the Principle of Serendipity
The theory of science which I propose is a naturalistic theory based on the evolutionist POR. It is evident from the
above discussion that the principle of discovery by serendipity is both descriptive and normative. It is descriptive
since I claim that science in fact advances by serendipitous steps. To be more precise, my approach is explanatory
rather than merely descriptive; we start with an evolutionary theory of science which, on the one hand, is checked
against historical evidence and which, on the other hand, attempts to explain scientists' decisions and acts. My
approach is normative, in the sense explained in the last chapter, since it maintains that in view of the basic
assumptions of our theory, science can make significant progress only by serendipity. Moreover, recommendation
for preparing the mind for serendipitous discovery may be derived from the theory. The theory explains, for
example, the fact that scientists prefer theories which yield unexpected predictions. It also may give us some more
insight on two phenomena which characterize modern sciencemathematization and cooperation.
5.3.1 Predictability and Epistemic Profit
The principle of discovery by serendipity sheds light on one of the most important methodological rules for
evaluating a scientific theory, i.e. that a theory, besides explaining known phenomena, should generate successful
predictions of unexpected events or phenomena. Thus, for example, Maxwell's electromagnetic theory
unexpectedly explained the phenomenon of light, and in addition predicted the existence of radio waves,
establishing relations between optical and electromagnetic phenomena. This requirement cannot be accounted for
by logic alone. We could perhaps be satisfied by "relegating" it to the psychological arena; yielding an unexpected
prediction has a dramatic effect, like telling the future or performing magic, thus persuading scientists to
< previous page
page_157
next page >
page_158
< previous page
page_158
next page >

Page 158
accept the theory. However, in view of our principle of serendipity we can avoid this sort of psychologism. If a
theory which was constructed in order to explain A also predicts B, it means that we explain B unintentionally, i.e.
by serendipity. Namely, in constructing the theory, the scientist could not be influenced by B, i.e. he was blind to
B. Thus, Maxwell did not set out to explain light; he calculated the velocity of propagation of electromagnetic
waves and was surprised to find that it fitted the known value of the velocity of light. When Dirac constructed his
electron theory, he intended to describe and explain properties of the electron as a quantum-mechanical particle
under relativistic conditions. The prediction of the existence of the positron came as an unexpected by-product. A
new "world," the world of antiparticles, was thus discovered, with the relevant particle-antiparticle symmetry. Of
course, after the positron was detected, its existence and its properties became parts of the theory's explanandum.
Thus, when we require high predictive power, it means that we require that the process of discovery of the theory
will be as blind as possible to the phenomena the theory eventually explains. The methodological statement that a
theory is confirmed by its successful predictions of novel facts can be translated into our evolutionary language by
saying that the theory is selected by facts and phenomena which were not taken into account in constructing the
theory.
In section 3.2, I cited some empirical studies which claim that the requirement of novelty is not always met in
scientific practice. However, in view of the principle of serendipity, novel prediction, including prediction of
known facts which were not taken into account in generating the theory, becomes a normative requirement (relative
to our evolutionist POR).
The ideal case would be when the discoverer generates the theory independently of any factual knowledge. The
process of discovery would then be totally blind to any data and if the theory yields successful predictions than the
"epistemic profit" will be maximized. Thus, the profit is maximal in a revolutionary change, following a blind
theoretical leap. The epistemic profit can be defined as the ratio of the amount of factual knowledge (or
information) predicted and explained by the theory to the amount of factual knowledge invested in constructing the
theory (in practice, of course, there is no simple measure for these "quantities" but in many cases scientists can
estimate the ratio). In Dirac's case, the input was just Lorentz-invariance, the output a doubling of the entire
particle world.
In view of our principle, we can explain the negative methodological attitude towards ad hoc modifications
imposed on a theory in order to explain some empirical data if they do not yield new predictions. Ptolemaic
astronomy, for example, which explained away every discrepancy between the theory and the facts by employing
epicycles, yielded no epistemic profit. Such ad hoc modifications are totally guided by the data, whereas our
principle demands that the generation of new variations be blind to some of the data they explain.
< previous page
page_158
next page >
page_159
< previous page
page_159
next page >

Page 159
Classifications, such as Linneus' tabulation of the living kingdom, Mendeleev's chart of chemical elements, or the
SU(3) symmetry of nuclear particles, are interesting when they predict new speciesso as to be falsifiable. However,
they then also take on a characteristic of "classifying A and finding B," with B's emergence representing a
broadening of the known universe.
A straightforward case of serendipity occurs when the theory yields an unexpected explanation for a known
phenomenon or an unexpected solution for a known problem. For example, as was mentioned above, Maxwell's
theory explained light, Newton's theory of gravitation explained the phenomenon of tides and Einstein's general
theory of relativity solved the problem of the abnormal behavior of Mercury. Thus, one of the major cases of
advance occurs when a new scientific theory solves a known problem which it was not intended to solve. The
discoverer in this situation solves the problem by using a theory which was constructed or discovered by scientists
who were not aware of the problem.
Thus, serendipitous events can be divided into two main classes:
1. intending to solve (explain) A, but solving (explaining) B instead;
2. intending to solve (explain) A, and solving (explaining) B in addition to A.
The case of an unexpected prediction belongs to class (2). Class (2) also includes cases where a research program
which solved the original problem A continues to evolve and solves problems, explains phenomena or leads to
ideas which were not dreamt of at the start. Ernest Rutherford, for example, was not satisfied with the idea of light
quanta, since he thought it lacked a physical basis, yet the Bohr-Rutherford model, which developed from his
initial model of the atom, contributed decisively to the acceptance of this very idea and lended it a high degree of
confirmation. In general, a research program starts with an initial version of a theory and ends up with a different
version due to ad hoc corrections and modifications made in the course of the development of the research
program. Thus, the later version can be seen as an outcome of the initial version, the initial discovery. The initial
version is intended to solve a given problem, whereas the subsequent versions of a progressive research program
solve additional problems. Hence, when we demand high predictive power we refer not only to a given theory as it
stands, but to its potentialities which can become actualized in a dynamic process of modifications and
improvements.
Lakatos treated the research program as the basic methodological entity in science which replaces the static theory
as the unit for appraisal. It is clear now that due to its plasticity, the research program or the dynamic theory can
also be characterized as the basic unit which may be subject to serendipitous developments; indeed, as we will see
below, the dynamic theory is developed in a social setting which provides the conditions for serendipitous
developments.
< previous page
page_159
next page >
page_160
< previous page
page_160
next page >

Page 160
5.3.2 The Mathematical and the Social Media

There are two characteristics of modern science which make it liable to serendipitous developments: the complex
deductive or mathematical structure of scientific theories and the highly cooperative nature of scientific research.
Although these two phenomena seem to take place on totally different levelscognitive and socialthey can be viewed
as serving a common purpose. Complex mathematical theory may yield far-reaching predictions of which the
scientist proposing the theory cannot be aware at the outset. Similarly, a scientist proposing a theory cannot know
in advance how the theory will be interpreted or exploited by other scientists and in what direction the theory will
be developed and modified by his collaborators or by his successors; after a theory or an idea appears in the public
arena it has a life of its own and does not remain any more under the control of its originator (see again
Rutherford's case). Thus, the discoverer of a new idea or theory is blind to some of its far-reaching consequences,
which are obtained either by mathematical development or as a result of its processing by the scientific
community. He may, therefore, unintentionally trigger a process which leads to a solution of a problem of which
he was not aware. The principle of serendipity, therefore, also sheds some light on the epistemic function of the
mathematical nature of modern science and of the cooperative nature of modern scientific research.
From the viewpoint of the socio-evolutionary POR, the important point regarding serendipity and unintentionality
in science is that it is mainly generated by the social dynamics of science. We encounter here a first example
where the evolutionary model is coupled with the social dimension of science.
5.4 Two Landmarks of Serendipity in Physics
Two of the greatest revolutions in physics can be viewed in the light of the principle of serendipity.
5.4.1 Kepler: The Conscious Sleepwalker
Johann Kepler is one of the heroes of Koestler's Sleepwalkers (1964). 6 Kepler's original problem was to explain
why there are exactly six planets and why the distances between their orbits are as they are. He was impressed by
the Pythagorean view that the world is governed by mathematical relations and hoped at first to find the solution to
his problem in the realm of arithmetics. He then looked for the solution in plane geometry. After he failed there
too, he turned to solid geometry. The erroneous number of planets, six, gave him the clue for his model of five
Platonic solids (the five regular convex polyhe-
< previous page
page_160
next page >
page_161
< previous page
page_161
next page >

Page 161
dra) on which he erected the universe. He associated the five perfect solids with the five interplanetary regions and
tried to find out how the proportions between the five solids are related to the distances between the orbits. After
many unsuccessful attempts to explain away the discrepancies between his model and the observational data, he
turned to another Pythagorean model: he tried to construct the heavenly motions as "a continuous song for several
voices (perceived by the intellect, not by the ear)." His model of the universe was thus constructed out of the five
Platonic solids and the musical harmonies of the Pythagorean scale. In his attempts to improve his model he
needed exact figures of the eccentricities and mean distances. These were supplied by Tycho Brahe. However,
Kepler encountered new problems in Tycho's data and found other parameters to investigate, instead of the relative
distances of the planets. The attempts to solve the new problems eventually led him to the discovery of his three
Laws.
The processes by which he arrived at his Laws involved many sequences of trials and errors. His Pythagorean
models stayed in the background and set the framework for his investigations. However, when he struggled with
the data, he arrived at his successful solutions mostly by error or by chance, not always recognizing their value or
significance. His struggle with the Martian orbit led to the First Law. His first assumption was naturally that Mars'
orbit is circular. Then he hypothesized an egg-shape orbit. After abandoning the egg-hypothesis, he constructed the
Martian orbit by very precise calculations and obtained a circle flattened at two opposite sides. He then found by
chance that a simple relation holds between two quantities related to the geometrical from of the orbit. As a result,
he obtained a simple formula expressing the functional relation between the planet's distance from the sun and its
position (Koestler, 33638). Since analytical geometry was not available in his time, he did not realize that the
formula characterizes an ellipse. And yet in his next step, he conjectured that the orbit is an ellipse. He thus
discovered his First Law twice: once by chance and once by making a hypothesis and testing it. This hypothesis
was almost the only one left for him after he had eliminated all other alternatives. The belief in circular orbits was
so deeply entrenched that Kepler could depart from it only by chance or by a lengthy process of trial and error.
The process by which Kepler arrived at the Second Law was hazardous in a different way. He discovered that the
radius vector of the earth's orbit sweeps out equal areas in equal times, after making three erroneous assumptions
which somehow led to the correct result. The assumptions were the following: "(a) that the planet's velocity varies
in reverse ratio with its distance from the sun; (b) the circular orbit; (c) that the sum of eccentric radii vectors
equals the area" (ibid., 591). With respect to assumption (b), note that he discovered the Second Law first. This
process of "error begetting truth" (to use Koestler's expression), which is characteristic of Kepler's investigations,
amounts again to blind discovery. Indeed, if we start a deductive inference
< previous page
page_161
next page >
page_162
< previous page
page_162
next page >

Page 162
from false assumptions there are equal chances to get a true and a false conclusion and the inference is thus
reduced to gambling.
The discovery of the Third Law came about after many years of search for a correlation between a planet's period
and its distance. Kepler arrived at the Law after innumerable trials. Hence chance did not play a significant role in
the final discovery.
In general, Kepler did not realize the importance of his Laws. Without Newton's theory the Laws looked arbitrary.
The belief in circular orbits of heavenly bodies was deeply rooted in the Ptolemaic and the Copernican world
picture and Kepler could see no reason why the planets would move in ellipses. He treated therefore the First Law
as a necessary evil. The Second Law was treated by him as a computational device. The Third Law was treated just
as one more step in the construction of Celestial Harmonies. Kepler thus thought that he was constructing his
Pythagorean models, the three Laws being just building blocks in this process. He was partially imprisoned within
the old traditions of Neoplatonism and Pythagoreanism. He could not therefore realize that he had made a decisive
step in transcending this "paradigm."
In summary, Kepler started with a problem in the old "paradigm," trying to give mathematical sense to the number
of planets and to their spatial distribution. He ended up identifying the mathematical regularities of planetary
motion, a discovery which led eventually to Newtonian Mechanics and to a new paradigm. The framework of
Kepler's grand research program was set while Kepler was totally blind to the final data and to the final problem
which he eventually solved. Although Kepler's original model of the Solar System has left no remnant in physics,
the mathematical element of his Pythagorean outlook did become part of the world picture of physics, established
by the Newtonian paradigm.
We can trace Kepler's train of reasoning from his writings. Other great scientists, such as Copernicus, Galileo and
Newton, mainly give us the final results of their investigations, hence we cannot trace serendipitous elements in
their work. Moreover, Kepler is partly aware of the serendipitous nature of his discoveries. He writes in the Preface
to his Astronomia Nova:
What matters to me is not merely to impart to the reader what I have to say, but above all to convey to him
the reasons, subterfuges, and lucky hazards which led me to my discoveries. When Christopher Colombus,
Magelhaen, and the Portuguese relate how they went astray on their journeys, we not only forgive them,
but would regret to miss their narration because without it the whole, grand entertainment would be lost.
Hence, I shall not be blamed if, prompted by the same affection for the reader, I follow the same methods.
(Cited in Koestler, 318)
< previous page
page_162
next page >
page_163
< previous page
page_163
next page >

Page 163
Indeed, if the notion of "serendipity" were available to Kepler he would have used it, since Columbus' story is one
of the most prominent examples of serendipity in human history. It seems, therefore, that Kepler felt that the zigzag
path of his investigations is significant enough to deserve a description in his scientific writings.
5.4.2 Planck: The Reluctant Revolutionist
Of the two revolutions which shook human thought at the beginning of the twentieth century the quantum
revolution seems to be the more radical. Relativity theory was based on classical theories, whereas the idea of
energy quantization and the subsequent principles of quantum mechanics sharply depart from classical physics.
(Indeed, nowadays relativity theory is included in textbooks as part of classical physics.) It is suggestive, therefore,
that the quantum revolution arose as a result of a serendipitous discovery, whereas the special theory of relativity
emerged out of a systematic analysis of known concepts and theories. And yet even in the latter case, a new theory
of space and time arose out of measurements that were meant to measure the velocity of the earth through the
aether.
Planck attempted to solve a problem concerning the Second Law of Thermodynamics and ended up solving the
problem of black-body radiation. The implications of the discovery had a radical impact on the whole world
picture of science. Planck treated the Second Law as an absolutely valid principle. Hence he did not accept
Boltzmann's statistical approach, which treated the increase in entropy, asserted by the Second Law, as "highly
probable" rather than absolutely valid. Planck spent many years trying to clarify and understand deeply the Second
Law. Before the turn of the nineteenth century he consequently became interested in the problem of black-body
radiation.
The problem of electromagnetic radiation emitted from a very small cavity in a hot furnace had occupied physicists
for half a century. The discrepancy between the observed distribution of intensity of the emitted light for different
wavelengths and the predictions of classical physics was termed "the ultra-violet catastrophe." The curve based on
the experimental data showed a very weak intensity for short wavelengths, in the ultra-violet region. The intensity
increased with wavelength until it reached a maximum at a certain wavelength (which corresponded to the
dominant color of the light emitted from the furnace) and then decreased and again became very weak in the infrared region.
The theory of black-body radiation was developed by Rayleigh, Jeans, Kirchhoff and Wien on the basis of classical
theories: the electromagnetic theory for treating light radiation and Newtonian mechanics for treating the oscillating
electrons within the walls of the furnace, which absorb or emit the light. Since the electrons in the walls must be at
equilibrium, they have to emit
< previous page
page_163
next page >
page_164
< previous page
page_164
next page >

Page 164
and absorb on the whole the same amount of radiation energy. Hence, a third theoryBoltzmann's statistical
mechanicswas incorporated in order to calculate the energy distribution in a state of equilibrium. Only in the infrared region was the curve of the energy distribution (or the intensity of the emitted light at equilibrium) predicted by
these theories indeed similar to the experimental curve. In the ultra-violet region, the intensity increased
indefinitely with decreasing wavelengths. Thus, at least one of the three theories employed for the calculation must
have been wrong.
As mentioned above, Planck's motivation in attacking this problem stemmed from his interest in the Second Law.
He had studied problems related to the scattering of electromagnetic waves by an oscillating dipole, which had
direct implications for the scattering of light by the furnace's electrons. His aim was to understand how the
radiation within the furnace is kept in a state of equilibrium at constant temperature. He was thus engaged in the
thermodynamics of radiation. In the course of his investigations he planned to derive the Second Law for a system
consisting of radiation and charged oscillators, in an enclosure with reflecting walls. I will not describe here the
details of these investigations with all their technicalities. I would rather refer the interested reader to Martin
Klein's detailed historical article on this subject (Klein 1966). I will only cite a few statements describing the aim
of Planck's research program. Klein writes: "The ultimate goal of this program would be the explanation of
irreversibility for conservative systems and, as a valuable by-product, the determination of the spectral distribution
of black-body radiation." In doing this, Planck hoped to "put an end, once and for all, to claims that the Second
Law was merely a matter of probability. How was Planck to know that he was headed in a very different direction,
that he had started on what he would later call 'the long and multiply twisted path to the quantum theory?'" It is
interesting to note that, in describing Kepler's thought, Koestler speaks about the "zigazag course of his reasoning."
This seems a paraphrase on the above-cited words of Planck, taken from his Nobel address. Needless to say,
Planck did not succeed in attaining his original goal (just as Kepler did not) and the by-product turned out to be the
major result of his enterprise.
In the last stage of his long serendipitous path he found that the problem of black-body radiation would be solved
if the energy of an oscillator could take only values 0, E0, 2E0, 3E0..., where E0 = hf, f being the frequency of the
oscillations and h a constant to be determined by experiment. Boltzmann had already used this idea as a
computational device, going to the limit E00. After six years of unsuccessful attempts to solve the problem,
Planck decided to employ E0 without going to the limit. Planck explains this move as "an act of desparation" and
he treats it, as he says, as "a purely formal assumption, and I did not give it much thought except for this: that I had
to obtain a positive result, under any circumstances and at whatever cost."
Thus, Planck's discovery was doubly serendipitous. First, his original
< previous page
page_164
next page >
page_165
< previous page
page_165
next page >

Page 165
goal was related to the Second Law of Thermodynamics but he ended up solving the ''ultra-violet catastrophe."
Second, when he arrived at the solution of the latter problem, he employed it reluctantly, without realizing at first
its far-reaching implications. He treated the energy quantization as a computational device rather than as a law of
nature. He did not commit himself to the existence of quanta as real physical entities. This may remind us of the
instrumentalist spirit of Osiander's preface to Copernicus' Revolutionibus, in which Osiander explains that
Copernicus meant his heliocentric hypothesis to be just a computational device, or Kepler's treatment of his second
Law. When Planck used his computational subterfuge, he did not dream that it would have immediate implications
in explaining phenomena such as the photoelectric effect and atomic spectra, not to mention the subsequent
developments and successes of quantum physics. Einstein was the first to treat quanta seriously, in explaining the
photoelectric effect in 1905.
5.5 Serendipitous Discovery of Natural Phenomena
The discovery of penicillin belongs to a different kind of serendipitous discovery which also contributes as an
evolutionary process to the advancement of science, but does not seem to be analogous to blind variation. I am
referring to the discovery of an unexpected phenomenon such as that of X-rays and radioactivity. As we shall see
in the following examples, the original aim of the investigation and the final result in these cases were entirely
differentresearching A and discovering B. The final result, however, was not a solution of a problem but the
emergence of a new problem awaiting a solution. The discovery of an unexpected phenomenon in the course of
normal-science research is similar to a new environmental pressure exposed by the species' activity (such as a
migration or an activity which undermines the ecological balance in the natural habitat). The new environmental
conditions pose a challenge to the species, which has to overcome new dangers and difficulties or exploit new
opportunities. There are two possible sources of variation which might enable the species to meet the challenge:
(a) Preexisting genes which are responsible for other functions. For example, in certain insects genes which are
responsible for metabolic functions also confer resistance to insecticides (in this example, the new environmental
pressure came about as an indirect result of the species' activity).
(b) Genes which have been kept in a dormant state, or genes with low frequency in the population, which are
activated by the new environmental conditions and spread through the population, since they endow high survival
value. This is typical of the evolution of resistant strains of bacteria.
< previous page
page_165
next page >
page_166
< previous page
page_166
next page >

Page 166
Similarly, novel phenomena may be explained by an existing idea or theory which was generated in order to
explain other phenomena, by a modified version of such a theory, or by an idea or theory that was latent in the
existing paradigm. In the following I will describe three examples of this kind of discovery.
Hertz, Roentgen and Bequerel A series of serendipitous discoveries at the end of the nineteenth century heralded
the approaching upheaval. The photoelectric effect, X-rays and radioactivity were discovered by serendipity and
were later explained by theories which stemmed from the quantum idea, itself the product of serendipity.
The photoelectric effect was discovered in 1887 by Heinrich Hertz while he was conducting experiments related to
the radio waves which he had discovered in 1886. He used a spark generator to produce the waves. In the course of
these experiments he discovered by chance that the behavior of the spark gap was affected by the illumination of
the electrodes. Other experiments following Hertz's observation showed that a piece of zinc illuminated by ultraviolet light became electrically charged. It was found that the effect is obtained for other metals and other
wavelengths of light, provided the wavelength is below some threshold, irrespective of the intensity of the light.
The metal became charged because negatively charged electrons were ejected from it by the incident
electromagnetic energy. It was further found that the speed of the ejected electrons was greater, the higher the
frequency of the incident light. Increasing the intensity of the light beam only affected the number of electrons
leaving the metal, which increased proportionally to the intensity. Below the threshold, however, no electron would
be ejected at any light intensity.
After Planck's discovery of energy quantization, Einstein was very quick to explain the photoelectric effect by
conjecturing that light waves of frequency f consist of light-particles, photons, each of which carries an amount of
energy E=hf. The intensity of the beam is proportional to the number of photons. Indeed, if the frequency of a light
beam is less then some threshold f0, no electron will be ejected, no matter how many photons there are in the beam,
since none of them has enough energy to knock out an electron.
Roentgen's discovery in 1895 came as a by-product of a long research program triggered by Faraday. Faraday and
his followers had investigated the phenomena which occur when an electric discharge is set up in partly evacuated
glass tubes containing two electrodes. Advances in vacuum pump technology led to the accumulation of new
experimental data and to the conjecture that the luminescence which appears near the anode is produced by what
was called "cathode rays." The research program culminated in 1897 with the results of J. J. Thomson's
experiments showing that the cathode rays were negatively charged particleselectrons. The aim of the research
program was
< previous page
page_166
next page >
page_167
< previous page
page_167
next page >

Page 167
to understand the nature of the cathode rays. However, before this aim was attained, X-rays were discovered by
chance, as a by-product.
Roentgen inserted into the cathode-rays tube a metal plate which formed an angle with the path of the cathode
rays. At each discharge within the tube, Roentgen observed a bright illumination of a screen covered with a
fluorescent salt situated outside the tube. It was evident that the cathode rays could not cause this glow since it had
been previously proved that they cannot penetrate the glass walls. It turned out that the higher the vacuum in the
tube, the more penetrating the new rays were.
The explanation of the X-rays phenomenon came later, with quantum theory and the atomic model. These rays are
produced when high speed electrons (such as cathode rays) bombard a target atom and as a result one of the
electrons in an inner shell of the atom is removed. The rearrangement of the electrons in the shells is accompanied
by a decrease in energy and an emission of an X-ray photon.
The discovery of radioactivity in 1896 was even more accidental than the above two discoveries. Henri Bequerel
knew about Roentgen's discovery and his aim was to find, as he says, "whether the property of emitting rays was
not intimately bound up with phosphorescence" (cited in Hurd and Kipling 1964, 363). As phosphorescent
substances he used uranium salts. He used a photographic plate wrapped with two sheets of thick black paper to
protect the plate from sunlight. He placed a plate of the phosphorescent substance on the paper and exposed the
whole thing to the sun. After developing the photographic plate, he saw the silhouette of the phosphorescent plate
in black on the negative. One day, Bequerel tells us, when "the sun only showed itself intermittently, I kept my
arrangements all prepared and put back the holders in the dark in the drawer of the case, and left in place the crusts
of uranium salt. Since the sun did not show itself again for several days, I developed the photographic
plates...expecting to find the images very feeble. The silhouettes appeared on the contrary with great intensity"
(ibid., 365). After conducting further experiments he came to the conclusion that the phenomenon was not caused
by radiation emitted by phosphorescence, and that the uranium salt itself emits radiation.
It was only after the famous experiments of the Curies that it was understood that radioactive elements such as
uranium and radium disintegrate and change their chemical identity. Bequerel's discovery led, therefore, to the
conclusion that chemical elements are not immutable and opened the way to the nuclear domain.
Let us now try to compare the theoretical explanations in these three cases to the above mentioned two possible
sources of variation enabling a species to meet new environmental conditions. The understanding of the
photoelectric effect is analogous to case (a), since it employed the idea of the quantum, which was invented for a
different purpose and was not treated seriously until 1905. The understanding of X-rays is also analogous to case
(a), since the phenome-
< previous page
page_167
next page >
page_168
< previous page
page_168
next page >

Page 168
non was explained by atomic physics, which had evolved from Planck's and Einstein's ideas and which was
intended to explain other phenomena. Radioactivity was explained by employing ideas borrowed from chemistry,
by using a heuristic of the Meyersonian kind for constructing matter-theories (case (b)).
Thus, a discovery of a novel phenomenon may foster theoretical advance, just as changing environmental
conditions may foster evolutionary progress. In each of the above examples the resulting theoretical advance was
serendipitous, since the new phenomenon was exposed following an experimental activity guided by an established
theory or by an entrenched conception, whereas the end result was the adoption of another theory or conception
which superseded the original one. In other words, the original research employed a guiding theory or conception
to solve a given problem, whereas the end result was its rejection or a radical change imposed on it. Hence the
epistemological significance of such a discovery is in triggering the process of replacing an entrenched conception
or theory. This sort of serendipitous discovery, just as the blind discovery of new explanatory ideas and solutions
of problems, serves, therefore, as a means of transcending an entrenched framework of knowledge. The discovery
of an anomalous phenomenon which is unexpected in view of the entrenched background knowledge must, indeed,
be serendipitous. In addition, as the example of the discovery of penicillin indicates, such a process sometimes can
be viewed as a serendipitous discovery of a solution of a practical problem. The discovery of penicillin solved an
acute medical problem. The discovery of X-rays and radioactivity solved experimental problems within physics
itself by providing new methods for probing the structure of matter.
We would expect that discovery of new phenomena by serendipity will be more frequent in sciences which do not
yet have a fully developed theoretical system to guide research. Indeed, this is very common in certain areas of the
medical sciences, such as drug research; the example of the discovery of penicillin by Fleming is repeated again
and again in this field. It has also typically occurred, as we saw above, in fluid stages in the evolution of physics.
Other classical examples are: Galvani's discovery leading to the electric battery, Brown's discovery of molecular
motion (two cases of interdisciplinary serendipity), and Oersted's discovery of the nature of the connection between
electricity and magnetism. Other examples are described in Cannon (1961), Shapiro (1986), Roberts (1989) and
Kohn (1989).
5.6 Cultivating Serendipity
There can be no method for generating serendipitous discovery, since a discovery generated by employing a
method which is directed towards the particular product of discovery cannot be unintentional. However, although it
is an
< previous page
page_168
next page >
page_169
< previous page
page_169
next page >

Page 169
unintentional phenomenon, we can enhance the chances for its occurrence by preparing appropriate conditions for
it; we can cultivate it. Below I will propose some "rules," or recommendations, for the cultivation of serendipity.
Most of them are practised intuitively. However, the principle of serendipity sheds new light on them. We can
interpret the role of these rules as helping to "prepare the mind" for serendipitous discoveries.
(a) The first rule or recommendation would be to adopt the policy of setting the target by encircling the spot after
the arrow has hit it. This recommendation is well known in its negative connotation. The principle of serendipity
sheds a positive light on it. Fleming adopted this policy when his bacterial culture was contaminated by mold and
thus spoiled his original experimental setting. What could be regarded as difficulty, was converted by Fleming into
a great discovery. Kepler also acted in this manner. He started by constructing a model of the universe out of the
five Platonic solids and the musical harmonies of the Pythagorean scale. His original goal was to give mathematical
sense to the number of planets and to their spatial distribution. As we have seen, his unsuccessful attempts led him
to solve other problems. He ended up identifying the mathematical regularities of planetary motion. The discovery
of the superstring came about as an opportunistic exploitation of the difficulty which arose by the appearance of
the undesirable spin-two massless state which was inappropriate for strong interactions, i.e. by reinterpreting the
quantum string as a theory of quantum gravity. Thus, sometimes the scientist is engaged in unsuccessful attempts
to solve a problem. He stubbornly tries again and again to proceed with Sisyphean efforts. If he were alert to the
possibility of solving other problems along the way, he might find out that his efforts are not in vain and that
another problem could be solved by the results already achieved. In this way an obstacle may turn into a victory.
George Polya describes a typical situation which the problem-solver encounters when he is working on a problem
A which is connected to another problem B. The study of problem B may bring him near his goal of solving A.
Dealing with B "may stir his memory and bring into the surface elements that he can use in solving his original
problem A" (Polya 1965, 90). Polya is employing here concepts which fit Simonton's account of the incubation
process to be discussed in the next chapter. The dilemma which, according to Polya, faces the problem-solver is
whether or not to invest time on B. This dilemma would not arise if we believe in the principle of serendipity. We
cannot know in advance which problem is related to A in the sense that dealing with it may help solving A. This
can be discovered only in hindsight. Serendipitous discovery happens when one does not expect in advance that B
is related to A. Hence, the answer to the above dilemma is that if one wishes to benefit from serendipity, one
should not deal with B only because it seems to him related to A. The serendipitous situation is an opposite one:
when one's
< previous page
page_169
next page >
page_170
< previous page
page_170
next page >

Page 170
goal is solving A, one might serendipitously solve a different problem B, which he did not intend to solve.
(b) The second recommendation is derived from the first: we have to be engaged in solving more than one problem
at a time, to be engaged with problems which do not seem to be related to each other, and to be aware of as many
problems as possible. And in general, one should not restrict his domain of interest to his current narrow area of
research. We have here a definite reason for being broad-minded.
(c) The next advice, which is directed to the scientific community, is to encourage freedom of research. This
demand does not stem from a sheer liberal-idealistic attitude. It can be justified, or rather explained, from the
viewpoint of the principle of serendipity. According to this principle, good science should not be controlled by
preconceived goals. A government agency might support research for solving a given problem. However, the
scientist should not be possessed by the initial goal. He should be alert to different problems which might be
solved by his research program. Hence, we arrive at the seemingly paradoxical conclusion that in order for society
to benefit from science, science should not be forced to solve the problems of society or to be controlled by
society.
Thus, the principle of serendipity has direct implications for science policy. First, basic research should not and
cannot be directed from outside the realm of science. According to Yuval Ne'eman:
We can only direct towards targets we know, but the more important ones are the unknown. A society or a state
that decides to concentrate on 'relevant' aims is in fact stopping progress. This happened in the USSR in the
Thirties to Fifties, in China during the 'cultural revolution' (which actually almost destroyed a culture) and even in
the West around 196570 in a milder fashion. (Ne'eman 1988)
The same advice should be directed towards the scientific community's policy for the granting of funds according
to submitted proposals:
In preparing such a proposal the researcher can only use extrapolation as an input. The really important advances
that will result from his proposal are those of which he has no inkling. The grantor should not take the proposal
seriously. He can only judge by past performance whether or not this researcher will detect something unexpected
in the course of his research. (Ibid.)
In the research report which the supporting agency asks the researcher to submit at the culmination of the research,
the following question should be
< previous page
page_170
next page >
page_171
< previous page
page_171
next page >

Page 171
avoided: "to what extent have you achieved the goal set up in your research proposal?" There is an essential
difference between an engineering project and scientific research: an engineer is supposed to achieve a predetermined goal. For example, if an aircraft is designed for certain purposes, then the project will not be considered
successful if the aircraft will not fulfill the predesigned task, even if the engineer discovers that it can perform
another task.
In general, the above recommendation is not followed in the scientific community. Indeed, in a reaction to our
principle of serendipity, James Baggot writes: "today's patterns of science funding are reducing the opportunities
for making...serendipitous discovery." What would encourage this kind of discovery is "the freedom to deviate
from a proposed research programme when such a discovery is made or appears possible" (Baggott 1990).
(d) It should be stressed that the principle of serendipity does not imply anarchy of the "anything goes" kind. In
order to enhance the chance for serendipity, scientists should be engaged in problems which appear on the current
level of scientific research, employing the most advanced knowledge and methods. Indeed, in arriving at the idea
of the quantum, Planck employed the latest ideas and mathematical tools of the time and had been engaged in
solving problems in the current paradigm. This is directly implied by the conceptions of stratified stability and
gradualism in scientific progress, which will be discussed in Chapter 7.
< previous page
page_171
next page >
page_173
< previous page
page_173
next page >

Page 173
Chapter 6
Intrapsychic Processes Of Creation
Even sleepers are workers and collaborators in what goes on in the universe.
Heraclitus, Fragment 124 (ibid. 79)
No doubt that some major scientific discoveries are generated by individuals who set out to solve a problem or to
construct a theory. In these cases, the process of discovery seems to be both intentional and creative. Newton's
discovery of universal gravitation, Fresnel's discovery of the wave nature of light and Einstein's discovery of
special relativity, are examples of revolutionary discoveries which are apparently intentional. (Although the
anecdote about Newton's arriving at his theory after observing a falling apple suggests that chance may have been
involved in his discovery). This seems to undermine the evolutionary picture of science. However, even processes
like these may be accounted for by the natural-selection model. But this time the natural selection is below the
threshold of awareness.
There are some indications that the brain brings order to information and creates an inner thought model of the
world during rapid-eye-movement (REM) phase of sleep (Winson 1992). As Karel Pstruzina puts it:
During REM-sleep...our brain slips into natural process in which the present day information becomes
involved with our endoceptive structure of thinking. Thinking that arises during REM-sleep is inarticulate.
It is not eas[y] or convenient to become aware of it. ... Behind the shut eyes there run products of our
imagination, thinking is diverted to remote association, and free imagination enters in. ... This sort of
knowledge touches the limit of our mental faculties. (Pstruzina 1992)
This is the kind of process I am referring to when I am talking about involuntary natural processes of creation.
Dean Keith Simonton's theory of creativity
< previous page
page_173
next page >
page_174
< previous page
page_174
next page >

Page 174
(Simonton 1988) 7 may acount for this phenomenon. Simonton provides us with a theory about the creative
process which takes place in the mind of the scientific genius. But we can treat the theory as describing the
intrapsychic process of creation in generalnot only in the genius' mind. Simonton's theory postulates the deep
mechanism which is responsible for the creative process of incubation. The theory explains some typical cases of
serendipitous events which occur when the scientist's mind hosts the incubation process.
The creative process is modeled on Donald Campbell's paradigm of blind variation and selective retention and it is
divided into the following personal and social components.
a) An intrapsychic process (taking place within the individual), which is comprised of two stages:
a1: The chance permutation of mental elements.
a2: The formation of stable configurations (internal selection).
b)Interpsychic (interpersonal) processes: the communication, social acceptance and sociocultural
retention of selected configurations (social selection). Let's look at each of these in more detail.
6.1 A Psychological Theory of the Creative Process
6.1.1 The Chance-Configuration Model
a. The Mechanism of Chance Permutation The fundamental creative process postulated by Simonton is a natural
selection process taking place in the scientist's mind. The raw material for the process are "mental elements,"
including cognitive entities, such as facts, principles, relations, rules, laws, formulas and images.
These mental elements can be deliberately evoked by retrieving them from memory. They can also be evoked
involuntarily, e.g. via an association. They need not be consciously entertained; they may be processed at the
periphery of consciousness. In fact, there are grades of consciousness. In this connection, Simonton quotes Einstein
saying that what we "call full consciousness is a limit case which can never be fully accomplished."
New combinations of mental elements are formed by chance in the scientist's mind. This is the basis of the creative
process. Simonton calls these combinations "chance permutations." He borrows this term from probability theory,
since the order of mental elements in the set is important. Indeed, in permutations the elements' order is important,
whereas in combinations it is not. For example, one kind of creative product is a mathematical proof, which
consists of an ordered set of logical steps.
< previous page
page_174
next page >
page_175
< previous page
page_175
next page >

Page 175
The chance permutation theory leads Simonton to conclude that "in a loose sense, genius and chance become
synonymous" (33). With this, the heroic image of the discoverer gets another decisive blow. The discoverer is
equipped with merely an efficient gambling machine. However, later we will see that, according to this theory, it is
not chance alone which generates discoveries.
b. The Formation of Configurations The selection process takes place on both the intrapsychic and social levels. In
the intrapsychic process, the most stable permutations, called configurations, are selected among the mental
aggregates formed in the chance-permutation process. The discoverer cannot become aware of the non-stable
permutations; only the stable permutations are processed consciously. Since a configuration is relatively stable, it
may be viewed as a new mental element which can serve as a unit in forming new combinations. Here Simonton
employs atomistic and chemical metaphors: the process of forming a stable configuration is analogous to the
chemical process of forming molecules from atoms. Mental elements tend to form stable units when they have
intrinsic affinities for each other. And "large clusters of elements also can spontaneously form highly ordered
arrangements out of chaos." In order to illustrate this process, he cites Campbell's example of crystal formation out
of a dissolved chemical. We can also resort to a geometrical metaphor, where a simple geometrial figure is formed
out of juxtaposing several geometrical figures which match each other.
What is the principle of selection? A permutation is stable if it is sufficiently coherent. The notion of coherency
here is too broad to be specified in more detail. It is context-dependent. Only if there is a shared sense of coherence
in a community of scientists, may a configuration generated by one of its members be widely accepted. The
principle of selection cannot be a priori specified. The sense of coherence would depend on the prevailing world
picture, standards of scientific explanation which have been internalized by the discoverer, as well as on his
personal preferences. Hence, Simonton's talk about "natural" or "intrinsic" affinity between two mental elements is
inappropriate, since these notions have an objectivist or absolutist flavor.
Stable permutations may be inventive as well as imaginative. When the stable permutation arises by chance, it is
restricted neither by logic nor by experience. It is thus fully imaginative. If on the other hand the configuration is
formed by applying logical or mathematical procedures, it is inventive, since it is generated intentionally, with the
guidance of rules (yet, as we have seen, there are inventions which are generated unintentionally). A genuine
creative process generates new configurations by chance. Simonton calls configurations formed in a rule-governed
manner, via mathematical or logical procedures, "a priori configurations," whereas "a posteriori configurations" are
those arrived at via experience. Both a priori and a posteriori configurations are not genuinely creative.
< previous page
page_175
next page >
page_176
< previous page
page_176
next page >

Page 176
The conjecture about the wave nature of light is given as an example of a creative configuration. Here the
association was formed by making an analogy between the two a posteriori configurations corresponding to the
empirical character of light and the wave phenomenon. Another example is the isomorphism between the a priori
configuration represented by the Balmer formula and the a posteriori configuration representing the spectral lines
of hydrogen. In both examples, there was an element of chance in the creative process. The conjectures about the
wave structure of light or the Balmer formula for the hydrogen spectrum could not be straightforwardly derived
from the data. Both new configurations resulted in a greater self-organization of the available information. The
new configurations were stable since they united diverse pieces of information in a coherent scheme. Such a
unified scheme occupies less memory space and increases information-processing efficiency. Thus, Simonton
suggests that ''the human intellect is programmed to self-organize its contents into hierarchical structures in which
knowledge is most efficaciously distributed" (14). Self-organization, rather than the goal of truth, drives the
creative process in science. Simonton subsumes under the title of self-organization, notions such as regularity,
structure, order, harmony and beauty.
c. Interpsychic Processes Not every configuration formed in the above manner in the mind of an individual will be
accepted as a scientific discovery. Two conditions are necessary for this. First, the configuration should be
expressed linguistically or mathematically in order to be suitable for communication. Second, the configuration
should bring about self-organization in the minds of the other members of the scientific community.
Hence, if the configuration is not yet expressed linguistically or mathematically, for example if the configuration is
a visual image or a vague metaphor, the discoverer should verbalize or explicate it. It should be presented in the
language which is acceptable in the particular discipline. The configuration is thus converted into communication
configuration. In some cases this process of articulation is by itself creative. For example, the discoverer may
create a new mathematical language in order to convey his ideas, as did Newton when he created Calculus in order
to express his physical ideas.
The acceptance of the proposed configuration depends on various factors. For example, other members of the
community should recognize the problem for which the new configuration is supposed to provide a solution as a
genuine problem. Other members should also share a common repertoire of mental elements, such as facts,
theoretical principles and methods. Rhetorical and stylistic elements are also important.
6.1.2 Phenomena Explained by the Theory and Evidence for Its Support
Simonton seeks support for his theory from introspective reports of some eminent scientists which provide us with
evidence about the three compo-
< previous page
page_176
next page >
page_177
< previous page
page_177
next page >

Page 177
nents of the personal process of discovery: the nature of the mental elements, the operation of the permutation
mechanism and the unexpected act of illumination or the eureka event, when a configuration emerges and floats
above the threshold of awareness.
For instance, in describing the early stages of thought, Einstein employs notions which correspond to Simonton's
mental elements and to the act of combination: "The psychical entities which seem to serve as elements in thought
are certain signs and more or less clear images which can be 'voluntarily' reproduced and combined. ... This
combinatory play seems to be the essential feature in productive thought" (quoted on page 25). Einstein claims that
only prelinguistic elements participate in the permutation mechanism: "the words of the language, as they are
written or spoken, do not seem to play any role in my mechanism of thought." The transition to a communication
configuration is made in the second stage of the process, after a stable configuration has been reached:
''conventional words or other signs have to be sought for laboriously only in a secondary stage, when the
mentioned associative play is sufficiently established and can be reproduced at will" (2526). Simonton suggests
that the more original the chance configuration, the more difficult the task of communicating it to other scientists.
The early phases of a genuinely creative process emerge from prelinguistic imagery.
Support for the theory is provided also by Poincar who draws from his experience of discovering the Fuchsian
functions. He describes the process of forming a configuration via a vivid mechanistic metaphor. "Ideas rose in
crowds; I felt them collide until pairs interlocked, so to speak, making a stable combination" (quoted on page 29).
He makes the analogy between these colliding ideas and the colliding molecules in the kinetic theory of gases. In
describing the mechanism of producing stable combinations between mental elements or ideas, Poincar resorts to
the mechanism by which the hooked atoms of Epicurus are combined; the hooks presumably represent the affinities
that certain mental elements have for each other.
From the above descriptions we see that the notion of a mental element is too vague and the mechanism of forming
combinations is very metaphorical. But this should be expected, since the experience from which this notion is
derived barely passes the threshold of awareness. Nevertheless, since several eminent discoverers have provided us
with similar vague descriptions, we cannot dismiss the phenomenon. In any case, if we adopt the view that
inferring a theory from the data (such as that provided by the above descriptions) only gives it initial plausibilty (or
prior probability in the Bayesian approach), we would treat the chance-permutation theory as a plausible
hypothesis which has to be confirmed by further indirect evidence. This is exactly what Simonton is trying to do
throughout his book.
Descriptions of scientists and problem-solvers suggest that in many cases the initial efforts to solve a problem set
up a subconscious process of
< previous page
page_177
next page >
page_178
< previous page
page_178
next page >

Page 178
incubation which is terminated by a sudden act of illumination or an eureka event. Simonton reports the
description of this kind of event by Poincar, Gauss, Darwin and Helmholtz. The event is described as "automatic,"
"spontaneous" or "involuntary." It is a culmination of a process which the discoverer hosts unawares in his mind.
This event is unintentional and it follows a period in which the discoverer pauses. This period of rest is perhaps
required in order to give this natural process the opportunity to develop without conscious intervention by the
discoverer. The discoverer voluntarily contributes only to the preparatory phase of the process. What he discovers
is the final result; the discoverer exposes the solution after it was generated by the involuntary process. We may say
that in this case a generational discovery is reduced to a discovery by exposure. The discoverer acts as a spectator;
he observes the outcome of an involuntary process which takes place in his mind. This phenomenon may remind us
of the kind of discovery made by the individual scientist when he finds out that a given idea has emerged and has
been widely and dramatically accepted by the scientific community (this phenomenon will be discussed in Chapter
7). In both cases, the individual scientist acts as a spectator watching the event external to him. This fits in very
well with the descriptions given by creative people quoted in section 1.3.4.
Finally, the chance-configuration theory can explain why a theory is strongly supported when it successfully
predicts novel facts, whereas less support is derived from its ability to explain already known facts which were not
taken into account when the theory was invented. The reason is that part of the process of generating the theory
may be a subconscious process of chance permutation. In this case, even if the known fact is not taken into account
in the conscious phases of generating the theory, it might still be represented by a mental element which
participates in the infraconscious process of chance permutation. If we demand that the theory will be generated
blindly, then the above methodological requirement is explained. Moreover, relative to the theory of chance
permutation and the evolutionary POR (which demands blind generation) the requirement becomes normative.
6.1.3 An Associative Mechanism of Generating Chance Permutations
Simonton provides us with a hypothetical mechanism which generates chance permutations (4448). He is mainly
concerned with styles of thought of different types of geniuses, but his model can be employed for analyzing
different styles of creative thought in general.
People can be distinguished by the quantity of mental elements they possess and by the distribution of associationstrengths between the mental elements. The capacity for generating chance permutations depends on the
distribution of association-strengths. Two types of creators can be distinguished: the intuitive type, whose mental
elements are linked by many infra-
< previous page
page_178
next page >
page_179
< previous page
page_179
next page >

Page 179
conscious associations, and the analytical type, whose mental elements are linked by a smaller set of stronger,
predominantly conscious, associations. The two styles of thought are sometimes called divergent and convergent
thinking, respectively. The analytical type has mental elements clustered into compact configurations arranged in a
hierarchical order, where there are strong links between elements within a configuration and weak links with
elements in other configurations. The intuitive type has many interconnections between mental elements, a higher
percentage of them being infraconscious. In ordinary discourse we would say that the intuitive type is more
imaginative, whereas the analytical type in more restricted to passing from one element to another by logical and
explicit connections. The "richness of associative interconnections provides the psychological vehicle for chance
permutations." In the intuitive type's mind each mental element is linked with many other elements via direct and
indirect routes so that a larger amount of permutations may be generated. Since there are many nearly equiprobable
possibilities of linking each element to other elements in the network, the permutations thus generated can be
regarded as random, or at least quasi-random. The intuitive type is more alert than the analytical type to "novel or
unusual stimuli on the fringe of focused attention." This what makes the intuitive type more "susceptible to
serendipity"namely, this type of scientist is alert to problems which are beyond his focal awareness. Consequently,
in the course of looking for a solution to a given problem, this scientist might notice that he arrived at a solution to
a problem different from the one originally under investigation. Or in attempting to solve a given problem, he
might hit upon a solution which was generated in response to another problem. Since many of the chance
permutations are based on infraconscious associations, the intuitive discoverer is not always fully aware of the way
the discovery was arrived at, and may be unable to reconstruct it correctly. In fact, as we have seen, in many
typical cases, he reconstructs it as if he had arrived at the discovery by reason or inference.
The following "law" is implied by the above model: "the stronger are the associations that tie the elements of a
configuration, the fewer will be the active bonds among the configurations." This means that the intuitive type
generates a set of configurations which are more diffuse in the sense that the interconnections between the elements
within a configuration are weak, and the configurations are not sharply distinct from each other. However, intuitive
configurations may be systematized or explicated to become progressively more clear and distinct. The more
advanced the process of explication, the more consolidated the configurations become and the less they are
interconnected. The result is that there is less room for creativity through chance permutation. This phenomenon
has implications for both the individual scientist and for science. The individual scientist becomes less creative in a
given field, the more clear and distinct elements of knowledge he has acquired in the
< previous page
page_179
next page >
page_180
< previous page
page_180
next page >

Page 180
field. And it is well known that the more advanced a science, the less room left in it for creativity.
6.2 Implications of the Theory
6.2.1 Explaining Serendipity
The theory of chance permutation provides us with a possible mechanism which explains serendipitous discovery.
When the process of incubation for solving a problem has already been evoked, some external stimulus might then
introduce new mental elements or associations which provide the missing pieces in the "jig saw" puzzle leading to
a stable permutation. Thus, the process of incubation may require the injection of some additional elements into the
blend in order to generate a novel configuration.
The prototypical serendipitious discovery occurs when one tries to solve problem A but unintentionally solves
problem B. Here the external stimulus for generating the new configuration which solves B is provided as a byproduct of the attempts to solve A. According to Simonton's picture, any external experience may provide the
external stimulus. Yet, in science the more specific kind of serendipity, that is, when one finds the solution while
trying to solve another problem, is very common. Indeed, in the last chapter I argued that the discoverer is more
likely to solve problem B while being engaged with another scientific problem A, preferably in the same field,
rather than when drawing his inspiration from extra-scientific sources. The reason I gave for this was related to the
gradual manner in which scientific knowledge grows, each new layer of knowledge being constructed from
elements belonging to the present layer. The theory of chance permutation provides us with a sharper reason for
this: in being engaged with current scientific research, the discoverer is more likely to hit upon mental elements
which have "intrinsic" affinities for the elements participating in the process of forming the solution of B. As Polya
describes it: when the problem-solver is working on a problem which is connected to his original problem, he is
more likely to "bring into surface elements that can be used in solving his original problem" (ibid.). This is
especially true in modern science since it employs very specialized and abstract concepts and principles which
differ significantly from those used in everyday life.
When Archimedes noticed the overflow of water in the bathtub, he hit upon the missing element of the solution to
a problem which had already occupied his mindthe problem which had been posed to him by King Hieron of
Syracuse. Almost everyone who had ever had this sort of experience, including Archimedes himself, had noticed
this phenomenon before, but only after the process of solving the problem had been evoked in Archimedes' mind,
could he benefit from this experience in solving the prob-
< previous page
page_180
next page >
page_181
< previous page
page_181
next page >

Page 181
lem. Since he was engaged with a problem which was not far removed from everyday experience, he could hit
upon the clue for solving it while being engaged in everyday activity. However, the particle physicist, for example,
is not likely to encounter in everyday experience a missing piece for solving a puzzle which is couched in terms of
quantum fields, quarks, etc. In order to have a chance to find a clue for solving such a problem today's physicist
has to be engaged with research in particle physics or in any field which employs physical or mathematical
elements which might have "intrinsic" affinities to the elements used in particle physics.
In view of the above picture, we can interpret Pasteur's claim that chance favors the "prepared mind." We can give
a weaker and a stronger interpretation. According to the weaker interpretation, the discoverer's mind is prepared for
solving problems of a certain kind if it is equipped with the appropriate mental elements and the appropriate
associations which enable the process of chance permutation to generate solutions of this kind of problem.
According to the stronger interpretation, the discoverer's mind is prepared for solving a specific problem if he has
already entered an incubation process for solving the problem so that he can benefit from appropriate hintsas if the
incubation process opportunistically "attracts" the missing element which was generated for another purpose.
Poincar provides us with a tangible description which may be interpreted as referring to the process of preparing
the mind in the strong sense. Employing his metaphor of hooked atoms, he says: "the role of the preliminary
conscious work...is evidently to mobilize certain of these atoms, to unhook them from the wall and put them in
swing." (quoted on page 31). The discoverer thus prepares his mind by starting this process of "free-associative
procedure'' which then goes on autonomously, ready to pick up the first clue for completing the task.
Thus, although, as Simonton says, "the permutation process is 'blind' in the sense of being devoid of any a priori
knowledge of the most profitable direction to search for combinations" (32), the process does not take place in
vacuo; the raw material for the creative process consists of the "atoms hooked in the wall." The mental elements
and associations with which the prepared mind is equipped constitute the "a priori" component of the process of
discovery. They constitute the raw material for the chance-permutation process.
Darwin's case illustrates the notion of the prepared mind. Darwin arrived at his theory of natural selection via a
long process. He tells us in his Autobiography that after reading Malthus' Principles of Population in September,
1838, "being well prepared to appreciate the struggle for existence...it at once struck me that...favourable variations
would tend to be preserved, and unfavourable ones to be destroyed. The result of this would be the formation of
new species. Here, then I had at last got a theory by which to work" (Darwin 1958, 120). Lamb and Easton
comment on this: "The influence of Malthus was unlikely to have occurred as an instantaneous event but was
< previous page
page_181
next page >
page_182
< previous page
page_182
next page >

Page 182
rather a contribution to a lengthy process of thought." Darwin came to Malthus not for the first time. But this time
"by September 1838 his researches were at a stage when he could assimilate Malthus's insights" (Lamb & Easton
1984, 65).
Thus, when we say that the time was ripe for the discovery, we mean two things: First, that at the time of the
discovery the appropriate background knowledge had been accumulated, i.e. the "collective mind" of the scientific
community was prepared for the discoveryfor its generation and acceptance. The ideas of selective breeding,
Malthus ideas and various evolutionary models were all in the foreground at the time Darwin had his eureka
experience. Second, that the mind of the discoverer was prepared, in the strong sense, for the discovery. So, in a
sense, both the collective and the individual mind were prepared.
Yet, the range of creativity of the chance-permutation process is restricted by the structure of the network of
mental elements. The initial conscious efforts to solve the problem trigger the process by evoking a train of
associations which lead to the mental elements which participate in the process of chance permutation. These
mental elements are not picked up totally at random since they are interconnected through a network of
associations. The initial mental elements and associations determine, through the structure of the network, the
subset of elements which will participate in the process of chance permutation. Thus, if the system is closed, there
will be no opportunity to arrive at a novel element or association.
However, an incubation process which ends up following an external stimulus may lead to a radical novelty. After
the incubation process has reached a "pending state," where it is awaiting an external stimulus, or a new piece of
information, which will bring about a transition into a stable state, the missing element, generated through an
activity which has not been aimed at solving the original problem, may be caught up "opportunistically." The final
step, which is the decisive step in the process, is not affected by the problem-solving pressures. Thus, due to the
openness of the system, serendipitous discovery can take place, leading to radical novelty.
6.2.2 Individual vs. Collective Creativity
When Simonton introduces the social component of creativity, he does not take into account the full implications
of the social dynamics responsible for the formation of configurations. He still refers to personal factors and treats
"creativity as a form of personal influence over others and therein as a special variety of leadershipsociocultural
rather than political, military, or economic" (22). The question is what about a great discoverer that is recognized
as such only posthumously? For instance, in the case of Mendel, the discovery was eventually accepted despite the
unsuccessful attempts by the discoverer to
< previous page
page_182
next page >
page_183
< previous page
page_183
next page >

Page 183
convince his contemporaries. Thus, the above description is not appropriate for cases where a great discovery is
accepted for reason other than efforts made by the discoverer. It also does not account for cases when the discovery
is a cooperative-historical process. Planck, for example, did not believe in the reality of quanta; only after the
discoveries made by his successors was the idea of the quantum widely accepted.
Simonton maintains that "collaboration per se may not necessarily contribute to creative productivity...collaboration
may contribute to the elaboration and verification of an initial creative idea, but the original concept normally
arises from an entirely intrapsychic chance-permutation procedure" (5455). The following objection may be raised
against this claim. The individual scientist does not employ only mental elements and associations stored in his
memory. Interpersonal interaction provides important external stimuli, such as new ideas, clues and missing
elements, to the intrapsychic process of chance permutation. The basic process in this case may still be viewed as
an intrapsychic process of chance permutation or incubation, nurtured by communication with other scientistsbut
the ideas channeled from external sources are sometimes essential to the process. The discovery may depend upon
the integration of these external ideas with the intrapsychic associations network.
Furthermore, a configuration proposed by an individual scientist may not only be subject to "elaboration and
verification"; it may also undergo substantial changes via a cooperative and/or historical process. The final creative
product in this process might have only the slightest resemblance to the original idea proposed by the individual
who initiated the process. The social process might also be viewed as a chance-permutation process on the
interpsychic level, where many individuals in a given community of investigators propose their ideas which are
quasi-randomly combined with other ideas to produce new ideas, most of which are rejected at first sight and only
a few of which survive for further scrutiny. This process may eventually yield a final configuration which is
accepted as a discovery. Thus, the act of acceptance is not applied to a finished product. Acceptance is built in the
dynamic process of creating the product of discovery. Hence, in the process of discovery, elaboration and
verification cannot be sharply distinguished from the process of creation, as we have already observed.
Another related claim (5355) is that there is an "intrinsic motive to engage in scientific research for its own sake"
which "ensues from the more fundamental drive toward self-organization" and ''the successive discovery of chance
configurations is accompanied by personal satisfaction, or subjective pleasure." Or: "The quest for selforganization...provides a powerful incentive that shoves aside other motives." Against this view about "the inherent
superiority of intrinsic over extrinsic motivation," we may state the motive of obtaining recognition by the
scientific community. According to the former
< previous page
page_183
next page >
page_184
< previous page
page_184
next page >

Page 184
view, self-organization, rather than truth, drives the creative process in science. According to the latter view,
acceptance and recognition by the scientific community, rather than the quest for truth, is the main goal which
drives the creative individual in modern science. This is the basis of social epistemology. An idea may be accepted
by the scientific community (by the leading figures or by the establishment) and yet it does not necessarily bring
about self-organization in the mind of every individual in that community. Thus, both intrapsychic processes and
epistemic cooperation may contribute to the process of discovery.
Simonton does consider the view that science is a more democratic enterprise than seems to be implied by his
theory of chance permutation (94). If "the scientific edifice is built piece by piece using small bricks mostly laid by
undistinguished craftspersons" and if "anyone with the appropriate training in the trade of science may participate
in the construction of even the most impressive monuments," then ''the term scientific genius may prove
meaningless." But this is exactly what is implied by the chance-permutation mechanism operating on the social
level. Ideas proposed by numerous scientists are continuously combined and recombined in this cooperative
enterprise.
Thus, Simonton seems to reject the idea that there is a chance-permutation process on the social level. Yet, he
supplies the most decisive argument for not ignoring the contribution of the mediocre to the process of scientific
advance: "Just as we cannot have winners of a race without having losers too, so we cannot really have successes
in the absence of failures. What is required, according to the current theory, are many variations that are open to
sociocultural selection" (97). We can understand the above statement as saying that a successful idea is not
objectively or absolutely a true idea, but an idea which has been accepted according to current standards or as a
result of the social dynamics. Hence, the unsuccessful individuals are as important for the selection process as the
successful ones. Moreover, if we treat sociocultural selection according to the model of (quasi-) random variation,
we would say that the configuration which is accepted, or which proves successful, is not necessarily generated
with a guidance of method or logic; otherwise it would not be random. If the winning configuration is a product of
chance, then we cannot attribute it to the wisdom of the discoverer. Yet, method or logic are required for the
selection of variations. We may say therefore that the mediocre contributes to the generation of variations, whereas
the analytical genius plays an important role in the process of selection.
6.2.3 Cultivating the Creative Potential
Innovation can be cultivated when the discoverer is isolated from the community so that he is not committed to the
tradition, or when his mind is fertilized by ideas from other fields. The tension between tradition and innovation
can
< previous page
page_184
next page >
page_185
< previous page
page_185
next page >

Page 185
be accounted for by the chance-configuration theory in the following way (115). There are two conflicting
requirements for cultivating the creative potential. On the one hand, the scientist's mind must be equipped with the
repertoire of mental elements shared by the community of investigators. This body of mental elements must be in a
comparable condition of disorganization so that the scientist would share the same problems with his collegaues.
On the other hand, he must be ready to depart from this shared body of knowledge and problems. The creative
potential of a scientist who is not ready to depart from tradition would be diminished for two reasons. First, the
number of mental elements would be limited by the tradition. Second, the chance-permutation process would be
constrained by the strong ties already existing between the prevailing mental elements so that there would not be
much room left for generating chance associations. Thus, an essential tension exists between tradition and
innovation (Kuhn 1963) i.e. between the conformism required for social acceptance and the non-conformism
necessary for generating novel chance permutations.
One way of bypassing tradition and nurturing the creative potential is through sociocultural or professional
marginality. A scientist who is isolated from the scientific community is less constrained by tradition. Sometimes
"ignorance...allows the chance permutations to proceed afresh" (127). But ignorance is not always an asset since
science cannot advance if it always starts afresh. This would not allow the combinations to become cumulative. A
major way through which scientific knowledge grows is when configurations become themselves elements which
participate in the combinatorial play. The discoverer should therefore be equipped with the repertoire of
configurations that have already been generated in the field.
As we will see, another social means of nurturing the creative potential is through what M. J. Mulkay calls
"intellectual magration" (Mulkay 1972). A scientist who switches fields brings along ideas and methods from his
original field to the new one, thus widening the range of associations. Simonton mentions two examples. Kekul's
interest in architecture was perhaps the motive behind his developing the structural approach to organic chemistry.
Helmholtz, who became engaged in medicine, was also interested in physics and this led him to the invention of
the ophthalmoscope. Thus, intellectual migration and, more generally, cross-fertilization between different
disciplines nurtures creativity.
6.2.4 Multiple Discovery
The traditional explanation of the multiple discovery phenomenon is based on the assumption that the process of
discovery is predominantly affected by the prevailing world picture or the zeitgeist, rather than by intrapsychic
factors. If there is a common world picture shared by the members of the scien-
< previous page
page_185
next page >
page_186
< previous page
page_186
next page >

Page 186
tific community, certain discoveries are inevitable since these discoveries are "in the air." Consequently, several
scientists who share the common repertoire of ideas and try to solve the same problem may arrive at similar
solutions. Simonton thinks that the zeitgeist interpretation seems to contradict the chance-permutation theory, since
the zeitgeist view amounts to sociocultural determinism which does not allow random variation. However, the
zeitgeist view does not necessarily entail generation of predetermined ideas or permutations. The zeitgeist may
entail only common standards of selection, while generation might still be quasi-random. Quasi-randomness means
here that the mental elements out of which permutations are randomly formed belong to the common repertoire of
ideas. The zeitgeist or the tradition limits only the range of variationsit does not predetermine them.
The "multiples" phenomenon seems peculiar to science, and does not appear in artistic creativity. This leads
Simonton to think that his view which treats artistic and scientific creativity on a par may be undermined. But then
he "rescues" his theory by claiming that the multiples phenomenon does appear in artistic creativity. But this
sounds implausible. The Fifth Symphony, Hamlet and the Sistine Chapel ceiling are unique, although their styles
reflect their zeitgeist. Of course, there are imitations, but these would not be considered to be genuine artistic
products or counterparts of scientific discoveries.
Nevertheless, we can hold that both kinds of creativity are governed by the chance-permutation mechanism and yet
scientific creativity is characterized by the multiples phenomenon, whereas artistic creativity is not. There are two
substantial differences between artistic and scientific creativity. First, unlike artistic creation, scientific discovery is
intended to describe the world. No wonder that several scientists investigating the same domain of nature might
arrive at the same results, or at similar results. Scientific discovery is thus constrained by experimental results as
well as by zeitgeist. In contrast, artistic creativity is constrained only by the zeitgeist. Thus both kinds of creativity
may be based on the process of chance permutation, but the selection procedures are different. The configurations
selected in scientific discovery must accord with observational data, as well as with the world picture. Even if
observational data partially depend on the zeitgeist, they depend also on nature.
The second difference is that much of the practice of science consists of inferences and argumentations. As we
have seen, much of the theoretical argumentation in science, in particular in the natural sciences, is modeled on
logical deduction. In logical and mathematical inference people are supposed to arrive at the same conclusion,
starting with the same premises. This kind of inference takes place in normal science, where the theoretical
framework is given and typical discoveries are made by performing logical deductions or mathematical
calculations, as well as by conducting experiments and making observations. We would expect, therefore, that in
normal science scientists who start from the same transparent presuppositions and employ the same
< previous page
page_186
next page >
page_187
< previous page
page_187
next page >

Page 187
theory and observational data would arrive at the same result. No wonder that several scientists may independently
arrive at the same discovery. This may be an ideal case, but it shows that multiple discovery is not an exceptional
case; on the contrary, it is the regulating ideal of empirical and mathematically oriented science. In artistic
creativity, logic does not play such a role.
Scientific revolutions are not restricted by the prevailing paradigm, i.e. by the zeitgeist, but they should conform to
some transparadigmatic standards which characterize science and which are responsible for some minimal
continuity in its evolution. A revolutionary configuration should conform with some general principles and
entrenched beliefs and with some observational data which were inherited from the old paradigm. We thus would
expect the multiples phenomenon to occur less frequently in revolutionary science than in normal science. In this
respect, revolutionary science comes closer to artistic creativity than does normal science.
As a consequence of the above discussion it would be highly objectionable to maintain that "scientific contributions
are no more inevitable than artistic creations," even were it true that "one theory of creativity may account for
significant contributions in both the arts and the sciences" (148). The discovery of Neptune was inevitable because
the planet is there. Of course, we can say it now in retrospect because the fact that Neptune is a planet in the Solar
System is now established beyond any doubt. Only if we are extreme antirealists, might we treat the existence of
Neptune just as a convenient instrument for organizing our observations. When we come to a full-fledged theory,
such as quantum mechanics or the theory of evolution, the influence of the zeitgeist on discovering the theory is
more significant than on an observational discovery, but still, the theory can be strongly confirmed by observations,
a factor which does not have a counterpart in the case of a piece of art such as the Mona Lisa.
In the domain of sense experience, we have more or less a definite set of natural kinds and phenomena and we seek
the causes of these phenomena. The causes may be independently discovered by many people. Yet, when science
significantly departs from the domain of ordinary experience, it can proceed along different paths. Many different
theories can explain a given set of data. We choose a theory according to the zeitgeist, but this does not guarantee
that we will be left with only one inevitable theory. There is an element of chance in the process. The physical
theories which prevail are not inevitable. Had modern physics not started from the Galileo-Newtonian tradition, it
is entirely possible that it may have proceeded in a completely different direction. We already have another
example: the Aristotelian tradition. Yet scientists, in particular physicists, have the feeling that there are some
inevitable discoveries. The reason for this is the fact that they are imprisoned in their paradigms which limit their
horizons. They do not believe that their paradigm may be eventually replaced. This is true in particular if they feel
that their cur-
< previous page
page_187
next page >
page_188
< previous page
page_188
next page >

Page 188
rent theories are in general successful. Indeed, within a given paradigm certain discoveries are inevitable and as a
result we encounter the multiples phenomenon. This is also true of biological evolution. Along a given evolutionary
line certain evolutionary developments are inevitable.
So although there is only one world, there are many possible evolutionary paths in it. There are different ways to
adapt to a given niche. And the same is true of science. Hence, although the products of scientific creation are more
inevitable than those of artistic creation, still they are not inevitable in any absolute sense, unless we believe that
science aims at absolute truth, a belief that would not conform to the evolutionary picture. Multiple discovery and
the independent appearance of similar species in geographically isolated places are two manifestations of the same
evolutionary phenomenon. The reason why the products of scientific discovery are more inevitable than those of
the arts is that the environment in which the arts operate is determined by human culture, whereas the environment
of science is determined by nature as well. And the latter imposes severe restrictions on the ideas which would
survive since nature is less flexible than human culture; there is a hard core in nature which cannot be changed by
human action. The cultural environment, on the other hand, is a human product and therefore can be more easily
changed by humans.
< previous page
page_188
next page >
page_189
< previous page
page_189
next page >

Page 189
Chapter 7
A Socio-Evolutionary Theory Of Science
7.1 Epistemic Cooperation and the Social Dimension of Discovery
Scientific knowledge is a socially produced knowledge and scientific epistemology is to a large extent a social
epistemology. Hence, scientific discovery is a social product. Individual scientists do not, in fact, cannot, make a
scientific discovery. In order that a hypothesis, an observation or an experimental result will count as a scientific
discovery, it has to be approved by the scientific community. Furthermore, the product of discovery is produced
collectively, synchronically (by cooperation) and diachronically (by relying on predecessors). Even Einstein's
theory of relativity cannot be seen as an isolated discovery made by a lone discoverer, since it relies on past
scientific results. If an Einstein or a Darwin would be raised up in a different cultural setting, they might not have
produced their discoveries. They synthesize the scientific results of their predecessors and their contemporaries.
Thus, they can be viewed as crucial links in a historical-cooperative process. One cannot envisage Einstein's theory
of relativity without Maxwell, Poincar and Lorentz. According to this view, the great discoverer supplies the
missing link in the cooperative process by which the discovery is generated. Discoveries are not made in a social
vacuum. The discovery is "in the air," or "the time is ripe" for it. This means that (a) the state of knowledge in the
scientific community is ready for generating the discovery and/or (b) the time is ripe for the acceptance of the new
idea or theory.
The idea that inventions are the products of their time is an Hegelian idea. It is related to the cooperative nature of
scientific discovery. The soil for the invention is prepared by many predecessors. According to Hegel, the inventor
"is like the man who finds himself among workers who are building a stone arch whose general structure is
invisibly present as an idea. He so happens to be the last in line; when he puts his stone into place the arch
supports itself. As he places his stone he sees that the whole edifice is an arch, says so
< previous page
page_189
next page >
page_190
< previous page
page_190
next page >

Page 190
and passes for the inventor" (Quoted in Lukacs 1975, 356). The discoverer, likewise, appears on stage at the right
moment and puts his stone into place. I will cite three examples described by Lamb and Easton (1984). The first
two examples are those of Newton and Darwin. Newton's theory of universal gravitation developed between 1679
and 1685 on soil prepared by Galileo, Kepler, Huygens, Descartes, Fermat and Hooke. Newton started to develope
the theory when Hooke showed him the analysis of motion along a curved trajectory. It was developed through
communication with his contemporaries. The theory gradually emerged by logical deductions and transformations
of existing ideas "out of the repeated construction of mathematical models of the universe whose consequences he
compared with known observations and laws of the physical world, such as Kepler's laws of areas and of elliptical
orbits" (ibid., 117). 8 The soil prepared for Darwin was fertilized by the ideas of selective breeding, Malthus's work
on population and various evolutionary theories; the discovery was "in the air" (ibid., 120).
The third example of a revolutionary idea which emerged and gradually accepted in a historical process until the
time was ripe for it is the idea of extinct species (ibid., 105). The idea of species becoming extinct without leaving
any descendants was revolutionary since it was against entrenched metaphysical beliefs going back to Aristotle.
Several investigators entertained this idea in the late seventeenth century, but eventually yielded to the
conventional view. By the beginning of the nineteenth century Georges Cuvier had become convinced of this idea.
His evidence in support of the mammoth's extinction undermined traditional views. In 1822, Gideon Algernon
Mantell discovered iguanodon teeth in Sussex. He compared them with the teeth of the South American iguana
lizard and calculated with reference to their respective sizes, that the teeth he found must have come from a
creature over sixty feet long. As a result, the hypothesis of extinct species was gradually accepted. Mantell's
quantitative evidence supported it strongly. This encouraged the search for further fossils. After similar fossils had
been found throughout the South England, as a result of the intellectual shift, the idea became established.9
The example of electroweak unification in particle physics, which will be described in Chapter 8, also demonstrates
the cooperative nature of theory-construction. In this case it wasn't Hooft who put his stone into place and made
the arch supporting itself, although he was not last in line.
A discovery may be produced by an intentional cooperative work of a group of scientists who work on the same
problems, not necessarily in the same place. The group, sometimes called "invisible college," may spread over
several continents, having a strong communication network connecting them. This is predominantly a synchronic
cooperation, where the scientists conducting investigations on the same issues are aware of each other's work and
react quickly to each other's suggestions or results, as if they were participating in an ongoing dialogue.
< previous page
page_190
next page >
page_191
< previous page
page_191
next page >

Page 191
The cases where cooperation is not a preplanned process are the most innovative. It may happen that a scientist or
a group of scientists propose an idea or a theory and other scientists who become acquainted with the proposal
through the communication network and publication channels of the scientific community may suggest some
modifications or improvements, or contribute their own ideas, and as a result a successful theory is produced which
is a collective product. The process of discovery may stretch over a long period of time. In this case it can be
viewed as a diachronic process where the scientists do not cooperate intentionally; only with hindsight, may we say
that the final result is a collective product. The diachronic process may not even be continuous in the sense that a
neglected theory or idea may be revived and modified after many years. The diachronic process in which one
generation of scientists relies on the achievements and results of previous generations, is reflected in Newton's
famous statement that he could make his discoveries since he was standing on the shoulders of giants. The final
result is again a collective product.
We have here, therefore, another kind of unintentional discovery processes, where the process of discovery is, in a
sense, autonomous and has a "vitality of its own." When we view the intrapsychic process of creation and the
social process of cooperative creation as involuntary processes, we can explain the tendency of scientists to express
themselves in a modest fashion. The discoverer knows that he owes much of his achievments to his collegues and
predecessors, or to subconscious processes. This may explain why the use of first person singular is avoided in
scientific literature. This mode of behavior is accounted for by the Mertonian norms of disinterestedness or
communism (communality). The first norm forbids the scientist to profit personally in any way from his research.
In conclusion, the scientist should not make the search for professional recognition his explicit goal (Merton 1973,
276). The second requires that the scientist should share his findings, as a common property, with his collegues
(ibid., 273) and thus it encourages epistemic cooperation.
Both diachronic and synchronic manners of interaction between scientists are social patterns of behavior; the first
expresses reliance on the works of predecessors and the acceptance of tradition and the second reflects
cooperation. The synchronic process has become more and more dominant in modern science. This kind of
cooperation is encouraged by the norms of the scientific community and is channeled through a communication
network, including journals, conferences and, nowadays, computer networking, all of which make the scientific
community highly liable to synchronic cooperation. Thus, the scientific community is a social system in which
cooperation in acquiring and producing knowledge is institutionalized.
An important stage in the discovery process is the evaluation of the product in order to decide whether it is indeed
a discovery. This stage also has a social dimension. In science, the individual's belief is not recognized as a legiti-
< previous page
page_191
next page >
page_192
< previous page
page_192
next page >

Page 192
mate knowledge claim before it is processed by the organs of the scientific community. Via this processing the idea
is assessed, improved or modified and integrated with other ideas. True, at any given period, there may be high
authorities whose opinions or beliefs are almost instantly accepted. However, it is the scientific community which
determines who are these authorities and when their opinions are accepted. There are cases where one or two
authoritative figure dominates in a given field. For example, in the 1830s40s Berzelius was "the arbiter of the
chemical world" (Mulkay, 12). Murray Gell-Mann and Geoffrey Chew where the "gurus" of particle physics in the
Sixties. On the other hand, in some cases the opinion of a very prominent authority is not accepted. For example,
Einstein's objection to the indeterministic nature of physical phenomena according to the newly created quantum
mechanics, and Rutherford's rejection of the idea of the quantum as a physical idea were ignored.
In order to clarify the epistemic situation in science, the following question is imminent: What is the evaluative
criterion which guides a scientist when he proposes a solution to a scientific problem or a new theory? Is it the
criterion of truth? Does he ask himself whether he has discovered some objective truth about nature or, perhaps,
whether his proposal will be accepted by the scientific community? As is very well known to active scientists or to
people who are acquainted with the patterns of behavior in the contemporary scientific community, the criterion of
acceptance is the first one which comes to the scientist's mind. This is by no means opportunistic behavior. Indeed,
for a scientist who has internalized the norms of the scientific community, the criterion of truth is acceptance by
the scientific community; this criterion is institutionalized. By "acceptance" I do not mean here simply acceptance
by all or most scientists in a democratic manner, but a kind of institutional acceptance which is expressed, for
example, by acceptance for publication in respectable journals, endorsements by leading scientists and inclusion in
textbooks. However, acceptance admits of degrees.
The social theory of science which will be expounded in this chapter will provide us with an explanation for the
social or collective nature of scientific knowledge. The main conclusion of this social epistemology of science is
that in the environment exposed by modern science, the individual scientist cannot rely on his cognitive capabilities
in arriving at an explanatory theory or in judging whether a theory is good or bad. My main epistemological thesis
is that in this case the scientist relies for good reasons on the collective wisdom of the scientific community.
Thus, in science we encounter an epistemic situation which seems to be diametrically opposed to traditional
epistemological conceptions. One of the main issues, if not the prime one, of the traditional theory of knowledge is
the criterion of truth, i.e. the method or the way of determining whether a claim is true. Cardinal Mercier, as cited
by Roderick Chisholm (1982, 63), says in his Criteriologie Generale Ou Theorie Generale de la Certitude that if
there is a cri-
< previous page
page_192
next page >
page_193
< previous page
page_193
next page >

Page 193
terion of truth, then this criterion should be, among other things, "internal." This implies that the criterion of truth
should not be provided by external authorities. For example, in our quest for truth, we would not rely on criteria
provided by our teachers or leaders or by the holy scriptures. Of course, we might be influenced by external
sources but in each case we are the final judges as to what to accept as true. The reason for accepting what the
external authority says must be our own reason. According to this epistemological principle, the criterion of truth
or plausibility cannot be an acceptance by the scientific community. However, for a scientist who has internalized
the norms of the scientific community through the process of scientific education and socialization, the criterion of
acceptance by the community becomes an internal criterion for him; he really believes that what is accepted is true
or plausible. Thus, just as according to the empiricist view, a belief is true only if it can be derived from, or based
on, sense experience, so in science a belief is true or plausible only if it is accepted by the scientific community. As
they stand, both truth criteria have no a priori justification.
What, then, is the difference between the tribesman who internalizes the truth criterion of believing whatever the
witch doctor or the chief says and the scientist who believes whatever is accepted by the scientific community?
Modern Western society would treat the tribesman's "criterion" as irrational, whereas science would be regarded as
the most rational human enterprise. The difference between the two attitudes may be found by employing the
notion of adaptability. The tribesman kind of behavior is appropriate for meeting the environmental conditions
prevailing within the boundaries of his restricted living spacethe tribal habitat. However, his truth criterion may be
detrimental under different environmental conditions. From the vantage point of the evolutionary POR, as we will
see, rationality requires adaptability to changing environmental conditions. If we equate rationality with
adaptability, we may explain why the tribesman truth criterion is irrational. Science is rational since it is adaptable
to a variety of environmental conditions. Science adapts to the artificial technology-intensive environment created
by science itself. The adaptation is achieved by employing highly abstract theories and by using elaborate
experimental and observational techniques which help scientists in testing and improving their theories.
Furthermore, the process of testing and improving the theories brings about further extension of the environment.
Hence, adaptability is built into the ongoing process of the growth of scientific knowledge.
Now, what is the role of epistemic cooperation in the process which makes science adaptable to extended
environments? An answer to this question will provide us also with an answer to the question: why is it rational to
replace individual judgments in science with epistemic cooperation. Cooperation brings about efficiency in the
acquisition of knowledge. However, efficiency is not epistemically essential. Cooperation would be epistemically
< previous page
page_193
next page >
page_194
< previous page
page_194
next page >

Page 194
essential only if it leads to discoveries that would not have been possible without it. Cooperation would be
epistemically essential, for example, in a hypothetical community where none of the members possesses all the
natural sensory or cognitive capabilities, which are possessed by all the members taken together. In science, it
seems the situation is not like this. In principle, a single gifted scientist can conduct all experiments and develop all
possible theories which are conducted and developed by the whole scientific community. Practically it is
impossible. However, traditional epistemology is not interested in pragmatic contingencies, such as the time taken
for making a discovery, or the availability of a human general-purpose supercomputer.
One possible reason why the social dimension might be essential is that the selection should not be made by the
scientist who proposed the idea. Thus, perhaps there should be a division of labor in this respect. However, even
this is not epistemically necessary. Indeed, a single Descartes would attempt to question and undermine any belief
he has; a single scientist may have both creative and critical faculties. Another widely accepted reason for the
importance of the communal acquisition of knowledge is that a community may eliminate personal bias. However,
there is no way to be totally objective; who will eliminate communal, cultural or, in general, human bias?
Yet, as I will argue in this chapter, the reasons for the epistemic indispensability of the social dimension of science
lie in its indispensability for science as an evolutionary process. Hence, epistemologically, the cooperation and the
social dimension of science cannot be divorced from the evolutionary view of science.
Now, I would like to list and to summarize some of the implications of epistemic cooperation for the process of
scientific discovery.
a) The process of scientific discovery is a collective process and the product of discovery is a collective, or social,
product. Hence, whenever someone proposes a hypothesis, he loses control of it since it is processed by other
members of the scientific community, including future generations of scientists. We can view the process as an
autonomous phenomenon. Thus, the process and the product of discovery are independent of the individual
scientist's mind. This reminds us of the realist's claim that the object of discovery is independent of the discoverer's
mind. For example, the idea of the quantum developed independently of Planck's intentions, and quantum
mechanics is a product of the collective efforts of Planck, Bohr, de-Broglie, Schroedinger, Heisenberg, Born, Pauli,
Dirac, Bohm and their followers. In this sense, collective discovery can be categorized as an involuntary
phenomenon on the social level, similar to the process of chance permutation or incubation which operates on the
intrapsychic level.
b) No scientist can know that he made a discovery before he has submitted it to the scientific community. An
essential part of the process of discovery is the
< previous page
page_194
next page >
page_195
< previous page
page_195
next page >

Page 195
processing of the proposed hypothesis by the scientific community. Only when this processing is terminated with
acceptance, is the process of discovery terminated. Hence, strictly speaking, even in case a proposal is accepted
intact, the discoverer is the whole scientific community, rather than the scientist(s) who proposed the hypothesis.
c) There is an implication of b) for the issue of realism. We might say that scientific discovery is made not merely
in nature, but also in the scientific community. This means that the object of discovery is not merely a natural
object, phenomenon or law; it is also a social product. This leads to a radical epistemological aspect of the
''socialization" of scientific discovery. When I say that scientific discovery is made "in the scientific community," I
mean that we discover something which happens in the scientific community; we discover what ideas or
hypotheses are accepted in the scientific community and how they develop. The totality of the accepted ideas and
theories constitutes the picture of reality formed by science. We may metaphorically describe this sense of
discovery by saying that the scientific community is an instrument by which we (human society) investigate the
world. This instrument is our only information channel from the external world. We do not have any alternative
way to observe or investigate the external world, either directly or indirectly. Hence, our only way of obtaining
information on what is going on out there is by observing the image or picture of the external world which is
generated or formed in the instrument. We discover, therefore, the image or the picture of reality, rather than
reality itself. According to this view, science is a "mirror of nature" (I employ this expression as it was used long
before Rorty). As we have already seen, this mirror is not a passive reflecting device. It is rather a mirror which
effects the color and the shape of the image. Hopefully, it is not a distorting mirror. The major question concerning
realism is whether a discovery in the scientific community reflects something real in the external world, or does it
only reflect the internal dynamics of the scientific community.
However, is there a real difference between this manner of viewing scientific discovery and the way we view an
ordinary discovery that we make in our ordinary experience? Viewing ordinary observation from an objective
point of view, i.e. from a vantage point external to the observer, we may say that the observer becomes directly
aware of the mental picture formed in his mind, not of reality itself. Yet, the observer, from his subjective, or
internal, point of view identifies the picture with reality itself. The difference between the mental picture formed in
the individual's mind and the collective picture formed in the scientific community is that with respect to the latter
no one has an internal vantage point; the picture is external to every member of the scientific community, although
all of them participate in the process of forming the picture. This is analogous to the following situation: none of
the members of a large group of soldiers or gymnasts in a stadium can directly perceive, or be directly
< previous page
page_195
next page >
page_196
< previous page
page_196
next page >

Page 196
aware of, the formation he is helping to create. We can discover the shape of the formation only from a bird eye's
view, which is an external point of view.
In the social game of science, a necessary condition for making creative discovery is the exposure of the collective
picture formed in the game. Indeed, a scientist who participates in the game is imprisoned in the collective picture
thus formed and he cannot intentionally change the picture without being aware of it. Hence, a creative or a radical
discovery, which has been generated in the scienitific community, can be exposed by someone who is capable of
ascending above the game of science or by someone who does not participate in the game.
Moreover, when we say that the creative discoverer proposes something that no one has thought about before, we
refer to the typical case when everyone is looking for the solution or the explanation in a traditional direction,
whereas the creative discoverer is looking in a new direction. This means that creativity is derived from the ability
to break out of the prevailing framework of conceptions and presuppositions. Sometimes the current paradigm is so
entrenched that most scientists are not aware of it. Indeed, a precondition for breaking out of the framework of tacit
beliefs, assumptions and conceptions in which the community of problem-solvers is imprisoned is identifying it.
This can be done only by someone who is capable of ascending above the framework, taking an external vantage
point. Since the framework is community-dependent, we may conclude that the creative discoverer should be a
non-conformist or someone who tends to be socially an outsider to the scientific community. Social isolation or
social non-conformism characterizes many revolutionary discoverers. The famous example of Einstein at the patent
office is well known.
Mulkay suggests a possible social mechanism of innovation in science (Mulkay 1972). Innovation is built into the
structure of the scientific community. Deviation from current orthodoxy is usually discouraged in the scientific
community. Yet, the very social processes which are responsible for maintaining intellectual conformity in the
scientific community generate also the conditions for innovation. One of the main driving forces in science is the
need for professional recognition. Scientists share the results of their research with their collegues. In return, they
receive recognition which is mainly given to contributions which conform to current "cognitive norms." So when a
scientific field matures and provides fewer opportunities for recognition by the scientific community, scientists
working in the field tend to look for other fields with more rewarding problems. This leads to the phenomenon of
intellectual migration mentioned in the last chapter. Scientists migrate from areas of declining interest into those
with greater opportunities for recognition. In their new area of research they apply the tools or techniques they
acquired in their old field for solving the new problems they face. The tools they carry with them to the new field
are specific ideas, theories or theoretical methods as
< previous page
page_196
next page >
page_197
< previous page
page_197
next page >

Page 197
well as experimental techniques and instruments. The resulting cross fertilization of ideas is a major source of
innovations.
Mulkay discusses another source of innovation. "Significant innovations emerge mainly at the top and the bottom
of the status hierarchy of science" (ibid., 8). Within a given community, innovation comes from scientists who are
new to the area and who have not acquired a strong commitment to the prevailing norms. They have little to lose.
And scientists at the top of the hierarchy can take risks since their reputation is well established.
However, many great discoveries were made from within the current paradigm. As we saw in Chapter 5, some
great discoverers started with problems which arose in the current paradigm, not being aware of the fact that they
started a process that eventually broke out of the framework. Kuhn maintains that "only investigations firmly rooted
in the contemporary scientific tradition are likely to break the tradition and give rise to a new one" (Kuhn 1963,
343). They do it, however, unintentionally.
Planck's example raises the question of whether creativity in scientific discovery can always be attributed to an
individual discoverer. The discovery of quantum mechanics, for example, was a historical process which cannot be
attributed to a single discoverer. Thus, as I have already indicated, major discoveries are sometimes historical and
collective processes which only in hindsight can be recognized as such. The individual participant in the process
may not be aware of the full significance of the process. In this case, our esteem should be directed to the whole
scientific community, rather than to individual scientists.
The social dimension of creativity is, therefore, evident. On the one hand, in some important cases a creative
individual is socially a non-conformist, a migrant or an outsider. On the other hand, many major creative
discoveries are collective and/or historical products. In other words, they could emerge only unintentionally from
the activity of conformist scientists. What can be concluded from these two social characteristics of creativity is
that there is a tension between creativity and social conformity in normal science. The two ways to overcome this
tension is when the discoverer is a non-conformist individual, or an outsider, and when the process of discovery is
a collective-historical process which transcends the intentions of individual scientists participating in it.
7.2 The Social Dimension of Blind Variation, Selection and Dissemination
One source of blind variation in science is the intrapsychic process of chance permutation. A second source is the
social dynamics of science. As we have
< previous page
page_197
next page >
page_198
< previous page
page_198
next page >

Page 198
observed, unintentionality and serendipity, i.e. blind variation, is built into the social dynamics of science. Due to
the the cooperative nature of scientific work, when a scientist proposes an idea to the scientific community, he is
not anymore in control over the idea, which is processed by the community synchronically and diachronically. The
end product may solve a problem the originator of the idea has not dreamt about; in suggesting the idea, the
originator was "blind" to the problem that is eventually solved in the process he initiated.
The mechanism for eliminating unfit variations which is part of the natural selection scheme also has a social
dimension. The elimination of unfit hypotheses is not a purely logical act as Popper claims. As is well known to
scientists, and as has been shown by philosophers (such as Duhem and Quine), a hypothesis can be protected from
refutation by observational data which contradict it by introducing ad hoc assumptions or by modifying some other
domains in the corpus of knowledge. Of course, such ad hoc maneuvers may complicate the theory which will lose
its original simplicity, without gaining any new information. However, there is no universal standard for deciding
when the auxiliary assumptions and modifications are unacceptable. It is not a matter of purely logical refutation
which leads to abandoning an hypothesis. Here again, the social dimension is operative. The mechanisms of
eliminating and accepting new hypotheses are regulated by social institutions such as scientific societies, the award
system and the funding agencies, which promote certain research projects and reject others, and scientific journals
and conferences, which accept certain research papers and reject others. The policy of these institutions reflect the
norms and methods prevailing in the scientific discipline. These norms are context-dependent rather than universal.
However, the acceptance or rejection of an idea or a theory may be a result of the social dynamics in the scientific
community, having nothing to do with any explicit norms.
If we try to view scientific decisions in a traditional methodological fashion, as an objective matter, then we may
conclude that the major factors contributing to the survival of a hypothesis or a theory which is under evaluation
are its problem-solving capability or the extent it fits the observational data. Problems may arise when trying to
explain particular phenomena or events, or when trying to relate the hypothesis to other theories. Hence, we may
define the environment of a developing theory as consisting of (a) the data which the theory ought to explain, (b)
other developing theories which attempt to explain the same data and (c) established theories with which the theory
should conform. This is analogous to the fact that the environment for an organism consists of physical
environment and other organisms. However, if we do not ignore the social dimension of science, we must take into
account the fact that the decision whether the given hypothesis is well adapted to the "environment" is made by the
scientific community, which does not necessarily follow objective norms. Thus, the mechanism of weeding out
unfit varia-
< previous page
page_198
next page >
page_199
< previous page
page_199
next page >

Page 199
tions is provided by the dynamics of belief-formation in the scientific community. Yet, the individual scientist
makes the first selection; he proposes only selected ideas to the community. Ideas proposed by individual scientists
are subsequently selected by the scientific community. Similarly, mechanisms for generating variations operate
both on the individual and the social levels.
The mechanism of dissemination of ideas and theories is clearly a social process, which encompasses the education
and publication systems. It neither belongs to the context of discovery or generation nor to the context of
evaluation. Nevertheless, dissemination, according to the evolutionist POR, is an essential component in the
evolution of science.
Thus, in the evolutionary POR the D-J dichotomy is converted into a trichotomy: the D-J-DS (discoveryjustification-dissemination) distinction. However, as there is not a sharp line separating discovery and justification,
so there is no sharp line separating justification and dissemination. Dissemination does not always temporally
follow justification or evaluation. Dissemination may be part of the evaluative process and vice versa. Since
evaluation has an essential social component, one of the determining factors in evaluation is the effectiveness of
the communication channels through which an idea is disseminated. If an idea is disseminated through an
established and influential publication system, and presented in prestigeous conferences, it would have more
chances to be accepted. The publication and communication systems are not regulated by objective norms and
standards of evaluation. Since an essential part of evaluation is persuation, effective dissemination is part of the
rhetoric which contributes to the process of evaluation and confirmation. An idea is accepted if most influential
scientists are persuaded to accept it. Thus, there are scientists who have more access to these systems than do
others, for various reasons, not the least of which is that they themselves are in control of tools of dissemination.
A logicist or coginitivist POR would not attribute an essential role to dissemination, since according to both
approaches science can in principle be done by a Robinson Cruso. In an evolutionist POR dissemination is
essential and is linked with the social dynamics of science.
7.3 Has Science Liberated Humankind from the Tyranny of the Genes?
7.3.1 Genetically Controlled Human Understanding
Viewing science as an evolutionary process sheds light on its epistemological significance. Evolutionary
epistemology tries to explain why science is so successful and reliable in view of its evolutionary character.
However, science
< previous page
page_199
next page >
page_200
< previous page
page_200
next page >

Page 200
emerged from our cognitive capacities and from our ordinary experience. Hence, science is restricted by, and relies
on, prescientific intuitions and commonsense which are guided to a large extent by our "hardwired" cognitive
apparatus or innate capacities which are common to all humans. Hence EE should try to explain first why our
genetically based cognitive apparatus is a relatively reliable guide for learning from experience and for our
orientation in the world, so that we have relatively high chance to survive.
For example, on the basis of our past experience, we make predictions about events we have never observed, and
miraculously most of these predictions turn out to be trueotherwise we would not survive. Why it is the case, for
example, that all (or most) of humankind's expectations that "the sun will rise tomorrow" have turned out to be
true? Induction cannot be logically justified. Furthermore, it cannot be justified by referring to the fact that most of
our inductive inferences have succeeded in the past, since this argument relies again on induction. Nevertheless, we
do not have to accept the following claim made by Alfred Ayer (1952, 73): "we shall have to accept it as a
mysterious inexplicable fact that our thought has this power to reveal to us authoritatively the nature of objects
which we have never observed." Even if induction cannot be logically justified, we still can treat it as a natural
phenomenon which is by no means inexplicable. Indeed, EE attempts at providing an explanation for this fact. EE
does not provide us with a justification, in the traditional sense, of beliefs based on induction. For example, it does
not supply us with a warrant for the prediction that the sun will rise tomorrow, or that the chair I am sitting on will
support me in the following second. However, it may provide us with reasons why these beliefs are rational, or
more rational than their negations. In addition to providing the basis for our inductive expectations, our innate
cognitive capacities may provide the basis for other kinds of expectations and conceptions which seem to us
necessary for our thought processes and for comprehending the world around us.
Biology can give an explanation for the miracle of the special adaptation of our "hardwired" cognitive apparatus to
the world. In the course of evolution, Homo sapiens and its predecessor species developed tools for acquiring an
orientation in the world and for learning from experience. These tools are genetically based and are inherent in
every individual. There is no guarantee that these tools enable us to perceive reality correctly, but they have
survived during the process of natural selection, hence we can assume that they have some advantages with respect
to their adaptability to the world.
When we try to explain the origin of our cognitive apparatus in the light of evolutionary theory, we can say that the
information which is responsible for the modes of perception and learning from experience have developed in the
evolutionary process where the DNA has been shaped by natural selection as a model of the environment. Thus,
human beings are born with a potential to develop a cognitive apparatus, which is destined to absorb the raw
material
< previous page
page_200
next page >
page_201
< previous page
page_201
next page >

Page 201
of sense impressions. Our innate capacities are not explicit ideas or beliefs. Rather they are dispositions or
potentialities which are actualized in the course of our experience and development. Namely, one becomes aware
of some genetically determined expectation or belief when one is exposed to some relevant experience, including
cultural experience, such as learning. For example, one may become aware of a basic logical rule after some
reflection or after being taught. The inborn schemes include the basis for the ways we learn from experience, e.g.
by induction, and the basis for our tendency to expect regularities in nature and for our belief in causality. It is also
plausible to assume that some of the basic elements on which classical physics is grounded stem from genetic
information which is common to all human beings. This is because classical physics was shaped in an environment
which did not much exceed the environment in which evolution of the hominid line took place. Indeed, Piaget
claims that the child discovers in a certain developmental stage the laws of mechanics, so we may say that the
child, through his activity in his environment, exposes his innate mental capability to comprehend these laws.
By assuming that there is a common genetic information responsible for some of our cognitive capacities which
has evolved through the evolutionary process, we may answer the question of how we can arrive at successful
theories such as evolutionary theory, Newtonian physics or the classical theory of matter. Such an explanation is
needed if we are not in possession of a method or an algorithm for inferring or constructing true or successful
scientific theories from a firm basis such as facts (if we are empiricists) or first principles (if we are rationalists).
I should add that it would be an exaggeration to assume that the world picture of classical physics is solely
genetically based. The classical world picture certainly draws some of its basic concepts and ideas from those
developed in cultural evolution and which are not genetically based. In any case, it is difficult to disentangle
genetical influence from cultural influences on concept formation and on our basic beliefs. However, the fact that
similar concepts and ideas have appeared independently in different cultures and that all natural languages have
common deep structure (Chomski 1957, 1966) may indicate that there are genetically based ways of learning from
experience and of comprehending the world. For example, inductive expectations and the concept of object are
perhaps universal. In fact, higher animals, too, behave inductively and they seem to behave as if they see objects in
their surroundings. Maybe also the concepts of space, time and force are transcultural. Also the method of
exploring the environment by intervening (e.g., by active experimentation) is characteristic of the whole human
species. The mechanistic outlook, on the other hand, seems to be dependent on European culture in the seventeenth
century. Some would claim that Aristotelian physics is more intuitive than Newtonian physics. However, nowadays
the world picture of classical physics does not clash with our genetically and culturally based intu-
< previous page
page_201
next page >
page_202
< previous page
page_202
next page >

Page 202
ition. And it is plausible to assume that some of the most fundamental concepts and beliefs upon which classical
physics is erected are inborn in the Piagetian sense; i.e. in the sense that at certain developmental stages we are
capable of comprehending, and even discovering or acquiring, these concepts and beliefs, provided we interact with
our physical and social environment. We do not have to accept the detailed theory of Piaget about a child's
development. I only endorse here the general principle that the actualization of our innate capacities is a process
which takes place only when we are exposed to experience, when we interact with an environment of the type in
which these capacities evolved or by interacting with our sociocultural environment, e.g. by learning.
Yet, the assumption that the genes responsible for the development of our cognitive apparatus were shaped in an
evolutionary process will not provide us with an explanation for the fact that modern natural science is successful
in those domains of experience which exceed the conditions under which organic evolution took place. It is
estimated that human anatomic and physiological constitution has not changed much in the last fifty thousand
years. There is no reason to believe that human genetically based cognitive capacities have changed much either.
Thus, "the actors in modern technical society are products of the past, of times and ways of life long gone"
(Washburn & Lancaster 1968, 221).
The actors in modern science are no better off; stripped of all cultural influence, they would not differ from the
hunters and cave dwellers of tens of thousand years ago. Nevertheless, modern physics, for example, is quite
successful in dealing with objects and phenomena on the microcosmic and cosmic scales. In these domains, the
physicist obtains his data by using high energy technology and high power telescopes, radiotelescopes and other
devices which extend his sensorimotor organs, exposing environmental conditions which are radically different
from those prevailing in our natural habitat, in which our ancestors, the cave-men hunters, evolved. Even when we
include the culturally shaped environment of industrial society in the environment where our genetically and
culturally based cognitive apparatus evolved, still the boundaries of the environment in which modern science
evolves are much wider. Hence, our genetically and culturally based cognitive capacities are inappropriate for
guiding us in modern scientific inquiry.
Indeed, the concepts and theories of modern physics are much more remote from our intuitive grasp than those of
classical physics and the commonsense. For example, the concepts and principles of quantum mechanics, such as
the wave function, the wave-particle duality or the uncertainty principle, are not easily comprehensible to our
intuition. In high energy physics, or particle physics, which stands in the frontier of science, the theoretical
concepts are extremely remote from everyday concepts. The quark, for example, departs from the everyday notion
of an object or a classical particle, more than
< previous page
page_202
next page >
page_203
< previous page
page_203
next page >

Page 203
does the ordinary quantum-mechanical wavy particle. Other examples are the notion of quantum field, guage field
or superstring.
The conceptual difficulties arising in the interpretation of quantum mechanics may be related to the fact that our
cognitive apparatus was shaped in the mesocosmos, which consists of medium-size objects and everyday
velocities, distances, temperatures, energies and so on. Namely, our cognitive apparatus is capable of
comprehending only mesocosmic phenomena which can be explored or exposed by our natural sensorimotor
organs aided by prescientific tools and technological devices. Indeed, our cognitive apparatus coevolved with these
organs and tools, which determine the boundaries of our natural habitatthe mesocosmos. It is therefore capable of
comprehending this environment only.
7.3.2 Transcending Our Natural Habitat
Serendipity enables science to break through the boundaries of the prevailing paradigm. It is only through
unintentional acts that a prevailing framework of beliefs can be transcended by conformist scientists. One of the
way an unintentional development takes place is via a collective or a historical process, in which the final problem
which is solved is different from the original problem which triggered the process. Serendipitous discoveries can
also be made by an individual scientist. However, as I will argue below, our main hope of freeing ourselves from
the framework of beliefs in which we are imprisoned is through serendipitous processes with socio-historical
dimensions. The process of transcending the Ptolemaic paradigm, initiated by Kepler, and the process of
transcending the Newtonian paradigm, initiated by Planck are two of the most salient examples of this
phenomenon. The product of a serendipitous process should be comprehended and accepted by the scientific
community before it is recognized as a discovery. My claim is that only the social dynamics of science can yield a
discovery which transcends an established paradigm. This is in particular true if the resulting discovery leads to
comprehending a new environment which significantly departs from our natural habitatthe mesocosmic
environment.
The clue for understanding how science progresses in explaining cosmic and microcosmic phenomena, for
example, can be found in the social dimension of science and in the manner by which science continues the
process of organic and cultural evolution. Our cognitive apparatus is mainly a product of organic and early cultural
evolution. Its limitations in comprehending wider environments can, therefore, be overcome by the continuation of
the evolutionary process on the socio-cultural level. The latter does not bring about modifications in human brain
and cognitive capacities. Our cognitive apparatus is essentially the same as that of our prescientific ancestors; an
offspring of a Planck or an Einstein is born with genetically based potential for developing
< previous page
page_203
next page >
page_204
< previous page
page_204
next page >

Page 204
cognitive capacities which are essentially the same as that of an Aristotle or a Ptolemy.
Of course, there is a variability in the gene pool of the human species. Perhaps some people have useful dormant
genetically based, or even culturally based, mental variations which can be activated only when the appropriate
environmental conditions emerge, e.g. when those people are exposed to an appropriate branch of science, to new
data or to a new mathematical apparatus. However, when such dormant variations become successful, they do not
spread biologically in the human population since they do not endow their owners survival value or reproductive
advantage; they just spread within science. Ideas are eliminated or disseminated by cultural means. Hence the
modifications of human cognitive capacities occur on the superorganic, or cultural, level. Culture and science
develop novel conceptual systems and world pictures which are not genetically inherited but can be acquired by
every individual via learning. This is why science is not adaptive for individual human beings. However, it is
adaptive for human society and this is what is important, since the evolution of humanity takes place at the present
stage on the sociocultural level. I would not accept, therefore, the following claim made recently by Michael Ruse:
Note that I am not saying anything so crude as simply that science is adaptive and that which we consider
better science is more adaptive than worse science. This is obviously false. Mendel, to the best of one's
knowledge, died childless and yet in respects he had a better grasp of the nature of heredity than any of his
fellows. Darwin, to the contrary, had many children but this had nothing whatsoever to do with his
brilliance as a scientist. (Ruse 1989, 193)
Other things being equal, a society which benefits from Mendelian genetics, or from modern biology, in general, is
more adaptive than a society which does not. There is no correlation between the adaptiveness of science and the
biological adaptiveness of single scientists. Mendel was childless but he fathered ideas which survived in science
and which contributed to the fitness of society at large.
Every individual who is brought up in a given culture acquires and assimilates since early childhood and in his
formative years the general world picture prevailing in that culture. Hence, this world picture is deeply entrenched
in his mind; he looks at the world through it. The hard core of the world picture is genetically based, in the sense
that every human being has an inborn potential for developing the expectations, concepts and beliefs constituting
that hard core. Furthermore, the world picture which was developed in any culture is adapted to the mesocosmic
environment. It is true that, to a certain extent, the scientific world picture penetrates the culturally based world
picture, but the
< previous page
page_204
next page >
page_205
< previous page
page_205
next page >

Page 205
average non-scientist does not look at the world through the updated world picture of science. Hence, we may say
that our genetically and culturally controlled cognitive apparatus limits us to comprehend the mecocosmos only. A
different view is proposed by Ruse who maintains that it is sensible ''to suppose that our reasoning abilities are, in
some very real sense, rooted in our biology" (ibid., 203). Ruse does not distinguish between ordinary reasoning and
scientific reasoning. For him the reasoning processes which led to quantum and relativistic physics are based on the
same logic which guides prescientific and everyday reasoning. Science is governed by commonly accepted rules
and criteria which "we humans use because they proved of value to our ancestors in the struggle for existence"
(ibid., 193).
However, although ordinary reasoning and scientific reasoning have some common core, it is implausible to
assume that the evolution of modern physics, for example, can be explained by this common core alone. Thomas
Nagel expresses this question in the following illuminating passage:
The question is whether not only the physical but the mental capacity needed to make a stone axe
automatically brings with it the capacity to take each of the steps that have led from there to the
construction of the hydrogen bomb, or whether an enormous excess mental capacity, not explainable by
natural selection, was responsible for the generation and spread of the sequence of intellectual instruments
that has emerged over the last thirty thousand years. (Nagel 1986, 80)
Indeed, something else is needed for comprehending the phenomenon of modern physics. However, Nagel looks
for the "enormous excess [of] mental capacity" needed for creating science in the wrong place. It is not an inborn
capacity. Rather, it is generated in a creative socio-evolutionary process. I will explain this cognitive capacity by
the same model of natural selection, operating on the sociocultural level. In looking for the additional factor
responsible for the growth of science, I will consider the possibility that in science new dimensions, in addition to
reasoning, play an essential role; perhaps natural science is a natural phenomenon in its own right, not just a matter
of reasoning. I shall look for these dimensions in the sociocultural arena.
However, one may object that to the extent that the capabilities of being a social animal and of developing culture
are rooted in humankind's gene pool, then everything humankind is achieving, including science, is "rooted in our
biology," as Ruse claims. But in what sense do we use here the term rooted? If we view science as an evolutionary
process, which is a continuation of human evolution, then we might say that everything developed in this
evolutionary process is rooted in human biology. In this sense of the word, every form or trait which emerges along
an evolutionary line, is rooted in any pre-
< previous page
page_205
next page >
page_206
< previous page
page_206
next page >

Page 206
decessor stage along the line. However, in the same fashion, we would be led to say that every new species or trait
which appears along an evolutionary line is rooted in the biological constitution of its predecessors, so that human
brain and human culture is rooted in the biology of some ape species. But this is a very broad and unilluminating
notion, which will lead us to the following argument: The capacity to do science is rooted in our biology, which, in
turn, is rooted in the biology of some hominid ape or even in the biology of some primordial form of life of which
we are descendants. Therefore, scientific reasoning is rooted in the latter.
However, evolution, including cultural evolution, is not deterministic; chance playing a major role in this process,
including the evolution of humankind, in particular in sociocultural evolution. Thus, if we use the word rooted in a
narrower, and more sensible sense, we would not claim that every quasi-random variation, of any order, imposed
on a given form along an evolutionary line, is rooted in this form. Novelty emerges along any evolutionary line,
and science is one of the novelties which have emerged in our evolutionary line. The evolution of science has been
affected both by our biology and by chance events or new environmental conditions which were exploited
opportunistically in the processthe same as it happens in organic evolution. Thus, the trait of social cooperation of
humankind, which perhaps was originally selected for the benefit of hunting, was later opportunistically exploited
for the creation of science. This does not mean that science is deterministically rooted in prehistorical hunting.
7.3.3 Two Patterns of Human Evolution
Yet, it is not only sheer chance which has been regulating novelty-generation along our evolutionary line. I would
like to suggest that there are two interlocked evolutionary patterns, or principles, which enable humankind to
transcend its natural habitat. These patterns in themselves were probably selected because of their survival value.
One principle refers to the strategy by which our species solves its problems and increases its adaptability: this is
the principle of growth by expansion. The second is concerned with the evolution of the capabilities which enable
humankind to carry out this strategy: the coevolution of the sensorimotor organs and the brain, or of human action
and human cognitive capacities.
a. The Principle of Growth by Expansion The preceding discussion raised a crucial question regarding human
evolution: whether and how human knowledge has broken free from, or has transcended, the limits imposed on it
by its genetically and culturally controlled cognitive capacities. Since our cognitive apparatus is an evolutionary
product, then by considerations of biological adaptation, we would expect human intelligence to be limited to
comprehend-
< previous page
page_206
next page >
page_207
< previous page
page_207
next page >

Page 207
ing only those aspects of reality which are vital to the survival of the human species. However, the question is
what does biological adaptation mean. Does it mean adaptation to humankind's natural undisturbed environment?
Furthermore, is there any biological meaning to such a static concept of environment? We know that the
environment of a given population of organisms is determined to a large extent by the organism's activity. In other
words, the environment is composed of those domains and aspects of the physical and organic surroundings which
the species tries to exploit. Thus, the environment is not independent of the species' traits and behavior. As the
activity of the population extends, the environment changes. This applies in particular to the human species and to
culturally based activity. Humankind explores its environment by actively disturbing it, e.g. by technological
intervention. Hence, humankind's environment is rapidly changing as a response to its activity. These culturally
induced changes are more extensive than the natural changes.
We may say that with the evolution of human extensive motor activity, and with the parallel evolution of the
cognitive capacities, the guiding principle of cultural development became the principle of evolution by expansion.
This means that the human species extends its activity, exposing new domains in the environment and adapting to
the changing environment. Natural science and technology, which continue this process most extensively, have
evolved, indeed, in a rapidly changing environment. The environment exposed by active scientific observation and
experimentation does not just change but expands; more and more phenomena are exposed and more and more
layers and aspects of reality are unveiled.
Thus, when we are asked whether the human mind is capable of comprehending only its natural habitat, our first
response will be that humankind's natural habitat itself is not static and is not limited to the niche in which Homo
sapiens evolved organically. A major trend in humankind is its adaptation to an expanding environment rather than
improving its adaptation to a given environment. However, the dichotomy implied by the last statement is only
apparent. By going beyond given environmental conditions, humanity overcomes problems which have been
unsolvable in the old environment. For example, in order to overcome food shortage, humanity changed its livingspace by agriculture. Another example is from science. In order to solve anomalous phenomena which arose in
chemistry and atomic physics, such as the radioactivity phenomenon, science entered into the domain of nuclear
physics, by developing new experimental and theoretical methods and techniques. Of course, every new "livingspace" raises new problems, some of which cannot be solved and a new "Lebensraum" is sought.
This principle of problem-solving by expansion can be related to C.H. Waddington's view of evolutionary change
(Waddington 1975). Waddington maintains that evolutionary progress (not only of humankind) means progress in
adaptability, rather than progress in adaptation. A species may be
< previous page
page_207
next page >
page_208
< previous page
page_208
next page >

Page 208
well adapted to a static environment. Such an environment does not exert selective pressures that give an advantage
to organisms which are more adaptable. The species may survive a long period in a stable state, but may become
extinct after the environment undergoes sudden changes. The human species does not wait passively for
environmental changes to occur but creates the changes by its own activity. This evolutionary pattern, which is
related to the evolution of the sensorimotor organs and the brain, brought about the high degree of adaptability of
humankind. Humankind's high degree of adaptability is not due only to the plasticity of the individual human
phenotype but also to the high plasticity inherent in the structure of human society. This includes the social
dimension of science and technology.
We often encounter an objection to the claim that science serves as a tool for survival, or that it constitutes a
continuation of biological evolution. The objection relies on the assumption that science expands into areas which
have no survival value for humanity. In other words, it is claimed that science "overdoes" the task of serving as a
tool for survival. According to this argument there is no survival value in theories developed for explaining
phenomena which to a large extent are created by science. Also there is no use in cosmological theories.
Furthermore, science aims at understanding rather than at practical usage. However, in view of what was said
above, a survival value is not attached to the adaptation to a given environment. Science is a method or a tool for
increasing adaptability, and the latter is the trait which has a survival value. Science prepares us for future hazards
or future opportunities to which we will more quickly adapt by employing scientific theories and methods. Since
we cannot anticipate what environmental changes will occur in the future, we cannot know at any given moment
whether a certain theory or an experimental result or technique has a survival value or not.
The phenomenon of serendipity is related to the process of increasing the adaptability of science. Serendipitous
discovery is a mechanism for expanding knowledge beyond any given framework or paradigm. Indeed, scientists
cannot know in advance what new environments will be eventually exposed as a result of their research. In many
cases a new environment is exposed serendipitously as an unplanned by-product of research done within the
confines of the old environment. The new environment provides solutions to problems that were unsolvable in the
old environment. This is, therefore, another important facet of serendipity. The phenomenon of unintended or
serendipitous discoveries, which is built into science, contributes to the adaptability of science.
Our cognitive apparatus which was shaped in the mesocosmic environment gives every human being the basis for
adaptation to this environment, including the basis for comprehending mesocosmic phenomena. Our genetic
heritage also gives us the capability of developing technology-based culture which is characterized by the pattern
of problem solving by expansion. This
< previous page
page_208
next page >
page_209
< previous page
page_209
next page >

Page 209
evolutionary pattern enables humankind to transcend the limits of the mesocosmic environment. However, at the
present evolutionary stage of cultural and scientific evolution the selection pressures exerted by new environmental
conditions, e.g. by the data gathered via scientific experiments, affect only our "idea pool" rather than our gene
pool. This is because ideas compete and "die in our stead" (to use Popper's phrase, 1972, 25). Hence, the resulting
increase in adaptability is not genetically transmitted from generation to generation. It is, therefore, the human
society, or rather the scientific community, which evolves and adapts to the new environmental conditions. The
scientific community with its institutional structure and its collective wisdom is capable of coping with the new
environments exposed by science. With this conclusion in mind, let us turn to the second pattern of human
evolution.
b. The Coevolution of Human Action and Human Understanding The above evolutionary pattern characterizes the
activity of the human species which results in adaptation to an expanding environment. The second pattern
characterizes the evolution of the human species' constitution: the coevolution of sensorimotor capabilties and brain
capacities. This is the very evolutionary pattern which enables humankind to generate the expanding environment
and to adapt to it. Furthermore, I will view sociocultural evolution, including science, as a process which is
governed by the same principle; technology and science extends both our sensorimotor capabilities and brain
capacities. Observational and experimental devices constitute our "extended sensorimotor organs" which generate
or expose the environment in which modern science evolves.
Now, if we want to complete the analogy, the question is what is the "extended brain" which coevolves with the
observational and experimental technologies and which developed the intricate mathematical and theoretical
apparatus of modern natural science? Our extended sensorimotor organs are based on technology. If we look for a
brain-extending technology we shall find, of course, computer technology. However, the most spectacular
achievements of twentieth-century science could not benefit from computer technology; quantum mechanics, the
theory of relativity and molecular biology were invented and developed before the digital computer. Furthermore,
the computer only magnifies our computational capabilities and our capabilities for storing, retrieving and
manipulating bits of information. For the time being, it does not help us much in inventing new conceptual systems
or scientific theories. Only when the concept-formation capabilities of the AI technology will be significantly
improved, might it aid science in inventing new concepts and theories.
What is, then, the extended brain which enabled humanity to invent or discover the great ideas and theories of the
physical and biological sciences? It seems that the answer to this question lies not in the realm of technology but
< previous page
page_209
next page >
page_210
< previous page
page_210
next page >

Page 210
in the realm of social dynamics. This was the conclusion I arrived at in discussing the first evolutionary pattern. I
propose to view the network of brains of the scientific community as the extended brain exhibiting cognitive
capacities which transcend those of an individual human being. The cooperative nature of scientific research is
what makes the progress of modern science possible (Kantorovich 1983, Levinson 1988, Hull 1988). Technology
does play a role in the formation of this network. Indeed, in the twentieth century, with the advance of
communication technology, science has become increasingly more cooperative and at the same time it has
penetrated environments far removed from the mesocosmos. Indeed, as we have seen, cooperation fosters
serendipity and the latter is one of the major driving forces in the evolution of science that enables science to break
out a tradition and to create new environments.
However, cooperative social dynamics fulfills an additional epistemic role in exploring foreign environments
created by this very social dynamics. This is the role of epistemic support. Cooperation is required in everyday life
whenever we attempt to solve a complex problem, or a problem which involves many uncertainties. An individual
who encounters a complex problem with uncertain elements needs support and confirmation from others. A
necessary condition for cooperation in problem solving and advancing knowledge is the existence of common
systems of basic concepts and beliefs. These systems are social products; they evolve as a result of interaction
between human beings and between them and their environment. In everyday life, a system of concepts and
beliefs, or a world picture, evolves as a result of the experience of human societies in the mesocosmos throughout
many generations. A common world picture enables human discourse and further cooperation in problem solving
and knowledge acquisition.
In technology-intensive science, the problem environment is much less familiar and predictable, much more
complex and much more rapidly changing than the mesocosmic environment. Our cognitive apparatus which is
capable of guiding us in the mesocosmic environment is not appropriate to do so in the extended environment.
Cooperation is what is needed for generating the extended cognitive apparatus for guiding us in this environment.
The extended guiding apparatus is the scientific world picture which is developed in an evolutionary process in
which ideas and theories are generated and selected by the scientific community. The notion of world picture as I
employ it here refers to the logic and methods of acquiring new knowledge, as well as to the basic concepts and the
entrenched beliefs and theories through which scientists look at the world. According to EE, all these are products
of the selection process in science. The world picture which thus evolves extends or replaces the genetically and
culturally based cognitive apparatus which evolved in the mesocosmic environment. It guides scientists in
exploring and comprehending the extended environment.
< previous page
page_210
next page >
page_211
< previous page
page_211
next page >

Page 211
The scientific community must, therefore, be so structured as to facilitate cooperation in advancing knowledgein
evaluating discoveries as well as in generating them. Thus, instead of relying on his genetically and culturally
based cognitive capabilities, the individual scientist must rely on the world picture generated by the scientific
community. The scientific world picture is based on the conceptual system by which the extended environment is
described. It includes the system of basic beliefs which guide scientists in constructing and selecting explanatory
theories and general methods of research. This notion is akin to some of the senses given by Kuhn (1962) to his
notion of the scientific paradigm.
The specific social organization of science aims at arriving in an efficient manner at a common world picture, or at
a consensus in the scientific community. According to John Ziman (1968), the goal of arriving at a consensus is a
distinctive feature of science. However, Ziman does not provide an explanation for why this goal by itself is
distinctive and what is its epistemological significance. According to the view I expound here, science indeed
strives at generating a consensus with respect to the basic world picture. The epistemological significance of this
goal is to create a collective guiding apparatus for comprehending the extended environment generated and
investigated by science.
The "collective brain" of the scientific community is not just a collection of brains; it is rather a network of
intercommunicating brains which are interlocked via an intricate social infrastructure which channels the social
dynamics of science towards creating a consensus or a widely accepted world picture. The social structure is
composed of the various institutions of the scientific community which foster cooperation and which support an
appropriate division of labor and specializations. This social infrastructure evolves with the advance of science by
modifying the division of labor, by splitting into subspecializations, by integrating separate fields and by creating
new institutions. The creation of the scientific Academies and Societies signified the emergence of one of the
crucial stages in the evolution of science. Other institutions which have evolved since then are the international
conferences, research funding, the award system and the publication system.
Twentieth-century physics have witnessed extensive changes in the structure of the scientific community, which
reflect the fast evolution of the world picture of science. In physics, for example, there is a sharp division of labor
between experimentalists and theoreticans. In biology, on the other hand, every scientist acts both as an
experimentalist and as a theoretician; theoretical biology, which has emerged recently as a separate discipline, does
not have yet a major influence on the mainstream of current biological research. Groups of experts which handle
new experimental or computational technologies emerges from time to time. For example, in recent years almost
every branch of science employs computer specialistsa function which did not exist in the fifties. Thus, the
computer revolution affected the evolution of sci-
< previous page
page_211
next page >
page_212
< previous page
page_212
next page >

Page 212
ence. In theoretical physics, there are groups of mathematically oriented physicists who are engaged in developing
the mathematical machinery which is applied by theoretical physicists. There are scientists who concentrate on the
foundations of their discipline. Others are engaged in applied branches of their science. All these functions are
controlled by the institutions of the scientific community and by its communication network. This network of
institutions is the "central nervous system" of the scientific community which controls scientific research. This
description may remind us of the organismic model of society which was employed in sociology and social
anthropology in the nineteenth century. This model described the evolution of society analogously to the evolution
of human organs where specialized groups played the role of organs. I do not mean to adopt this model literally.
Rather I employ this model metaphorically in order to make the point that observational and experimental
technologies constitute an extension of our sensorimotor organs and that the evolution of these tools is
accompanied by changes in the structure of the scientific community which uses these tools. The principle of
coevolution of organs and brain is similarly extended to the structure of the scientific community.
Thus far I have described the evolution of science as a coevolution of experimental technology and the social
infrastructure of the scientific community. Now I will turn to the evolution of the scientific knowledge, which is
generated by the "collective brain" and the extended "sensorimotor organs." This will lead us to a social theory of
knowledge, or to social epistemology.
7.3.4. The Epistemological Significance of Cooperation in Science: The Evolutionary Perspective
The theory of science which I outline here is a naturalistic philosophy of science. In a traditional logicist approach,
the social element does not have any epistemological significance. Only logical facts matter. However, the logicist
approach cannot account for scientific explanation. It is a logical fact that any finite body of observational
statements can be deduced from an unlimited number of different theories. Namely, for any finite number of
observational statements for which we have to provide an explanation, we could construct an unlimited number of
theories, in conjunction with initial conditions, from which these statements could be deduced. The problem of
generating an explanatory theory is, therefore, formidable. If scientists had a logical method or algorithm for
inferring or constructing a theory from observational data, then the problem would not arise, but they do not. In
constructing a physical theory, for example, physicists since the seventeenth century have had at their disposal
theoretical entities such as epicycles, vortices, media, fluids, powers, forces, poles, fields, waves, particles, strings,
charges and radiations in any number, combination and configuration, to mention but a few. The theoreti-
< previous page
page_212
next page >
page_213
< previous page
page_213
next page >

Page 213
cal system can always be adjusted to yield the requested predictions. In the history of humankind's attempts to
explain natural phenomena we encounter also prescientific or mythical entities such as sympathies, attractions,
tendencies, natural loci, ghosts and spirits and the list is only beginning.
As we have seen, although there is an unlimited number of possible hypotheses, in most cases scientist have
difficulties in finding even one explanation. The anarchistic policy recommended by the slogan "anything goes"
(Feyerabend 1978) has never been practised in modern natural science. Logic alone cannot tell why the Ptolemayic
system was replaced by the Copernican. Logic alone cannot distinguish between acceptable and unacceptable
explanations. Logic also cannot prevent scientists from inventing any theory they wish, provided it is selfconsistent and consistent with the data. When we add methodological requirements such as simplicity and high
predictive power, we may reduce the number of possible theories. However, we face the problem that there is no
unique way of measuring these methodological properties (we cannot measure simplicity or count predictions and
weigh them). Scientific explanation requires extra-logical criteria and standards for choosing plausible theories
among all those which logically account for the data.
Thus, in actual science not every logically possible theory is accepted as an explanation. The logical requirement
that the observational data should be deducible from the theory in conjunction with initial conditions is a necessary
but not a sufficient condition for explanation. The additional requirement for an explanatory theory is that it should
conform with the established body of knowledge, including the prevailing world picture, and with the general
extralogical criteria. At any given period there is a general conception of what an explanatory theory should look
like, and a repertoire of exemplary models of explanation, which narrow the range of acceptable hypotheses. For
example, in the nineteenth century, the mechanistic-curpuscularian picture dominated physics; mechanics was
regarded as the paradigmatic theory. Hence, attempts were made to construct explanatory theories in fields such as
light, electricity and magnetism in terms of mechanical models. Lord Kelvin, a very respectable physicist,
proposed, for example, the following model for the structure of the luminiferous aether, which was supposed to be
the weightless elastic medium carrying the action of the electric and magnetic forces:
...a structure is formed of spheres, each sphere being the center of a tetrahedron formed by its four nearest
neighbours. Let each sphere be joined to these four neighbours by rigid bars, which have spherical caps at
their ends so as to slide freely on the spheres. ... Now attach to each bar a pair of gyroscopically mounted
flywheels, rotating with equal and opposite angular velocities... (Whittaker 1951, 145)
< previous page
page_213
next page >
page_214
< previous page
page_214
next page >

Page 214
All these gadgets were supposed to be massless and to fill up all space, including the space inside and between the
molecules of this paper. It should be emphasized, however, that the notion of world picture as we use it does not
necessarily refer to a naive realist picture of reality; Lord Kelvin, for example, did not believe that his machinery
literally occupies all space. The world picture, as we conceive it, determines the repertoire of acceptable
explanatory models, whether they refer to reality, or just serve as explanatory devices. The repertoire of
"legitimate" explanatory models of contemporary physics does not include the kind of explanation employed by
Lord Kelvin. However, the contemporary world picture is not logically necessary; there is no logically valid proof
which shows that it is preferable to the mechanistic world picture which guided Lord Kelvin.
Without the extralogical requirements for explanation, e.g. that an explanatory theory will conform to the world
picture, science will have too many paths along which to proceed and progress will be impossible. The further
removed we are from the domain of immediate sense experience or from the mesocosmic environment, the less
adequate becomes our genetically and culturally based cognitive apparatus, leaving us with too many possible
explanations for any set of phenomena. Thus, without a guiding world picture, we would be totally at loss in the
new environment. Hence, narrowing the range of possible explanations constitutes a necessary condition for
scientific progress.
It is not logical inference which generates the world picture of science. At any given period the scientific world
picture is a product of a process of selection. The individual scientist, who cannot rely on his cognitive apparatus to
orient himself in the foreign and complex environment of modern science, turns to the world picture thus formed.
The world picture is a collective entity which emerges as a result of the mental and experimental activity of many
intercommunicating scientists. An individual scientist does not rely only on his own experiencehe relies mainly on
the experimental and theoretical results achieved by other scientists. For example, an individual physicist does not
directly encounter by his own sensorimotor organs the environment explored by modern physics; he relies on the
work of his colleagues who conduct experiments, i.e. on the collective sensorimotor organs (so to speak) of the
scientific community. And even if he is a member of an experimental team, he might be just one member of a
large group of scientists and technicians conducting an experiment. It is the whole community which collective
encounters the phenomena through complex experimental devices operated by many research groups and through
the application of theories for analyzing and interpreting the results. The philosopher of science expresses this
situation by saying that modern physics investigates theoretical entities which cannot be directly observed or
sensed. According to the above outlook, it is the scientific community as a collective whole which observes and
interprets these phenomena involving theoretical entities. Thus, theoretical entities are distinguished not only by the
fact that their observation requires sophisticated experimental
< previous page
page_214
next page >
page_215
< previous page
page_215
next page >

Page 215
arrangements and theoretical machinery. They are also characterized by the fact that the act of observing them is
not an act of an individual but a cooperative act. And perhaps one of the measures of the degree of theoreticity of
an entity is the number of scientists and groups of scientists needed for the act of ''observation."
Contrary to what is going on in many areas of philosophy, a scientist is not engaged most of the time in criticizing
his fellow scientists; he relies on their work. I have already mentioned Newton's words to the effect that he stands
on the shoulders of giants. They reflects this attitude; scientists rely on other scientists both synchronically and
diachronically. Thus, the scientific world picture summarizes the collective experience of a community of
scientists. Since the reliance on other scientists' work is so crucial, the scientific community is engaged not only
with investigating natural phenomena but also with surveying the capabilities of its members and "grading" them;
the scientific community is a complex instrument for probing remote territories of nature, hence all its parts or
components should be constantly "calibrated." With the emergence of science as a cooperative enterprise which
relies on experts operating sophisticated technology and employing sophisticated mathematical machinery, this
activity of assessing the performance of the community members has become institutionalized in the award system
of the scientific community, in paper- and citation-counting and in many other formal and informal means of
ranking.
Thus, the scientific community is a social system in which cooperation in acquiring knowledge is institutionalized
and the individual's opinion is not recognized as a legitimate knowledge claim before it is processed by the organs
of the community. Via this processing, a proposed idea is assessed, improved, or modified and integrated with
other ideas, or it is rejected. As we have seen, for the scientist, the ultimate truth-criterion is the acceptance by the
scientific community. According to this view, this is the only possible and recommended criterion, especially in
normal science, since there is no way for the individual scientist to know whether he discovered some truth in the
extra-mesocosmic domains dealt with by modern natural science. A similar view is expounded by Hull (ibid.) who
claims that the main aim of a scientist is in disseminating his ideas rather than in discovering truth.
Just as our genetically based cognitive capacities do not necessarily reflect absolute truth, so is the scientific world
picture. Both reflect those aspects of reality which have something to do with the needs, preferences and interests
of humankind, on the organic and cultural levels.
7.4 The Tension between Change and Stability
The tension between innovation and tradition is built into the evolutionary POR. A proliferation of ideas and
hypotheses is vital for the progress of sci-
< previous page
page_215
next page >
page_216
< previous page
page_216
next page >

Page 216
ence as an evolutionary process, since variability is the source for evolutionary progress; the generation of variant
genes (or ideas) which are usesless at the time of their appearance might be useful for the future adaptation of the
species (or of science) to its environment. Natural selection favors, therefore, variability and adaptability. The
social system of science provides a fertile soil for producing a variety of ideas which constitute the raw material
for the process of selection in science. The strive for originality is a major driving force in the scientific
community. However, originality is restricted by tradition. Hence, the phenomenon of serendipity is essential for
generating ideas which deviate from tradition.
Thus, science is a cultural form which exhibits two seemingly contradictory trends: on the one hand, it narrows the
range of acceptable ideas, and on the other, it fosters variability of ideas. These two complementary trends are
necessary for progress. Conformity with tradition is essential for progress in a given direction; it enables the
exploitation of the potentialities hidden in the existing world picture. Whereas variability is necessary for
enhancing the chances of hitting upon new ideas in order to modify or replace the present world picture by a better
one. We face here, therefore, a dialectic situation in which change and stability play an equally important role. In
organic evolution the parallel situation is that a large number of mutations endanger the phylogenesis of the
species. The mutations should bring about useful changes in the gene pool of the species without undermining the
central characteristics of the species. In cultural evolution we have a similar situation: the deviation from
traditional norms is necessary for cultural progress, however cultural changes should not endanger the information
contained in the cultural tradition, including scientific tradition.
The dialectics between change and stability can be illustrated by J. Bronowski's model of stratified stability
(Bronowski 1970), as applied by Ervin Laszlo (1972) to the evolution of science. According to Laszlo, science can
be treated as an open system, analogous to an organismic system, which develops and grows as a result of
information input from its environment ("nature") through a series of intermediary states, on its way to equilibrium
with its environment. These intermediary states are far from equilibrium. Some of them are more stable than
others, i.e. they resist disturbances from the environment for a relatively long period of time. The information
inflow, which consists mainly of empirical data, enables the system to grow, so that the potentials inherent in the
systemthe hidden strata of stabilityare actualized. In a state of stability the system retains its main characteristics in
the face of disturbances from the environment such as anomalies or experimental data which do not fit the
theoretical system. A stable state of science corresponds to something like normal science with a stable world
picture or paradigm. When the disturbances exceed a certain limit, the system will lose its stability and after a
period of instability will eventually reach a new state of
< previous page
page_216
next page >
page_217
< previous page
page_217
next page >

Page 217
stability. This transition corresponds to a scientific "revolution" or to a radical theoretical change, or to a radical
change in the world picture. When the system climbs up the strata of stability, it becomes progressively more
resistant to disturbances; the new state of the theoretical system accounts for the phenomena explained in the
preceding state plus the unexplained anomalies and new data. Thus, unlike in a Kuhnian revolution, in which the
successive paradigms are incommensurable, a higher stratum of scientific knowledge in the above model can be
compared with the preceding one; the transition to a new state of stability means progress. This agrees, for
example, with the physicists' intuition that the transition from classical physics to quantum and relativistic physics
constituted a progress.
This kind of evolutionary model requires, therefore, a variability of ideas which will enable the system to
accomodate the inflow of information and to climb up the evolutionary ladder. However, when the system stays in
a stable state it attempts to exploit the potentialities of that state of knowledge and to explain as much data as
possible. It would not be wise to try climbing the ladder prematurely, before the present rung is stabilized and fully
exploited, since the next stratum will be erected on the present one. For example, Newtonian physics was not
replaced before its potentialities were maximally exploited. Thus, at a period of stability, i.e. in normal science,
mainly ideas which conform with the world picture should be employed and investigated. Novel ideas will be kept
aside provisionally to be activated at a time of crisis. But the activation of these ideas will not be by intention;
rather it will be a result of a serendipitous event.
This model can therefore provide the basis for the concept of "gradualism" underlying the principle of serendipity.
Serendipitous discoveries are variations generated unintentionally in the course of methodical research taking place
in normal science, which represents a stratum of stability in the evolution of science. Eventually they may bring
about a transition to a new state of stability. A serendipitous development is more likely to occur at a time of crisis,
when the system loses its stability as a result of disturbance caused by some acute problems awaiting a solution.
These problems trigger off an interpsychic "chance-permutation" process in the collective mind of the scientific
community. If we make the analogy with the intrapsychic process, we may treat this process as an "incubation"
process. Thus, when scientists encounter the missing piece in the puzzle, a stable configuration is formed and the
incubation process terminates. We may identify the system of "mental elements" which are distributed throughout
the scientific community as the open system referred to by Laszlo. When a major new configuration is formed in
this manner, the system undergoes a transition to a new state of stability. Thus, the system climbs up the strata of
stability via serendipitous discovery.
< previous page
page_217
next page >
page_218
< previous page
page_218
next page >

Page 218
7.5 Implications for Discovery

7.5.1 The D-J Distinction Revisited
At first sight, it seems that the D-J distinction is valid in the evolutionist POR since there is a clear distinction
between the process of variation-generation and the process of selection. Furthermore, the main import of the
natural selection model is that the way by which a variation was generated is irrelevant to the selection-evaluation
process, in the sense that the environment "opportunistically" selects quasi-random variations.
Since in organic evolution variations are not produced by agents who are capable of acting intentionally, there is no
sense in talking about a mutation which is produced "in order to" overcome an environmental pressure. In science,
however, variations can be produced by intentional acts. Hence, if we believe that genuine discoveries must be
produced "blindly," we would treat favorably only ideas and theories which have been produced without knowing
in advance the phenomena which will be eventually explained or the problems which will eventually be solved by
these theories or ideas. Only in this respect is generation relevant to evaluation. However, in this case information
concerning generation is relevant to evaluation in the negative sense. As we have seen, scientists treat unfavourably
an ad hoc hypothesis, which was produced with the knowledge of the phenomena and the data which it eventually
explains.
Thus, we encounter here the antithesis to the D-J distinction thesis. The D-J thesis, which was inspired by the
empiricist attitude, can be seen as directed against rationalism. In the Cartesian system, a conclusion which is
validly derived from first principles is considered to be necessarily true, irrespective of any empirical observations.
Empiricism implies an opposing view reflected in the D-J thesis: a hypothesis is refuted or confirmed by
observation, irrespective of its source. However, even without being guided by the evolutionist POR, a
straightforward objection would arise: what about the case where the hypothesis was constructed in order to match
already known data or to explain already known phenomena? If the same observational results which were
employed in constructing the hypothesis serve also as the ultimate court of appeal for judging it, something is
wrong with this kind of "justice." The evolutionist POR and the principle of serendipity give us a definite reason
for having this intuitive feeling of "injustice." In other words, the theory of natural selection provides a deep
explanation to this methodological rule. Thus, the empiricist requirement that evaluation should depend only on
observation will lead us to the absurd result that one can "cook" his hypothesis to match the data and continue to
do so indefinitely; whenever some new predictions of the hypothesis do not agree with the observation, one would
adjust his hypothesis to the recalcitrant data. Even if in the course of doing so, the flexible hypothe-
< previous page
page_218
next page >
page_219
< previous page
page_219
next page >

Page 219
sis does not yield any successful predictions, the empiricist requirement will still be obeyed and the D-J distinction
thesis will still be valid, without gaining any genuinely new knowledge in the process. The evolutionist POR and
the principle of serendipity remedy this deficiency by demanding that information concerning the process of
discovery or generation be relevant to the evaluation in the following respect: the hypothesis should be arrived at
blindly or serendipitously. This includes the widely acknowledged requirement that a hypothesis would yield
unexpected predictions which agree with the observation, including predictions of already known effects which
have not been taken into account in devising the hypothesis. The evolutionist theory provides us with an
explanation for this evaluative requirement. According to our justificatory-explanatory scheme, we should make
sure that this requirement or rule is indeed practised in science. If the rule has been practised in most cases, or in
most important cases, we still would not be entitled to recommend it to scientists since it might be an accidental
rule. Only if the rule is backed by an explanatory theory, then it has a normative strength, according to our nonabsolutist conception of justification.
7.5.2 Cultivation: Preparing the Collective Mind
Socio-evolutionary processes of discovery are not controlled by the individual scientist and the products of
discovery cannot be evaluated by him. Thus, the distinctive features of the sociologist POR is that "rules" of
discovery are not directed solely towards individuals but also, and perhaps mainly, towards the institutions of the
scientific community.
If we employ the metaphor of the farmer who cultivates the soil and scatters the seeds, we would say that the seeds
in this process are the ideas which initiate the cooperative process of discovery. These ideas are sown in the minds
of the members of the community. Thus, the scientists' minds are analogous to the ground or the soil in which the
seeds are sown. The process is nourished with other ideas which are already present in the minds of the members
of the community. Some are brought from other fields. This "soil" is cultivated only if two main requirements are
met: First, the members of the scientific community should be knowledgeable in the field so that they can
"nourish" the body of knowledge which is growing out of the original ideas. A major source of "nourishment"
comes from information obtained from ongoing observations and experiments. Second, there should be a
communication network conveying the ideas to all members of the community, and a social infrastructure which
encourages scientists to submit their ideas for public scrutiny through this network. The institutional advice will be
that the scientific community will foster epistemic cooperation by setting up appropriate norms and appropriate
institutions. If these requirements are met, we may talk about the "collective mind" of the scientific community,
which provides
< previous page
page_219
next page >
page_220
< previous page
page_220
next page >

Page 220
the fertile soil for discoveries to flourish; by the act of cultivation, we prepare the collective mind.
The dependence of the final product on observational data that nourishes the process of discovery reflects the
legacy of empiricism. Different data would lead to a different product. The product is also cultivated by ideas,
knowledge and information embedded in the scientific community. However, also the social characteristics affect
the final product. Without an efficient communication network, the process would not benefit from all the ideas
and information spread throughout the community; actually the whole process, which is cooperative in nature,
would not come into being. This social element is absent from traditional empiricism. The recommendations for
cultivating discovery may, therefore, be of two categories. First, recommendations regarding the social structure
and dynamics of the scientific community. Second, recommendations regarding the kind of knowledge processed
("nourished").
The above recommedations are directed to the whole community. In parallel, there are recommedations for the
individual scientist. The first recommedation to the scientist would be to integrate within the scientific community.
The individual scientist should not keep his ideas to himself, trying to be original. He should submit his ideas to
public scrutiny not only for giving other scientists the opportunity to criticize them, but also for employing the
ideas in different contexts. This also applies to long-range historical processes which may be triggered off by the
original idea and which result in great discoveries. Another recommedation is to be socially involved in order to
pick up the climate of opinions, or the emerging collective world picture, which determines what kind of ideas
have more chances to be accepted for processing by the community.
From a related view, according to which the collective world picture of the scientific community is mainly a tacitknowledge, one might derive a similar recommedation. It is not advisable to spend all the time sitting in the library
and reading textbooks and even research papers. In a less developed science and in contemporary science
education, the prevailing conception is that the alternative to "indoors" activities such as textbook reading and
attending lectures is to go to the field, make observations and experiments. According to our socially oriented
view, a different advice is offered: Travel as much as possible and observe your colleagues at work and at
meetings. Indeed, the twentieth-century scientist is a travelling scientist, no less, and maybe more, than an
experimenting scientist.
7.5.3 Strategies of Discovery
The two evolutionary patterns discussed in section 7.3.3 have immediate implications for the strategy of generating
new theories. The principle of growth by expansion implies that solutions for problems arising in a given
< previous page
page_220
next page >
page_221
< previous page
page_221
next page >

Page 221
domain may be found by going beyond that domain. Serendipity may carry us unintentionally beyond the present
domain or stratum of knowledge. Hence, rules for cultivating serendipity will have implications for this strategy. A
problem solver who deals with problems in a given domain may discover a solution to a problem in another
domain. However, scientists can try intentionally to look for broader domains in order to solve unresolvable
problems or to explain recalcitrant anomalies. This advice may be implemented by trying to generalize present
concepts and theories and by trying to perform experiments in wider domains. For example, in particle physics
anomalies were resolved and explanations were provided by a chain of generalizations: first, the notions of electric
charge and charge independence of nuclear forces were generalized to the notions of isospin and isospinsymmetry. Then, the notion of hypercharge was invented and isospin was generalized to SU(3) symmetry, where
the isospin symmetry was included in this symmetry. Every such generalization yielded new predictions, which
had to be tested by new experiments and new experimental methods which involved collision experiments at
higher energies. Thus, theoretical expansion proceeded hand in hand with aggressive experimental intervening.
There is no general recipe as to how to implement this strategy of problem solving by expansion and intervention;
it is left for the creative and skillful scientist to find the direction of generalization, to invent fruitful concepts and
to perform new kinds of experiments. In so doing, he may learn from preceding cases and try to imitate them. In
fact, the above chain of generalizations was generated by scientists imitating successful models.
Polya offers the following rules for problem-solving. The first rule is: "Stay as close to the problem as possible."
And he adds: "Yet we cannot predict how close to the problem we shall be able to stay." He offers a principle of
gradual advance: "we first explore the proposed problem itself; if this is not enough, we explore the immediate
neighborhood of the problem. If even this is not enough, we explore a wider neighborhood; whenever our
exploration fails to discover a path to the solution, we are obliged to go further. ... Yet be prepared to go as far
away from the problem as circumstances may oblige you to go" (Polya 1965, 9192). These rules clearly express an
expansionist strategy of problem-solving, although a step-by-step expansion.
The methodological implications of the principle of coevolution of the sensorimotor and cognitive capacities are
not always clearly recognized or practised by scientists. When a scientist proposes a radically new theory
prematurely, without having sufficient experimental data, he is violating this principle. For instance, we can view
Prout's model (nineteenth century), which described all atoms as composed of hydrogen atoms, and the 1932 model
that described them as composed of nucleons and electrons as two theories based on almost the same fundamental
idea: that every atom is composed of basic units which have the characteristics of the hydrogen atom.
< previous page
page_221
next page >
page_222
< previous page
page_222
next page >

Page 222
However, Prout's theory was premature. The nucleon theory could benefit from the data accumulated in chemistry
and physics since Prout's time. This required the evolution of experimental techniques, such as those employed by
Rutherford, Geiger and Marsden for unveiling the structure of the atom. And vice versa: when a scientist preforms
novel experiments without having a sufficiently elaborated theory for guiding him, he is violating the principle. In
the first case, we would say that the theory is too speculative, or premature, and in the second casethat the
experiments are going ahead of theory. In general, the principle implies that the rate of theoretical growth will not
exceed the rate of experimental growth and vice versa.
< previous page
page_222
next page >
page_223
< previous page
page_223
next page >

Page 223
Chapter 8
Tinkering And Opportunism: The Logic Of Creation
Heraclitus says that man's conjectures are like children's toys.
Iamblichus, De Anima (Wheelwright, 88)
8.1 Evolutionary Tinkering in Science
In this chapter, I would like to introduce the notion of tinkering, which will shed light on the significance of all
kinds of unintentional, serendipitous and opportunistic processes of scientific creation. Levi-Strauss (1962)
introduced this notion in describing savage thought and Francois Jacob (1977) borrowed it for characterizing
evolutionary progress. The basic idea can be described in the following way.
In addition to the model of blind-variation-and-selective-retention, which underlies evolutionary changes,
evolutionary progress is characterized by the following historical pattern. The evolution of organs along an
evolutionary line is not a preplanned process. Rather than reflecting a program or a purpose, the structure of an
organism reflects the past, namely the history of the evolutionary line. If the evolutionary line starts with a certain
structure, imperfectly adapted to a certain niche, all future adaptations to changing environmental conditions will be
"ad hoc" modifications imposed on the initial structure. Note that these "ad hoc" modifications are not generated as
a response to environmental pressures. According to the natural selection model, they are generated as blind
variations. Environmental pressures select them opportunistically from the repertoire of variations thus formed. In
this way, organs undergo changes which make them capable of performing new functions.
< previous page
page_223
next page >
page_224
< previous page
page_224
next page >

Page 224
Jacob offers several examples of tinkering. For instance, he describes Ernst Mayr's hypothesis (1964) about the
formation of the lung of terrestrial vertebrates:
Its development started in certain fresh water fishes living in stagnant pools with insufficient oxygen. They
adopted the habit of swallowing air and absorbing oxygen through the walls of the esophagus. Under these
conditions, enlargement of the surface area of the esophagus provided a selective advantage. Diverticula of
the esophagus appeared and, under continuous selective pressure, enlarged into lungs. Further evolution of
the lung was merely an elaboration of this themeenlarging the surface for oxygen uptake and
vascularization. To make a lung with a piece of esophagus sounds very much like tinkering. (Jacob, 1164)
A second example offered by Jacob is the evolution of the human brain. The latter was formed by imposing new
structures on old ones. The neocortex, which controls intellectual and cognitive activity, was added to the
rhinencephalon of lower mammals, which controls emotional and visceral activities. "This evolutionary
procedurethe formation of a dominating neocortex coupled with the persistence of a nervous and hormonal system
partially, but not totally under the rule of the neocortexstrongly resembles the tinkerer's procedure. It is somewhat
like adding a jet engine to an old horse cart" (ibid., 1166).
Other examples may be added: according to some evolutionary theories, jaws were developed in fish from gill
arches, and legs developed from fin supportsto serve perhaps for locomotion in very shallow water. Homo sapiens'
hands, which paved the way to human culture, were probably developed from the forelimbs of a predecessor
species which served mainly for climbing trees. It should be stressed that not every structure is amenable to every
ad hoc modification, and if the latter is successful, it is due to the potentialities inherent in the original structure.
The above pattern has its counterparts in the evolution of science, when scientists exploit existing tools for new
tasks. This might happen to an individual scientist, or to a local group of scientists, when they employ a tool they
happen to have, or to a whole scientific community, when they use an element of their tradition for solving a new
problem. This phenomenon exhibits "opportunism" or tinkering.
The notion of evolution as tinkering is expressed by Jacob as follows: "Natural selection does not work as the
engineer works. It works like a tinkerera tinkerer who does not know exactly what he is going to produce but uses
whatever he finds around him...uses everything at his disposal..." (ibid.). Each of the materials and tools he finds
can be used in a number of dif-
< previous page
page_224
next page >
page_225
< previous page
page_225
next page >

Page 225
ferent ways. The use he makes of the materials and tools around him depends on opportunities. ''Evolution behaves
like a tinkerer who, during eons upon eons, would slowly modify his work...cutting here, lengthening there, seizing
the opportunities to adapt it progressively to its new use. ... It does not produce novelties from scratch. It works on
what already exists" (ibid.).
According to this view, weour organs and brainare the products of tinkering. I suggest that the theories of
molecular biology and particle physics, for example, are also the products of tinkering. This would be a plausible
conclusion of a theory which views science as the extension of our sensorimotor organs and cognitive apparatus.
Since we expound here an evolutionary theory of science, and since we have already seen some evidence in
support of the serendipitous and opportunistic facets of scientific discovery, it would be natural to conjecture that
the above conception applies to science. The main point implied by the notion of tinkering is that evolution "works
on what already exists." This point was emphasized in Chapters 5 and 7, where it was related to the gradualist view
of scientific progress.
Knorr-Cetina (1981, 34) illustrates the notion of tinkering as it is reflected in laboratory practice, when the scientist
uses the local material resources at his disposal and exploits situational contingencies in an opportunistic manner.
However, in science, opportunism and tinkering are not restricted to the manipulation of tangible resources.
Opportunism is exhibited, for example, in the phenomenon of intellectual migration; the scientist "carries" with
him his tangible and intangible tools of research when he moves from one field to another, seizing opportunities
for using the tools in different contexts. A scientist acts as a tinkerer when he makes a serendipitous discovery; he
exploits an opportunity which occurred to him for solving a problem. A scientist may exploit situational
contingencies, such as an idea which occurred to him, as well as a piece of equipment which happens to be in his
possession.
In particular, tinkering is characteristic of the situation when there is no widely accepted theory in a field and
scientists proceed by employing theoretical tools and models they find "around them," which have been used for
other purposes. In other cases, when a general theory is available, scientists try to explain unexpected phenomena,
by making ad hoc modifications, "cutting here, lengthening there." In fact, these are the situations which prevail in
most fields of science most of the time; either a general theory has not been discovered, or there is a general theory
which has been successfully applied only for solving problems in a certain domain, and in order to extend its
domain of application, it has to be modified. The difference between this kind of ad hoc modification and the kind
treated unfavorably is that in the latter kind scientists devise the modification in response to the problem to be
solved, whereas as tinkerers they employ existing elements they find ready for use.
For the sake of illustration, I will give two examples; one is from Greek science and one from modern times. The
first example of tinkering can be
< previous page
page_225
next page >
page_226
< previous page
page_226
next page >

Page 226
found in the Timaeus. Plato (429348 B.C.) set out to construct a theory of matter. He worked in the Pythagorean
tradition which determined the range of (intangible) materials and tools he found "around him." He incorporated
into his theory elements from the theories of his predecessor, Empedocles (490-430 B.C.), and his contemporary,
Democritus (460370 B.C.). He wanted to erect Empedocles' theory of four elements on geometrical foundations. At
his disposal was a recently discovered mathematical theorem. Theaetetus (415?369 B.C.) found that there are only
five regular convex polyhedra: the tetrahedron, the cube, the octahedron, dodecahedron and icosahedron. Plato
seized upon this opportunity and tried to match the four elements and the five solids. He devised intricate
arguments for choosing the tetrahedron as the atom of fire, the cube as the atom of earth, the octahedron as that of
air and the icosahedron as that of water. No earthly element remained to match the dodecahedron, which consists
of twelve pentagons. Plato matched this polyhedron with the boundary of the universe. Thus, for solving his
problem, Plato used a mathematical theorem that happened to be discovered at his time. He opportunistically
exploited an existing tool for constructing his theory. This reminds us of cases where the modern physicist employs
a recently developed mathematical tool for solving his problems.
Toulmin and Goodfield (1962) describe Plato's theory, and they add the following remark in brackets: "The close
approximation of the dodecahedron to a sphere was well known to the Greeks, who made their footballs from
pentagons of leather sewn together in sets of twelve" (ibid, 86). This is presumably an insinuation that in devising
his theory, Plato may have drawn upon this piece of information. As a support for this speculation, we may note
that among the five solids, it is the icosahedron, rather than the dodecahedron, which has the largest number of
facesi.e. twenty triangles. So that at first thought, one might think that the icosahedron is the closest approximation
to a sphere. If this plausible conjecture is correct, it might further confirm the view that in devising his system of
geometrical atomism, Plato acted like a tinkerer. The intricate system of argumentations by which Plato justifies his
choice gives the impression that he arrived at his discovery only by pure reason. Thus, according to the above
interpretation, we can distinguish here between the context of justification and the context of generation, where the
latter was dominated by tinkering.
The second example is the chain of events which led to Rutherford's discovery of the emptiness of the atom.
Rutherford discovered alpha particles in 1899. He and his colleagues had been conducting experiments with these
particles for several years. Rutherford later recalls: "One day Geiger came to me and said, 'Don't you think that
young Marsden, whom I am training in radioactive methods, ought to begin a small research?' Now I had thought
that too, so I said, 'Why not let him see if any alpha particles can be scattered through a large angle?'" (ibid.). The
rest of the story was told in section 2.2.
< previous page
page_226
next page >
page_227
< previous page
page_227
next page >

Page 227
Geiger tried to exploit his trainee's expertise for "a small research." He had the tool and he looked for a problem.
Rutherford suggested one. Rutherford did not believe the exercise would yield anything of great interest. The
experiment which led to the greatest revolution in humanity's conception of the structure of matter was a result of
sheer tinkering. Here it was a skill, or a human resource, that was exploited.
In general, when a scientist uses an existing model for solving a new problem, it looks very natural. However, this
can be viewed as the work of a tinkerer. In particular, solving problems by making analogies can be viewed as
tinkering. This is one of the major ways of generating novelty in science. When the scientist encounters a
phenomenon which requires an explanation, he tries to explain the phenomenon by making an analogy with a
familiar one which is explained by a familiar model. He does not look for the best explanation possible or the
optimal solution to his problem. Rather he employs familiar models. Thus, he adopts the strategy of the tinkerer. If
the analogy is successful and applies to a wide range of phenomena, it may develop into a full-fledged theory. In
this case, the way to novelty is paved by tinkering.
The scientist does not try to find or devise the best conceivable solution to his problem. He picks up whatever
model he can find around him, a model he is already familiar with. This is, for instance, what Leverrier did when
he proposed his conjecture about the existence of the planet Vulcan. Leverrier had suggested this model for
explaining the anomalies in the motions of Uranus and it had worked successfully. In the case of Mercury,
however, it did not work. In fact, in normal science, every scientist who employs a model drawn from the restricted
repertoire of models of the tradition acts as a shortsighted tinkerer who employs Polya's rule that was mentioned in
the last chapter: "Stay as close to the problem as possible" and "be prepared to go as far away from the problem as
circumstances may oblige you to go." This recommendation is diametrically opposed to Popper's policy. The latter
would require that scientists propose bold conjectures that go as far as possible from what is already known.
8.2 Tool-Oriented Scientists: Intellectual Migration
Tinkering provides the raw material of variations for the selection process. If all scientists find around them a
number of similar models that belong to the tradition, the number of variations will be restricted. One way of
increasing the range of variations is to encourage the flow of ideas from other fields. In the last chapter, I
mentioned the phenomenon of intellectual migration as a
< previous page
page_227
next page >
page_228
< previous page
page_228
next page >

Page 228
novelty-generating mechanism. Scientists move to a new area of research and apply the tools or techniques they
employed in their original field for solving the problems they encounter in the new area. We may say that these
scientists are "tool-oriented," rather than "problem-oriented." They are not primarily interested in a specific
problem, rather they look for new opportunities for using their tools. Their tools are "solutions in search of
problems." This opportunistic behavior is characteristic of the above pattern of biological evolution. Genes which
were selected for their function in overcoming certain environmental pressures, are opportunistically exploited in
overcoming new pressures, possibly in a new niche. Thus, these tools which were developed for solving a given set
of problems A are exploited for solving a new set of problems B. The process whereby these tools were generated
was, therefore, ''blind" to the problems they eventually solved.
Drawing upon Dubos' book (1951), Mulkay relates the example of Louis Pasteur (Mulkay 1972, 917) "who moved
continually from one area of research to another, seeking significant problems and attempting to resolve them by
the introduction of ideas and techniques developed during his earlier work." No one would blame Pasteur for being
an opportunist, in the negative sense of the word. Rather opportunism was his method of research, and it proved to
be quite successful. He started his work in a well-established tradition of crystallographic research. First, he solved
the problem of why tartrate solution rotates the plane of polarized light, while the paratartrate is optically inactive.
Both are chemically identical, but, as he found out, the optically active tartrate has an assymmetrical crystalline
structure, whereas the paratartrate is a mixture of right-handed and left-handed crystals. After Pasteur acquired
recognition for his research on tartaric acid and molecular and crystalline dissymmetry, he began to pursue his
hypothesis that life is a function of molecular dissymmetry. This work evoked strong objections in the scientific
community. Then he moved on to the problem of fermentation. Since the processes of fermentation gave evidence
of molecular dissymmetry, he conjectured, against the accepted view in the field, that they are related to the
activity of living organisms. Then he turned to spontaneous generation and attempted to prove that it does not
exist.
Another example which is offered by Mulkay is the growth of the phage network within molecular biology (ibid.,
3435). During the 1930s, a group of physicists, including Delbruck and Szilard, "concluded that research in
physics was unlikely to reveal any interesting problem for some time to come; that biology seemed to offer the
most promising opportunities for using the physical methods with which they were familiar." They undertook
research into bacteriophages. Since they used the distinct methods they brought with them from physics, they
formed a distinct group which developed later into the new field of molecular biology. Following the discovery of
the structure of DNA by Watson and Crick, the field grew explosively. There was a continuous flow of physicists,
< previous page
page_228
next page >
page_229
< previous page
page_229
next page >

Page 229
chemists, biophysicists, physical chemists, bacteriologists, medical researchers, virologists, biologists and
biochemists into the field. In fact, the high turnover of intellectual migrants became a characteristic feature of the
field.
There are plenty of cases of intellectual migration in contemporary science, although not necessarily in the sense of
moving to a different scientific discipline. Nowadays one does not have to move to a different science in order to
qualify as an intellectual migrant. For example, many physicists switch areas or work simultaneously in two or
more areas within physics. In contemporary physics, since the distance between two subspecialities may be quite
large, we can treat a scientist who moves from one area of specialization to another as an intellectual migrant. By
the same token, a scientist who works in two areas may be treated as a person who has dual intellectual
"nationality." Furthermore, in order to exploit an existing tool for solving new kinds of problems, it is not
necessary to move to a new area. One can stay in his native area of research and borrow tools created in a different
area. Or, one can apply an existing tool for solving a new kind of problem in the same area. These phenomena
exhibit serendipity, opportunism and tinkering.
8.3 Tinkering in Particle Physics
One might not be surprised if one is told that biology, or even chemistry, has been dominated by tinkering. What I
would like to claim is that important developments in contemporary physics have been characterized by tinkering.
More specifically, tinkering, as a "method" of research, has been most distinctly exhibited in (elementary) particle
physics. So I will dwell on it at some length. This field has been in the forefront of physics since the late 1940s.
Indeed, tinkering is characteristic of explorations into unknown territories of nature.
The main problem in elementary particle physics was that no general theory was available in the field. Until the
early 1970s, quantum field theory gradually declined and no general theory was widely accepted. Physicists
considered this to be a shortcoming. Their desideratum was a model of science where one general theory reigns,
although already pre-World-War-Two physics had deviated significantly from this ideal. But gradually they
adopted the "method" of the tinkerer. They used whatever tools were available to them for solving the problems in
the field. Or "worse," they generated the problems that would be suitable for their tools. They took elements from
different theories and models and "tailored" their solutions. The renormalization program, Regge poles theory,
current algebra and the quark model are but a few examples of this phenomenon. Yet, if science is an evolutionary
process as described by Jacob, there is no reason to think that the ideal of having a unifying theory will be ever
attained. At most, from time to time there might appear a theory which will dominate a given field. But sooner or
later physicists will find out that the scope
< previous page
page_229
next page >
page_230
< previous page
page_230
next page >

Page 230
of the "general" or the "unified" theory is restricted to a more or less narrow range of phenomena, and in order to
adjust it to a wider range, they will have to resort to tinkering again. This pattern of development has been
characteristic of physics since the advent of quantum mechanics. But in particle physics it was pushed to its
extreme.
The opportunistic exploitation of existing tools for new tasks, as it has appeared in particle physics, is described by
Andrew Pickering under the title "opportunism in context" (Pickering 1984). However, Pickering treats this
phenomenon on a "shallow" level, without offering a deeper explanation for it. Equipped with the evolutionary
model of tinkering, we can exploit this opportunity and offer an explanation for this phenomenon.
According to the picture offered by Pickering, each scientist, in the course of his scientific experience, has acquired
tangible and intangible techniques and methods and he seizes upon opportunities to use the resources at his
disposal in different contexts. If an individual scientist successfully exploits an existing tool for solving a new
problem, this can be viewed as personally motivated opportunism since it would lend him recognition. But when
the tool belongs to the shared resources of the scientific community, the community might adopt it for solving the
problem. In this case it may be viewed as communal opportunism. In some cases the tool may not be familiar to
most members of the community. And yet, if it is successfully exploited for solving current problems, it might be
adopted by the community, although some time will elapse before the new tool is assimilated into the shared
resources.
Employing traditional terminology, we would say that particle physics has been developed through a process of
making analogies and constructing models. From our present viewpoint, we would say that, in a typical case, when
particle physicists have faced a new problem they have tried to employ the theoretical or experimental expertise
they had at their disposal for solving new problems. In describing hadrons as quark composites, they used the
model which proved successful in atomic and nuclear physics, describing the atom as composed of electrons and
the nucleus, and the nucleus as composed of nucleons. Quantum electrodynamics (QED), that treated
electromagnetic interactions with great success, served as a major resource for constructing theories and models for
weak and strong interactions. It inspired the development of relativistic quantum field theory and later, of current
algebra and gauge theories of quark and lepton interactions. The composite model and QED were among the major
items included in the standard tool-kit of the particle physicist. Let us turn to some major examples.
8.3.1 Symmetries without Dynamics
In order to illustrate the mechanism of the analogical extension of existing theoretical tools into new areas, let us
consider first the era of internal symme-
< previous page
page_230
next page >
page_231
< previous page
page_231
next page >

Page 231
tries in the history of elementary particle physics, when particle physicists extensively employed group theoretical
tools. Some of these tools were inherited from pre-World-War-Two atomic, nuclear and relativistic physics, and
some were borrowed directly from mathematics. These tools were applied to symmetries and conservation laws
which, in the absence of a satisfactory dynamical theory, were supposed to bring order into the plethora of particles
and their interactions.
The application of symmetry groups to elementary particle physics began with the introduction of the notion of
"isotopic spin" (later called "isospin"). The first step in the invention of isospin was made by Werner Heisenberg in
1932, immediately after the discovery of the neutron, when he introduced the notion of nucleon. Heisenberg
suggested to look at the proton and the neutron as two states of the same "particle," the nucleon. He employed here
the quantum-mechanical machinery of spin, introduced by Wolfgang Pauli in 1927 for describing the electron spin.
The electron had been described by a wave function with two components, corresponding to the two spin states of
the electron''spin up" and "spin down." Spin was an angular momentum magnitude, describing spatial properties of
the electron. The spin formalism had been invented to explain the fine structure of atomic spectra. It was not
supposed to distinguish between charge states.
Thus, in order to adapt to the new situation which came about with the discovery of the neutron, Heisenberg
exploited the spin formalism. We might say that he made here an analogy with spin. But the notion of analogy will
not capture the significance of the innovation. Everything can be made analogous with everything. Which analogy
is acceptable and which is not is determined by the standards prevailing in the scientific community. According to
the prevailing conceptions, the two phenomenathe existence of two spin states of a particle, and the existence of
two particles which are close in their mass and differ in their electric chargewere far from being similar. By
exploiting the spin concept for describing the proton and the neutron as two states of one "particle," Heisenberg set
up a new standard of similarity, which paved the way to the notion of "internal" symmetry, eventually leading to
the SU(3) symmetry and the quark model.
The spin operators generated an SU(2) group of rotations in ordinary space. Isospin did not correspond to rotations
in ordinary space. Eugen Wigner later described the isospin operators as the generators of a new SU(2) rotation
group in an abstract isospin space, in analogy with the SU(2) group of ordinary spin. Thus, the notion of isospin
brought with it the resources of group-theory into elementary particle physics. Hadrons were classified into isomultiplets, in analogy with the splitting of the energy levels of a spin multiplet. These corresponded to the
representations of the SU(2) group of isospin. This led to a conservation law for isospin, in analogy with the law of
coservation of angular momentum. One of the variations in the analogical
< previous page
page_231
next page >
page_232
< previous page
page_232
next page >

Page 232
transfer from spin to isospin was that isospin conservation was violated by electromagnetic and weak interactions,
whereas angular momentum was universally conserved. The symmetry-breaking due to these violations of the
symmetry was responsible for the relatively small mass difference between the members of an iso-multiplet, i.e.
between the pions (forming an iso-triplet) or between the nucleons (forming an iso-doublet). Thus, the above
variation was represented by the notion of symmetry-breaking. Symmetry-breaking within an iso-multiplet was
small in comparison to the strong interaction and this was expressed by the small mass differences between the
members of an iso-multiplet.
The particle physicist in the fifties was equipped with the conservation laws of energy-momentum, angular
momentum, electric charge, baryon number, lepton number and parity, which were regarded as having universal
scope. The appearance of isospin introduced the notion of conservation law which is not universally obeyed. This
has been the evolutionary pattern to follow. The price of extending the application of an existing tool to a new
area, or to solving a new kind of problem, was the modification of the tool. As agents capable of acting
intentionally, physicists devised some of the modifications in response to the problems which they faced. This kind
of modifications was, therefore, treated unfavorably. The general feeling among particle physicists was that these
were ad hoc modifications, in the "bad" sense.
Strangeness was the next quantum number which was introduced into the elementary particle tradition by GellMann (1953) and Nishijima (1954), for solving the problem of why some strongly interacting particles were
decaying slowly as in weak decay processes. The conservation of strangeness was modeled on the conservation of
charge, but with a slight variation. It was not universally conserved; it was not conserved by weak interactions.
After the isospin symmetry was established, it served as a model for constructing theories of "internal" symmetries.
In the next step the SU(2) group was extended to include strangeness. And finally the SU(3) symmetry, or the
unitary symmetry, proposed by Gell-Mann (1961) and Ne'eman (1961), was selected. Since the unitary symmetry
was modeled on isospin, it was sometimes called "unitary spin." As Pickering rightly notes, both Gell-Mann and
Ne'eman aimed at constructing a gauge theory as a particular version of quantum field theory of strong
interactions, as well as providing a classification scheme of hadrons. But in the following years, quantum field
theory quickly went out of fashion. So that "very soon after its invention, SU(3) was divorced from its roots in
gauge theory and left to stand or fall on its merits as a classification system" (ibid., 57). This was, therefore,
another example of a partially serendipitous event. The main point was that in the construction of the theory, the
tools of both quantum field theory and the armory of symmetry groups were employed, without knowing in
advance what would be the nature of the final product.
< previous page
page_232
next page >
page_233
< previous page
page_233
next page >

Page 233
Again, a further ad hoc variation was imposed on the model. The relative mass-differences between the members
of an SU(3) supermultiplet were much bigger than in the case of an iso-multiplet. The large broken symmetry was
accounted for by the Gell-Mann-Okubo mass formula (Gell-Mann 1961, Okubo 1962). This formula enabled the
prediction of the mass of the missing member of the unitary decuplet, the W-, which was detected in 1964 at the
Alternating Gradient Synchrotron at Brookhaven (Barnes et al. 1964), having the predicted mass.
The next step was SU(6). Gursey, Pais and Radicati (1964) constructed this group by analogy with the approximate
SU(4) symmetry of the nucleus proposed by Wigner in 1936 (Wigner 1937). The latter corresponded to the
combination of two SU(2) symmetries: isospin independence of nuclear forces and spin-orientation independence
of the nucleons within nuclei. The SU(2) of isospin was extended to the SU(3) of unitary symmetry. The result was
an approximate SU(6) independence of strong interactions. The variation here arose from the problems related to
the fact that the symmetry was good only for non-relativistic cases. The attempts to extend the group so that it
would accommodate both the Lorentz group (which represented the relativistic effects) and SU(6) were not
successful. But, in the meantime, these efforts yielded a variety of groups, which were not symmetry groups, and
which included the Lorentz group, having infinite representations.
The termination of this evolutionary line came as a result of the appearance of the quarks (as theoretical entities).
But the origin of quarks was in this evolutionary line: in the SU(3) symmetry. Eventually, the unitary symmetry
was abandoned and the quarks have remained. This is a perfect example of a cooperative-historical creative
process. It started with Heisenberg's idea of isospin and ended up with the discovery of quarks. When Heisenberg
proposed his idea for explaining the difference between the proton and the neutron, he did not envisage that the
idea would eventually lead to the discovery of quarks. Many minds contributed their ideas and expertise to this
enterprise. So, who discovered the quarks? If we use the notion of discovery in a narrow sense, the answer would
be: Gell-Mann and Zweig. But if we take into account the whole process, Heisenberg made a decisive step in the
discovery process. (When I say that quarks were discovered, I do not mean, of course, that quarks have been
observed. Rather I mean that a successful theory in which quarks appear as theoretical entities was discovered.)
8.3.2 The Resources of Quantum Field Theory
The use made of quantum field theory by particle physicists in the 1950s and the 1960s is an architypical example
of tinkering. Particle physicists inherited QED from the 1920s and the 1930s, when it was successfully applied to
many phenomena in atomic physics. However, the theory suffered from a severe dif-
< previous page
page_233
next page >
page_234
< previous page
page_234
next page >

Page 234
ficulty. It was impossible to calculate certain integrals corresponding to experimentally measurable quantities. The
second order perturbative corrections of these quantities yielded infinite integrals. The problem was solved by a
typical tinkerer's stratagem which turned out to be successful. This was the "renormalization" procedure of
absorbing the infinities into physical constants. The trick gave the correct result in calculating the Lamb Shift, with
a very high degree of accuracy. The physicists did not have full understanding of the physical meaning behind this
procedure. But the theorists were satisfied, since this technique enabled them to calculate measurable quantities.
Following the success of the renormalization technique developed by Tomonaga, Schwinger and Feynman in the
late 1940s, QED became a very powerful tool for calculating measurable quantities of electromagnetic interactions.
Via this process, the meaning of the term theory was shifted. Traditionally, one of the main functions of a physical
theory had been to provide a description or an explanation. With this shift of meaning, a theory was concieved
mainly as providing a tool for calculations. Elementary particle physicists in the early Forties, therefore, tried to
exploit QED for calculating the quantitative properties of weak and strong interactions. I will not dwell here on the
application of quantum field theory to weak interactions. I will mention only that already in 1934 Enrico Fermi
constructed a quantum field theory for the weak interactions in beta-decay. He modeled his theory on QED, but
one of the difficulties was that the theory was non-renormalizable.
The first application of quantum field theory to strong interactions was proposed by Yukawa in 1935. Again, it was
modeled upon QED. But it was a non-starter. Since the coupling constant of strong interactions is greater than one,
the higher order terms in the perturbation expansion are increasingly greater than the first-order term. This led to
insurmountable difficulties in the calculations when the renormalization procedure was applied. Having no other
alternative, physicists turned then to the S-matrix theory which was available since the mid-1940s. Being a "blackbox" theory, it treated only the transition probabilities of the input/output states. The S-matrix was treated as an
analytical function and the tools of the mathematical theory of complex-variables functions were exploited
extensively. Nevertheless, quantum field theory was still used as a guide in investigating the analytic properties of
the S-matrix. Meanwhile, those who developed the "bootstraps" approach to the S-matrix abandoned field theory
altogether.
A widespread tool, which was picked up from a work on non-relativistic potential scattering in the late 1950s and
applied to the S-matrix tradition, was Regge poles theory. Regge introduced the concept of complex angular
momentum, purely as a mathematical technique for solving his nonrelativistic problem. "Regge himself never
backed the cavalier applications of his results (which had been proved only in norelativistic Schroedinger theory)
to the relativistic problem" (Cushing, 131). Regge poles theory was a central research topic that
< previous page
page_234
next page >
page_235
< previous page
page_235
next page >

Page 235
engaged many particle physicists in the 1960s and the early 1970s. Since then, it has been totally abandoned. 10
8.3.3 Playing with Quarks
The idea that strongly interacting particles are composed of elementary constituents goes back to 1949. According
to the Fermi-Yang model (1949), the pion was not an elementary particle; it was a nucleon-antinucleon composite.
The Sakata model, which extended this model to include strange particles, followed in 1956 (Sakata 1956). The
fundamental particles were the three Sakatons, i.e. the two nucleons and the "strange" baryonlambda. The rest of
the baryons and all the mesons, including the strange ones, were Sakaton composites. With the advent of SU(3)
symmetry, the Sakatons were replaced by the three quarks. The quarks corresponded to the fundamental grouptheoretical representation of SU(3), the triplet, while the baryons and mesons occupied octets representations. An
earlier version of the model was suggested as a mere mathematical possibility in 1962 by Goldberg and Ne'eman
(1963) and a more elaborate version was independently developed in 1964 by Gell-Mann (1964) and Zweig (1964),
who discussed the physical application of the model. The main difference between the Sakata model and the quark
model was that in the former the nucleons and the lambda were not composite particles, whereas in the latter all
observed hadrons were composed of quarks. At this point quarks have gradually become independent of their
group-theoretical origin. Thus, the quark model followed in the footsteps of the S-matrix theory and current
algebra which became semi-independent of their field theoretical origin.
The main difficulty in this model was that a free quark had never been detected. Nevertheless, this non-realistic
model yielded very good results and the quark-game continued. It was an effective tool for classifying hadrons and
for calculating their dynamic properties. Particle physicists employed the composite system model because it was
familiar to them from nuclear physics and it served as a good tool for calculations.
8.3.4 Tool-Oriented Particle Physicists
In particle physics we very frequently encounter the phenomenon of tool-oriented scientists. I will describe several
cases taken from Pickering (1984) which can be viewed according to the tinkerer's principle.
Dalitz
Richard Dalitz was trained in nuclear physics and specialized in the analysis of composite systems. In the late
1950s he moved into elementary particle physics and dealt with strange hadronic resonances. He borrowed the tool
of spectroscopic analysis from atomic and nuclear physics and employed it in dealing with quark systems. The
observed hadrons played the role of the energy levels of composite quark systems. Many theoreticians followed
him
< previous page
page_235
next page >
page_236
< previous page
page_236
next page >

Page 236
and since the mid-1960s the quark model was developed with the tools of nuclear physics (ibid., 9697).
Ting
Experimental high-energy physicists are prone to be tool-oriented since their equipment is so expensive and since
they invest many efforts to gain expertise in using it. Samuel Ting had a lot of experience in the detection of
electron-positron pairs following a series of experiments he had conducted since 1965 at DESY in Hamburg. At
first he was mainly engaged in the production of e+e- pairs for testing QED. Then he turned to investigating the
properties of neutral vector meson by detecting e+e- pairs. In the early 1970s he was looking for a new subject
where he could use his expertise. He decided to look for vector mesons at higher energies. In 1974 his group
conducted an experiment at the Alternating Gradient Synchrotron in Brookhaven, bombarding a beryllium target by
a proton beam. They scanned the energy region from 2.4 to 4 Gev, counting e+e- events. Around 3.1 Gev they saw
a strong and exceptionally narrow peak. They interpreted the results as the decay products of a new vector meson.
This was the discovery of the J-psi particle (discovered simultaneously by Burton Richter's groupin a colliding
beams experiment, where hadrons were produced from electron-positron collisions, at Stanford)a turning point in
the history of particle physics.
Yang-Mills Theory
The history of Yang-Mills gauge theory provides us with instructive examples of the phenomena of tool borrowing
and tinkering. It also demonstrates the cooperative nature of theory-construction.
The gauge field theory which was first proposed by Yang and Mills (1954), was modeled on QED. The two
physicists intended to extend QED beyond electromagnetic interactions. C. N. Yang started the process by
employing his knowledge in group theory, which he had acquired in his undergraduate studies. His BSc thesis was
entitled: "Group Theory and Molecular Spectra" (Pickering, 161). Thus he was one of the first physicists who
introduced the group-theoretical tradition in particle physics. He applied the QED formalism to strong interactions.
He replaced the electron field in the QED Lagrangian by the nucleon field, for example, which belonged to an
isospin doublet. The local gauge transformations which kept the QED Lagrangian invariant, generated the group
U(1). With the nucleon field, he aimed at constructing a Lagrangian which would be invariant under the isospin
symmetry group SU(2). However, in order to keep the Lagrangian invariant under the SU(2) group, he had to
introduce, instead of the photon, an isospin triplet (W+, W0, W-) with spin 1, as the quanta of the field. The
difference between these particles and the photon was that they carried an electric charge and this led to the
appearance of self interaction terms in the Lagrangian, in addition to the terms describing the propagation of free
Ws in space.
Since 1954, when the theory was proposed, many authors made contributions to it. Among them were Sakurai,
Gell-Mann, Schwinger, Glashow and
< previous page
page_236
next page >
page_237
< previous page
page_237
next page >

Page 237
Salam. They were mainly engaged with the masses of the vector particles. Since the photon is massless, the
original QED Lagrangian contained no mass terms for the photon. But the Yang-Mills theory was intended to treat
strong and weak interactions which are short-range. In particular, the possibility of identifying the Ws as the
mediators of weak interactions was attractive, since this may have led to the unification of the electromagnetic and
the weak interactions. For example, the W+ and W- and the photon may belong to the same family. But the
electromagnetic force is much stronger than the weak force and the photon is massless, whereas the mediators of
weak interactions are massive. So the idea to put them in one family raised difficulties.
In 1961 Glashow suggested, therefore, that the gauge symmetry would be SU(2)xU(1) so that the photon may be
singled out. Salam and Ward made a similar suggestion. In these theories, the masses of the intermediate vector
bosons were not derived from the theory but were inserted in the Lagrangian in an ad hoc manner. But this caused
the theory to be non-renormalizable. A solution to this problem was found by the model of spontaneous symmetry
breaking, where the intermediate vector bosons acquired mass, whereas no mass terms for the Ws appeared in the
Lagrangian. This model had been introduced by Nambu and Jona-Lasinio in 1961, in another context.
Yoichiro Nambu had specialized in the application of quantum field theory to many-body phenomena in both solid
state physics and particle physicstwo relatively disconnected specialities. In the late 1950s he worked on the
phenomenon of superconductivity which was then a subject of central interest. He continued to work also in
particle physics. He applied his expertise in solid state physics to construct "A Dynamical Model of Elementary
Particles Based Upon an Analogy with Superconductivity" (Nambu and Jona-Lasinio 1961). According to this
model the Lagrangian which describes a physical system may be invariant under a symmetry group which is
"spontaneously broken" by the physical states of the theory. Originally, in the early 1960s, the model was
employed to account for the symmetry breaking of SU(3). Since then, the resources of solid state physics have
continued to fertilize particle physics.
Jeffrey Goldstone, who also was an expert in the theory of superconductivity, argued that spontaneous symmetrybreaking implies the existence of massless, spin-zero particlesGoldstone bosons. Salam and Weinberg joined him
in this conclusion in 1962. Yet, as another solid-state physicist, P. A. Anderson, argued, massless particles do not
appear in superconductors. This problem led to further developments of the theory, to the Higgs mechanism and
finally, to the Weinberg-Salam model which did not differ much from Glashow's model.
Yet, one obstacle remained in the way of the construction of a unified electroweak gauge theory: the problem of
renormalizability, which Yang and Mills had failed to solve. The renormalization procedure would bridge the gap
between the theory and the calculation of measurable quantities.
< previous page
page_237
next page >
page_238
< previous page
page_238
next page >

Page 238
Two figures played a decisive role in solving the problem. The graduate student 't Hooft who solved the problem
and Veltmanhis supervisorwho brought the problem in its final form to his student's attention. Now we come to the
most crucial step in the process: Veltman and 't Hooft's contributions to the development of Yang-Mills theory. I
will describe their contributions in some more detail, since they were typical of the whole process of discovering
the unified elecroweak model and since they demonstrate our thesis. I shall follow here mainly Pickering's
description, and all citations will be from there (ibid., 174178).
The story begins as Martin Veltman tried to understand current-algebra, using the techniques developed in his PhD
work on Lagrangian field theory. He looked for a subject which could be treated by the techniques and methods he
had acquired. He attempted to derive current-algebra results from manipulation of quantum fields. This was GellMann's strategy in a reverse order. Gell-Mann constructed current-algebra with the heuristic guidance of gauge
field theory and then discarded the gauge theory and treated the resulting current-algebra independently of its
origin.
In 1966 Veltman found that some of the current-algebra results could be derived from two field-theoretical current
equations. Richard Feynman told Veltman that the equations' structure was characteristic of Yang-Mills theory. But
he argued that it was related to strong interaction rather than to weak interaction. By this he followed the Sakurai,
Gell-Mann and Ne'eman's approach to gauge theory, which was the mainstream approach at that time.
Later, Veltman's collaborator John Bell became interested in Veltman's equations and in 1967 he published a paper
"arguing that what was needed was a Yang-Mills structure for the weak interactions. Veltman later recalls that
Bell's paper on the subject 'became a great mystery for me for a while. It kept on going in my mind. ... If finally
dawned upon me that the current equations were a consequence of a Yang-Mills type structure of the weak
interaction'." Here we have an event which has all the ingredients of an eureka moment.
Veltman then went on "to investigating the renormalizability of gauge theory, drawing further upon the resources
he had acquired in his thesis work." His interest was in massive Yang-Mills theory, where the masses were
inserted "by hand," since the massless gauge invariant theory was unrealistic. The received opinion was that
massive theories were non-renormalizable. He found that the obstacle to the renormalizability of the theory were
certain higher-order calculations in perturbative field theory which involved self-interacting intermediate massive
vector bosons. Technically speaking, these were the calculations which corresponded to Feynman diagrams
containing two loops.
On the other hand, he had learned that massless gauge theories similar to his massive theory had been showed to
be renormalizable. At this stage he began to investigate the possibility that perhaps appropriate scalar (spin-zero)
< previous page
page_238
next page >
page_239
< previous page
page_239
next page >

Page 239
fields could be introduced into the massive Yang-Mills Lagrangian in such a way as to cancel the two-loop
divergences due to the the self-interacting vector bosons. This was the stage where Veltman's student, Gerald 't
Hooft, entered the scene.
In 1969, at the suggestion of Veltman, 't Hooft studied the renormalizability of the "sigma-model." This was "a
simple Lagrangian field theory used in heuristic current-algebra calculations. It was not a gauge theory, but it
displayed spontaneous symmetry breaking. ..." When he was going to start working for his PhD under Veltman's
supervision, he proposed to work on gauge theory. "Veltman agreed, although he felt that it 'was so much out of
line with the rest of the world that very likely one was producing specialists in a subject that nobody was interested
in'" and he suggested that 't Hooft work on the renormalization of massless gauge theory, since some problems still
remained even there.
't Hooft published a paper where he presented the first detailed argument that massless theory was renormalizable.
Then Veltman asked him whether he could devise a renormalizable theory with massive vector bosons. His
immediate answer was positive, since he had the appropriate tool for carrying out the taskthe sigma-model. In a
paper published in 1971 ('t Hooft 1971) "'t Hooft used the technique of spontaneous symmetry breaking, familiar to
him from his earlier work on the sigma-model, to give masses to the vector bosons of the pure gauge theory. By
adding multiplets of scalar particles into the massless Yang-Mills Lagrangians. ..." With this result he found that
gauge theories, in which vector bosons acquired mass by spontaneous symmetry breaking, were renormalizable.
This was the breakthrough which brought the Yang-Mills gauge theory into the focal attention of particle physicists
and which paved the way to electroweak unification.
The last episode demonstrates the role played by professional marginality in the process of innovation. Pickering
quotes Ne'eman:
"Current opinion" rejected quantum field theory regarding it as hopelessly wrong. ... Hence the quantization
of Yang-Mills dynamics was finally achieved by workers who happened to be immune to that consensus'
view. It required geographical remoteness (Fadeev-Poppov, Veltman-'t Hooft [in the USSR and Utrecht,
respectively]), professional remoteness (De Witt, working in gravitation theory). ... (Ne'eman 1983, 2, note
2)
After quoting the above citation, Pickering adds that Veltman was both geographically isolated from the
mainstream of particle physics at Utrecht and "professionally remotehis route to entanglement with gauge theory
and the resources he brought to bear were unique and distinctive" (note 60, 199200). Thus he could contribute
novel variations to the pool of ideas, methods and
< previous page
page_239
next page >
page_240
< previous page
page_240
next page >

Page 240
techniques. This supports the thesis that intellectual marginality is one of the sources of novelty in the scientific
community.
Intellectual marginality as a source of innovation corresponds to the evolutionary phenomenon in which genes with
low frequency are activated by new environmental conditions, or pressures, and spread through the population.
Mayr (1959) describes the following mechanism for the evolution of a new species, which demonstrates the
contribution of marginality to the generation of novelty. Speciation occurs according to this theory as a result of
the separation of a "peripheral isolate" from the original population. A new species may arise from a small
population cut off in a remote corner at the edge of the range occupied by the "parent" species. The conditions in
the isolated niche, which differ from those prevailing at the center of the range, together with the small size of the
isolated population, may bring about the evolution of radically new characteristics on the basis of the existing gene
pool. The new species formed in this manner may return to the area occupied by the parent species, be better
adapted and displace the parent species. The case of 't Hooft demonstrates this phenomenon. The new species in
this case can be identified with the gauge theories and unified field theories which were brought to the forefront of
particle physics following the success of electroweak unification.
<><><><><><><><><><><><>
To summarize, the history of the Yang-Mills gauge theory exhibits the following characteristics.
1.The theory was gradually evolved. New variations accumulated in an evolutionary manner.
2.Most of the novelties were introduced by tool-oriented scientists: the theory's inception was inspired
by the tools of group theory brought from the field of molecular spectra. Then, ideas and methods
drawn from the resources of solid state physics were injected into the research program.
3.Marginal scientists who were "immune to the consensus" employed their tools.
4.The process was cooperative, wherein many minds participated. The decisive step was done on soil
prepared by all participants. The discovery was a product of the contributions of numerous scientists.
5.The breakthrough was achieved by a scientist who set out to solve a very specialized problem, which
turned out to be a crucial problem. 't Hooft was just a "cog in the wheel." He appeared on stage at the
right moment and "put his stone into place," to use Hegel's metaphor. He was fortunate to get the
right question from Veltman. Although he had specialized in
< previous page
page_240
next page >
page_241
< previous page
page_241
next page >

Page 241
using a marginal tool, it turned out to be the right tool for solving the crucial problem.
6.The process was not preplanned. Veltman did not expect the research to result in a breakthrough. He
agreed somewhat reluctantly to his student's proposal, thinking that no one would be interested in that
subject. This also reminds us of Rutherford, who gave young Marsden an ''exercise" for practicing his
skill, while he did not believe that any significant result would be obtained.
According to the tinkerer's principle, scientific products are generated by situational contingencies. Knorr-Cetina
describes the tinkerer's principle as it emerges from "laboratory studies." And the conclusion is that "similar
problems tackled by different people in different environments will yield different solutions" (Chubin and Restivo
1983, 70). The impression is made that "anything goes" in the scientific laboratory. But the point is that tinkering,
in theoretical as well as in experimental science, is a mechanism for generating variations. Eventually, certain
variations, presumably the "fittest," will be selected, or accepted by a given community. And in normal science,
different laboratory sites in the same research area have similar resources. Therefore, similar situational
contingencies are encountered in different laboratories, and the range of variations is restricted. Furthermore, the
"environmental" conditions and the standards of selection are also similar. The same happens in theoretical science.
Scientists in different places have at their disposal more or less the same repertoire of models, methods and
theories, which are characteristic of their research tradition. And the standards or mechanisms of selection are
similar. Therefore tinkering will generate radical novelties only in an open system, where scientists can draw upon
marginal or external resources. Thus, marginality, migration and other sources of cross-fertilization are among the
factors contributing to the generation of radical novelty. Historical contingency might therefore be exhibited at the
time of radical scientific change. The present world picture of physics, for instance, reflects the historical
contingencies which were incorporated by the scientific revolutions of the seventeenth and twentieth century.
Contemporary physics is a very successful science, but the present theories of physics are by no means "perfect."
The same can be said about the living world; no species is perfect, included Homo sapiens. This is a direct
conclusion of the tinkerer's view of science and natural selection.
The progress of science is not attained by erecting monumental systems of thought. Only in rare cases are the
products of science cathedrals planned by an ingenious architect. Rather, scientific creation is in many cases the
work of a tinkerer. Jacob intended this principle to be applied to natural selection. He did not envisage that the
model of natural selection might be applied to
< previous page
page_241
next page >
page_242
< previous page
page_242
next page >

Page 242
science. The conclusion which he perhaps did not intend to reach is that the same principle governs savage thought
and scientific thought. And yet, this very conclusion is a product of tinkering; the principle of tinkering was
borrowed by Jacob from the context of savage thought and was transferred to the context of biological evolution
(this creative transfer was perhaps the result of the following situational contingency: Levi-Strauss wrote French,
making his work more accessible to Jacob, the Frenchman), whereas here it is borrowed from the latter context and
transferred to the context of science.
< previous page
page_242
next page >
page_243
< previous page
page_243
next page >

Page 243
Chapter 9
Completing The Picture: Is There A Role For The Genotype-Phenotype Process?
Two kinds of novelties are exhibited by organic evolution. The first kind is the new variation (mutation or
recombination) which is selected by the environment. Blind or serendipitous discoveries in science correspond to
this kind of novelty. The second is the grown up organism. The more complex is the organism and the more
versatile is the environment, the larger is the novelty which may be embodied in the mature organism. The process
of ontogeny, i.e. the growth of the organism after conception, yields a grown up "product" which has some
physical characteristics which depend on both the environmental conditions under which the process took place
and on the genotype's constitution. The phenotypic variability in our species is much larger than in other species,
since in addition to the physical variability there is also a personal and mental variability. Correspondingly, the
environment includes the sociocultural environment in addition to the physical and organic environment. The
grown-up human being constitutes a novelty, in particular in personality and intellectual capacities. Personal and
intellectual novelty depends much on the sociocultural environment.
Ontogeny happens to be the most concrete manifestation of the life phenomenon. The following question therefore
arises: is there a counterpart of ontogeny in an evolutionary theory of discovery? The mainstream program of
evolutionary epistemology models the evolution of science on phylogeny and leaves ontogeny without an analogue.
I will try, therefore, to exploit this neutral analogy in our evolutionary model, which has remained untouched, and
to convert it into a positive analogy (Kantorovich 1989).
On the other hand, there is a fundamental scientific process that is left in our model without a biological
counterpart. This process turns out to be the commonest mode of scientific research in "normal" science, i.e. the
process of developing a newly generated idea or model. In this process we
< previous page
page_243
next page >
page_244
< previous page
page_244
next page >

Page 244
exploit the potential inherent in an embryonic idea, while being engaged in problem solving. Here logic, in the
wide sense, enters into the game of science. This process is what we have called "research program" or "dynamic
theory." It seems natural, therefore, to consider the possibility of modeling the research program on ontogeny.
Phylogeny emerges from the totality of all ontogenies. Hence, understanding ontogeny is essential for
comprehending and explaining phylogeny. Similarly, the evolution of science is the sum total of scientific research
programs; the scientist, qua-scientist, is only engaged in research programs. Hence it is essential for the
evolutionary epistemologist to study the mechanism by which a research program develops. Organisms survive
environmental pressures, and before they die they grow and change. Living theories, likewise, have a finite
lifetime; before being replaced they develop and undergo modification. We will see that in addition to these
parallelisms, the two processes, the growing organism and the developing theory, have structural similarities. This
would strengthen the conjecture that the counterpart of ontogeny is a research program. Thus, I will propose to
extend the evolutionary model for science such that it will do justice to the life history of an idea or a theory. I will
argue that the latter can be treated analogously to ontogeny, and that in science the "logic" of ontogeny is
complementary to the "logic" of selection.
9.1 Non-Creative Discovery: The Genotype-Phenotype Logic of Growth
As we have seen, there is an epistemological significance not only to the death of a theory, i.e. to its refutation or
replacement, but also to its growth and life history. The discovery of the idea or model which initiates the research
program is the most creative phase of the process. It is a product of blind variation exhibited in our approach by
serendipity, tinkering and involuntary processes. The continuation of the process whereby the initial idea "ends
with algebra" is no less important; it consolidates the product of the creative process, exhausts its potentialities and
makes it susceptible to evaluation. It is carried out as a process of exposure, guided by some kind of logic.
The analogy between the growth of a dynamic theory and the growth of an organism can be demonstrated as
follows. First, both a research program, and an organism are engaged in problem-solving. A research program
solves, for instance, problems arising from the need for adjusting its basic ideas (which constitute the "hard core"
in the Lakatosian terminology) to new observational data. Similarly, an organism adapts to new environmental
conditions on the basis of its genetic makeup. The metaphor of an organism as a problem-solver is employed, for
example, by Karl Popper.
< previous page
page_244
next page >
page_245
< previous page
page_245
next page >

Page 245
Secondly, there is a structural similarity between the growth processes in both cases. The growth of a dynamic
theory in a research program resembles ontogeny, where the action of the environment transforms the genotype
information from potentiality to actuality in the form of a growing phenotype or organism, including behavior and
learning. I suggest, therefore, to consider the following analogy: the basic ideas of the dynamic theory are
analogous to the genotype, whereas the explicit theory develops as the phenotype. The analogy implies the
following main points:
1.The basic ideas are determined with the initiation of the research program, as the genetic information
of an organism is fixed at conception.
2.The basic ideas of the model can be seen as providing a program or a blueprint for developing the
explicit theory, as the genetic information represents a blueprint for the development of the organism.
3.The environment within which a research program evolves is comprised of observational data and
other developing research programs. This is analogous to the fact that the environment for an
organism is comprised of other organisms, as well as the non-organic environment.
4.The "central dogma" of molecular genetics says that information flows in the cell only in one
direction: from the DNA molecules to the RNA molecules, which direct protein synthesis on
ribosomes. The process cannot be reversed; phenotypic modifications that are brought about by the
interaction between organism and environment cannot influence the hereditary information encoded
in the DNA. Similarly, the relation between basic ideas and theory is unidirectional; the basic ideas
direct theory construction, but are never modified as a result of theory-observation interaction, i.e. by
the attempts to adjust the theory to the data.
5.The decoded genetic information can guide cell development only if the appropriate environmental
inputs are provided. Similarly, the basic ideas can be decoded, that is expressed explicitly in the form
of instructions for constructing the theory, only if the appropriate experiments are made and the
appropriate observational data are provided. The basic ideas of a research program are by themselves
devoid of empirical content in the Popperian sense, since they are not falsifiable. They are not
falsifiable since they do not entail possible observations that may refute them, and sometimes they are
not explicitly formulated. This is the reason why they can be kept untouched. The empirical content
is built up as the research program progresses. For example, as we have seen, in the Bohr-atom
research program some of the model's elements were left without clear specification at the beginning.
The initial version of the model left open, for example, the implications for atomic structure of the
finite size of the rotating electron-planet, such as its degree of freedom of rotation around its axis.
The open points constitute the neutral analogy of the model. Only the attempts to explicate and
formulate the basic ideas and to utilize unambiguously the neutral analogy give these ideas empirical
content.
< previous page
page_245
next page >
page_246
< previous page
page_246
next page >

Page 246
Thus, the epistemology manifested in the genotype-phenotype (G-P) model for the growth of scientific
theories stands midway between "nativism" and empiricism; the growth of scientific knowledge within
the framework of a research program is a product of empirical data and preexisting "innate" (basic)
ideas. The "innate'' ideas are not a priori valid or necessary; they only remain unquestionable
throughout the lifetime of the research program. These basic ideas guide theory-construction in the
framework of the research program and in this process they acquire empirical content. Thus, the
heuristic rules for theory-construction are derived from these ideas (Lakatos makes the distinction
between the "hard core" and the "positive heuristic," but as we can see here, heuristic rules are derived
from these same ideas, which are kept untouched). These heuristic rules include the tacit, or
transparent, presuppostions and, in particular, the material rules of inference shared by the community
of investigators. Thus, a research program starts to develop after these conceptual tools have been
creatively generated and have become black-boxed. The research program can be viewed as a tool for
exposure, i.e. for unfolding the theory's content, or for exposing its potentialities. In other words, the
research program signifies the inference phase in the discovery process, which is governed by the
material rules of inference.
Schematically, the growth of an organism and the development of a theory can be represented by the
following parallel formulas:
Ontogeny: environmental conditions + genotype developing phenotype
Dynamic theory: data + basic ideas developing theory,
where the "plus" sign represents interaction, and the "data" refers to theoretical results obtained in other
research programs, as well as to observational data. Thus, following this analogy, the logic of
discovery by exposure in the framework of a research program is the logic of ontogeny. In fact, if we
compare the above formula for a dynamical theory with the data-driven inference pattern:
RP: e&T[HEU]T'
mentioned in section 2.2.6, we find that they are identical, provided the heuristic rules are derived from
the basic ideas and the latter do not change in the transition from T to T'. In the RP pattern each new
piece of data (e) entails, via the inference rules (or heuristic rules), the manner by which the theory will
be modified or reconstructed. It is by no means an inductive pattern, since the basic ideas determine the
theory-change no less than the data. The basic ideas may be regarded as the "innate" ideas of the
research program.
In the above model the conflict between nativists and empiricists can be translated into the "nature vs.
nurture" debate. The attempts to construct a balanced biological theory that integrates genetic and
environmental influences on the developing phenotype suggest that historiographers of science
< previous page
page_246
next page >
page_247
< previous page
page_247
next page >

Page 247
should attempt to construct a theory of science that will give appropriate weights to the influence of
the basic ideas inherited by the research program, as well as to the external data, in constructing a
scientific theory. As an organism cannot begin life without genetic information, so also a research
program cannot start without basic or "innate" ideas. As an organism cannot grow without input from
the environment, so a dynamic theory cannot grow without interacting with observational data and
other external information. The manner in which the genetic information is expressed in the growing
organism depends on the organism's diet and exposure to environmental pressures. Similarly, the
manner in which the basic ideas are expressed in the developing theory is influenced by the exposure
to experiments and to inputs from other research programs. Thus, theory building is not blind to the
data; the same master plan or blueprint will yield different theoretical constructions when different
"environmental" inputs are provided, although they will have some basic characteristics in common.
6.The category difference between the genotype and the phenotype is reflected in the parallel category
difference between the "metaphor" and the "algebra," the basic idea and the formulated theory, the
implicit picture and the explicit theory. On the one side of the relation stands a potential entity, an
innate design, which materializes into an actual entity; a visible or explicit construction stands on the
other side of the relation. Furthermore, in biology the genotype is a "theoretical" entity, whereas the
phenotype is "observational." A similar distinction may be made in the framework of a naturalistic
theory of science. The most direct and unambiguous data available for the theorist of science are the
explicit scientific claims and the manifest behavior of the scientists, whereas their tacit assumptions,
transparent rules of inference and the basic ideas that guide them can mainly be inferred or theorized.
Hence, for the theorist of science these background models, tacit ideas or material rules of inference
are "theoretical'' entities, whereas an explicit theory-version or a statement is directly "observable."
7.From the evolutionary point of view, the difference between the genotype and the phenotype is that
natural selection operates directly only on the latter. Selection in science operates in a similar way:
only explicit theory-statements can clash with observational data and be selected by the scientific
community, whereas basic ideas of a research program cannot, since they are not linguistic entities
and are not fully communicable. The question of what the unit of selection is, whether it is an idea, a
(dynamic) theory, or perhaps an entire world picture or "paradigm" has a direct counterpart in
biology. The parallel question in biology pertains to "the level (or levels) at which selection can take
place" (Hull 1981, 23). In our model we adopt the view that parallels the current majority view in
biology: "that genes mutate, organisms are selected, and species evolve" (ibid.). A basic idea is
analogous to a gene, and a theory is analogous to an organism. Our unit of direct selection is the
theory, whereas
< previous page
page_247
next page >
page_248
< previous page
page_248
next page >

Page 248
basic ideas are indirectly selected through the selection of entire theories, just as genes are selected
indirectly through the selection of entire organisms.
8.We still have to determine what in our model constitutes the analogue to the species. It is natural to
choose the population of research programs that evolve within a given paradigm or world picture as
the analogue of a species. If we take the definition of a species to be a set of individuals that can
interbreed, then we can see that the analogy will work if indeed only research programs within one
paradigm or tradition can communicate and contribute ideas to produce new research programs. This
reminds us of the incommensurability thesis: research programs from different paradigms cannot
intercommunicate. The species can also be characterized by its genotype (the genotype of the species)
or by its gene-pool. Similarly, a population of intercommunicating research programs is characterized
by their common world picture, or by its repertoire of ideas ("idea pool"). A research program that
survives and succeeds in "mating" with other ones contributes some of its basic ideas to its
successors so that these ideas disseminate in the normal science and will appear in high percentage in
the repertoire of ideas of the paradigm. At any given time, the world picture in normal science is a
product of natural selection in antecedent generations of research programs. The information stored in
the world picture reflects the long phylogenetic history during which environmental pressures
produced and selected the currently selected ideas ("genes").
In other words, only via selection operating on dynamic theories can information about the world be transmitted to
the repertoire of ideas of the paradigm, i.e. to the world picture. Thus, according to this model, there is no way to
pass on to the common world picture the information that has been acquired in the course of the development of
the theory. Hence, our model would predict, for example, that the information included in the advanced versions of
Bohr's theory will not be transmitted to the hard core of successive research programs. This seems to be a serious
shortcoming of the model. Although we do not expect the idea-pool to include explicit theoretical developments of
the model, it should include new significant ideas, such as spin, which were developed in a successful research
program. I shall give three arguments for answering this question.
The first one is the following. The G-P model has descended from the Popperian model, which claims that nothing
can be learned positively in science; the only thing we can learn from experience is that a theory was refuted. The
G-P model constitutes in this respect a considerable advance over falsificationism without being inductive: it
implies that our repertoire of ideas gains positive knowledge by increasing the relative weights of basic ideas which
have guided successful research programs. This pool of ideas serves as a source of basic ideas for subsequent
research programs. However, this is only a slight consolation. Let us turn, then, to a stronger argument.
< previous page
page_248
next page >
page_249
< previous page
page_249
next page >

Page 249
There are two ways through which information acquired in advanced stages of a research program will be
transmitted to other research programs, though not to their "hard cores" of basic ideas. (a) A progressive research
program leads to novel experiments which produce new types of data; namely, it changes the environment to
which current and future research programs will have to adapt. For example, following Bohr's theory, research
programs have had to explain the vast amount of new spectroscopic data gathered throughout Bohr's research
program. (b) Also the theoretical results arrived at in an advanced theory-version of a research program constitute
part of the environment to which other research programs should adapt. In other words, a research program
transmits its advanced knowledge through the changes it brings about in the experimental and theoretical
environment; future generations of research programs should adapt to this changing environment. Thus, although a
research program does not pass on information obtained during the course of its development to the "hard cores" of
ideas of its offspring, it communicates this information to other research programs, which should cope with it and
accomodate it to their theoretical systems.
The third argument is the strongest. The "hard core" may include "dormant" ideas which were inherited from
previous research programs but remained inactive in the present research program. These ideas may be activated in
order to solve a problem which arises in the research program. If the problem is successfully solved, the research
program survives and bequeaths the ideas to its successors. The idea of spin (inherited from classical mechanics)
which was activated in the Bohr-atom program might fit the above description. The result is a serendipitous
development, where the "dormant" ideas, which were not explicitly included in the hard core of basic ideas and
were not intended to solve the present problem, are disseminated throughout the population of research programs.
A similar kind of serendipitous development which affects the idea pool can be represented as the counterpart of
the following kind of evolutionary phenomenon (mentioned in earlier chapters). A species which was "designed" or
selected to cope with a certain kind of environmental conditions, A, manages to cope with a new kind of
environmental conditions, B, by exploiting a "redundant" and rare trait which had not endowed its carriers with a
selective advantage. For example, a hypothetical species of short-necked giraffes who are fed by leaves growing on
low branches encounters a new situation where only tall branches are available. Individuals who possess the very
rare trait of having a long neck exploit this trait for coping with the environmental change. As a result this trait
spreads throughout the entire population via differential reproduction. The corresponding serendipitous
development takes place when a research program which was designed to solve a certain kind of problems, solves
a new kind of problems by exploiting a redundant and probably problematic idea which was implied by its hard
core. The idea may be, for
< previous page
page_249
next page >
page_250
< previous page
page_250
next page >

Page 250
example, an undesired negative analogy of the basic model which guides the development of the program. The idea
is then disseminated throughout the entire population of research programs belonging to the same "paradigm."
9.2 The Selection Cycle in Science
We can summarize the picture outlined above by a cybernetic model of the selection process offered by C. J.
Bajema (1971, 2). The model describes how the information flows from the organism-environment interaction to
the gene pool and back. The organism (the phenotype) is produced via genotype-environment interaction in the
ontogenetic process. The organism in its turn adapts to (interacts with) the changing environment. Subsequently,
the process of differential reproduction affects the content of the gene pool. The gene pool undergoes further
variations via mutations and recombinations. This affects the genetic makeup of the organisms in the next
generations, and so on. There are two sources of change in this cycle: changes in the environment and variation
input to the gene pool.
We can translate the above biological feedback cycle into its counterpart in the evolution of science. The diagram
below describes the corresponding information cycle in science.
It must be noted that there is some overlap between the G-P process (theory-construction) and the theory-data
interaction. The latter refers, however, mainly to a mature stage in the development of the theory. The sources of
novelty in the evolution of science are two: discovery of new observational data and generation of new ideas. Both
sources are necessary for the process to go on. Radically new observational data could not be accommodated, i.e.
explained, without generation of novel ideas. New ideas could not become assimilated into the world picture
without being confirmed by some new data (or without being selected by the new data).
< previous page
page_250
next page >
page_251
< previous page
page_251
next page >

Page 251
<><><><><><><><><><><><>
Finally, I would like to make the following remark. There have been some attempts to identify the G-P process in
science with the intellectual development of the individual scientist. However, according to our approach, science
is viewed as the continuation of the evolutionary process on the sociocultural level, extending our collective
cognitive apparatus and sensorimotor organs. Hence, ontogeny on this level does not coincide with the individualscientist ontogeny. The unit of selection according to our theory is a theory or maybe an idea; an individual
scientist does not die with his ideas or theories. In our model it is thus a mistake to look for the G-P process in the
intellectual development of the individual scientist.
Peter Munz demonstrates the similar roles played by an organism and a theory in the following way. He argues that
since an organism is equipped with a cognitive structure it can be viewed as "an embodied theory about its
environment" (Munz 1989, 241). "It takes signals or stimuli from the environment and immediately interprets or
decodes them in terms of its structure and then responds. This is exactly what happens when we test a theory."
Thus human beings "propose laws to nature" (rather than prescribe laws to nature, in the Kantian fashion).
However, this argument leads to another conclusion. The laws that we propose to nature are determined by our
cognitive apparatus which develops in an ontogenetic process (as was described, for example, by Piaget).
Therefore, when we say that we are theories, the latter should be understood as dynamic theories. On the other
hand, Munz claims that ''consciously proposed hypotheses can be regarded as disembodied organisms" (ibid., 243).
Thus in science we "continue and extend the process of evolution in our minds" by generating and selecting
theories. We therefore arrive at the unavoidable conclusion that the disembodied organisms which we create and
propose to nature are dynamic entities.
In summary, the basic G-P formula is not a rule or a method of discovery. Rather it describes a general kind of
process, which may have psychological and social dimensions and which may be cultivated by supplying to it the
appropriate environmental conditions. The main lesson for cultivating discovery which can be drawn from this
model is that a proposed idea or theory cannot be treated as a discovery in isolation from its environment, where
the environment includes, in addition to the data, metaphysical, technological and social dimensions. A theory or
an idea may develop into a great discovery in a given context but not in another. As we have observed, the
question whether a given idea or theory is a discovery is context-dependent. In the same way, the question whether
an idea will develop into a discovery is also context-dependent. Thus, in assessing the prospects of a proposed
idea, we cannot be satisfied with examining only the idea itself; we have to examine also the environment: what is
the state of knowledge, the experimental techniques available, and the climate of opinion within the scientific
community.
< previous page
page_251
next page >
page_252
< previous page
page_252
next page >

Page 252
In the above scheme I have exploited the evolutionary model in some more detail then it is customary to do in the
practice of evolutionary epistemology. Of course, there is a limit to the possibility of exploiting the neutral analogy
of a model. A scientific theory is not a living creature in the physical world, nor is normal science a population of
such creatures. Nevertheless, I have proposed the model since there seems to be a high degree of structural
similarity between the growth of a theory and ontogeny. True, there is also some negative analogy in the model, so
that the balance between the negative and the positive analogy will determine the fate of the model.
< previous page
page_252
next page >
page_253
< previous page
page_253
next page >

Page 253
Conclusion
In Chapters 5 through 9, I have proposed a naturalistic, or an explanatory, theory of science. According to the
theory, science advances via two intertwined creative processes which are manifestations of evolution through
natural selection. These two processes feed one another with the raw material of variations. The intrapsychic
process is fed by ideas created and selected in the interpsychic process, whereas the interpsychic process is
nourished by ideas generated in intrapsychic processes. Novel ideas are injected into the system via tinkering. Thus
the system remains open and proliferation of variations is ensured. The coupling of these two kinds of creative
processes form a system that can be viewed as the collective brain of the scientific community, which extends
humankind's cognitive capacities. The extended cognitive apparatus, therefore, magnifies creativity and the
resulting creative process is very powerful and enables science to transcend the limits imposed on human
knowledge by the evolutionary heritage of our species. Modern science may, therefore, be viewed as a new
evolutionary line on the sociocultural level, which emerged around the sixteenth to seventeenth centuries. From this
perspective, science is viewed as a contingent phenomenon.
Arthur Fine (1986, 174175) considers the possibility that science is a contingent "historical entity ... like a
particular speciesthe horse for example." 11 He argues that as there is no "science of the horse," so there is no
science of science; there can be no science devoted to an individual (a historical entity). Yet, there is one
exception: the human species. Fine's claim would not apply to this contingent historical entity. More than one
science is devoted to the study of this species: for example, psychology, sociology, human biology, human
ethology, anthropology, the medical sciences and economicsin shortthe human sciences. There is no reason why at
least one science will not be devoted to the study of science, which extends our natural capabilities and carries us
beyond our evolutionary "home-base.''12
The above theory of science has implications for the discoverer who participates in the process of science. An
inevitable consequence of the natural-selection model is that the creative steps in the evolution of science are the
products of serendipity, opportunism and tinkering. These creative facul-
< previous page
page_253
next page >
page_254
< previous page
page_254
next page >

Page 254
ties are not method-governed, but they require a prepared mind. Furthermore, these notions imply that science has
no goal. From the vantage point of the individual discoverer, there are two major involuntary or unintentional
novelty-generating mechanisms in science. One is a creative process which the scientist hosts in his mind. The
other is the cooperative enterprise in which the scientist participates. In both cases the discoverer is a passive
observer on the arena of creation.
Yet, there is an indispensable role for logic and method in scientific discovery. When a new conceptual tool or a
new information channel with nature is created, logic enters into the game after the new channel is black-boxed
and becomes transparent. At this stage, scientists try to expose new phenomena and new aspects of reality by using
the new information channel. And they do this by using logic, in the broad sense, which serves as a tool for
uncovering the information conveyed by the new channel. In this case the discovery is not generational or creative,
rather it is discovery by exposure. Logical rules, or material rules of inference, serve as tools of exposure. This is a
situation analogous to the uncovering of the information hidden in a set of statements by using deductive rules.
This role of logic is no less important for discovery than the role played by involuntary processes and tinkering; the
latter create new ideas and logic consolidates them as discoveries. The general pattern of scientific progress is that
of creation followed by exposure and vice versa. A new idea or theory is created and then its potentialities are
exposed. Novelty may arise from an unintentional deviation from a strict procedure. The intentional activity of the
scientist takes place within research programs, where he is trying to solve problems and to convert his ideas into
explicit formulations. Unintentional creative leaps may appear as by-products of this activity.
< previous page
page_254
next page >
page_255
< previous page
page_255
next page >

Page 255
Epilogue
Implications For Science Education
Today's science teaching reflects yesterday's philosophy of science. Several decades ago, it was inductivism or
positivism which dominated educational thought. It was taught that the scientist learns from experience by
collecting data and generalizing or finding regularities. And so does the child in the classroom; he is just a small
discoverer of objective truth. The impact of the Popperian revolution on science teaching was in undermining
scientific dogmatism. The most sacred theories did not fall from the sky; they are manmade and therefore
replaceable. This brought into the classroom a fresh pluralistic and permissive atmosphere. The Kuhnian revolution
contributed to the overthrow of the myth that observation and experiment determine the fate of science. The
theories of science and the views of discovery expounded in this book reflect the naturalistic turn in the philosophy
of science. The hard sciences, such as physics, become somewhat softer. In fact, they are explained by the softer
sciences. Instead of method-governed discovery, we have here psychological and social phenomena dominated by
tinkering.
The notions of exposure and generation, the socio-evolutionary view of science, the principle of serendipity and
the strategy of the tinkerer have direct implications for science education and science teaching. In the following, I
list some of these implications.
A Lesson from the Generational Character of Discovery
One of the modern trends in science teaching is that the child should learn a new concept or a law of nature by
discovering it himself through observation, experimentation and reasoning. This view implies that the child is faced
with a question which has a unique answer like an arithmetic exercise. The fact that physics and mathematics are in
general taught together, in many cases by the same teacher, may reinforce this impression in the child. A unique
answer to a problem may be given when the child is faced with a normal-science situation,
< previous page
page_255
next page >
page_256
< previous page
page_256
next page >

Page 256
when the teacher supplies him with a hint such as what law or theory to use in order to explain a given
phenomenon. A problem in physics may become an exercise with a unique answer if the teacher leaves only a
small missing gap between the premises and the conclusion. The child learns how to apply a well defined law to a
well defined situation. This is very far from treating the child as a "little scientist."
There are two contexts of science teaching. The first is the context of applied science, where problems can be
solved in the above fashion, without entering into a "philosophy." The second context is that of learning a new
concept or law. In this context, the child should be taught that a scientific law or theory is not a final truth. And,
further, that scientific concepts and theories are not derived from observation or experiments. Rather they are
invented or generated by scientists and then tested by experiment. Here, it might be helpful to give examples of
unsuccessful theories, such as the theory of phlogiston, the caloric or Praut's theory, as well as of successful
theories which were replaced, such as classical mechanics.
In class we can hope to reconstruct some cases of discovery by exposure (discovery by observation, classification,
computation and inference). We cannot reconstruct generational discovery, since different theories can be
constructed and there is no unique "right" answer; theories are underdetermined by the data. The standards of
explanation change with the scientific paradigm or world-picture. Creativity in science is influenced by
metaphysical beliefs and cultural-technological background. The child's background differs from that prevailing at
the time of the real discovery.
A Lesson from the Social Dimension of Scientific Creativity
If we adopt the view that scientific discovery is a cooperative enterprise, we should not expect the child to achieve
what even individual scientists do not achieve. Most scientists do not make even a single discovery by themselves
throughout their lifetime. They participate in the process of discovery or they contribute to it. The only thing which
can be done in order to simulate some facets of the discovery process in class is to try to set up working groups
which will develop ideas, without expecting them to arrive at the "right" ideas, since those do not exist.
The above view seems to be confirmed by empirical research showing that the cognitive process of learning
scientific concepts, such as the concept of velocity, is much more efficient when done as part of a small group
activity than when carried out individually through reading textbooks or through frontal teaching. The general
hypothesis is that internalization of unfamiliar concepts is more efficient as collective learning. This bears some
similarity to what happens in science, when novel ideas are accepted only after being
< previous page
page_256
next page >
page_257
< previous page
page_257
next page >

Page 257
processed (and possibly modified) by the scientific community. Brainstorming (Osborn 1963) is the best available
technique for encouraging epistemic cooperation. According to this method, a discussion group is set up for
solving a problem. Each participant in the group raises ideas, as "wild" as possible, without criticism or disturbance
from the other members. The initial ideas are improved, and combinations of different ideas are generated, to
produce a collective solution. This stage is later supplemented by a mutual criticism.
A Lesson from Serendipity and Tinkering
Perhaps most major creative discoveries were made unintentionally: when the discoverer solved problem B while
aiming at solving a different problem A. We cannot expect the child to intentionally discover laws, theories,
principles, etc. which were discovered in a serendipitous process. In class, the child is expected to solve problems
in a routine fashion. However, one way to stimulate creativity is to ask the child to be alert to different problems
which might be solved along the way in addition to, or instead of, the original problem. In order to encourage this
kind of creativity, the teacher should not insist on solving the original problem. By this we can simulate
serendipity. We can also encourage the child to try solving the original problem by using tools and methods which
were devised for other purposes. This lesson is good for solving problems in any area, such as natural science,
mathematics or in practical fields.
"Ontogeny Recapitulates Phylogeny"
The above slogan is well rehearsed by the students of science education. One of Jean Piaget's central claims in his
genetic epistemology is that the intuitive views of the child mirror earlier stages in the history of science. There is
a parallelism between the history of science and individual development (Piaget 1977). This view even led Kuhn
to say that in order to understand dead scientists, he had to study living children. Some people go as far as to claim
that they found Aristotelian physical thought in erring children. This view may sound plausible if we refer to
genetically based scientific conceptions. If our intuitive conceptions are inborn, we have the potential for
developing the same conceptions as our ancestors.
As long as we are dealing with teaching empirical knowledge and some basic laws of classical physics, the
Piagetian view might be true. However, in view of the fact that modern science has carried us beyond our
evolutionary "home-base" and at the same time has become increasingly more cooperative, the Piagetian
conception is inadequate in accounting for the evolution of sci-
< previous page
page_257
next page >
page_258
< previous page
page_258
next page >

Page 258
ence, in particular of twentieth-century physics. Hence, I would not agree with the following statement of Nancy
Nersessian, which appeared in a recent collection on the impact of philosophy of science on science education:
"there should be a single cognitive model for conceptual change in science and in learning science" (Nersessian
1989, 178). For example, the conceptual change leading to the quantum revolution involved a long historical
cooperative and serendipitous process and tinkering. The process which started with Boltzmann and was continued
by Planck, Einstein, Rutherford, Bohr, Sommerfeld, Goudsmit and Uhlenbeck and then carried on by the collective
efforts of some of the greatest minds of twentieth-century physics cannot possibly be reconstructed by a single
innocent child. This conceptual change was a social and cultural change rather than a cognitive change. It was so
complex, that even some of the major participants in the process, such as Planck and Einstein, could not
comprehend its results.
Skill, Mechanized Discovery and the Child
The above objections are similar to the objections I raised against mechanized discovery. The latter is inappropriate
for replicating unintentional, involuntary or natural processes of scientific discovery. These include serendipitous,
subconsious and collective processes and discoveries generated by discernment and by tacit faculties. What is
common to the machine and the child is that both lack discerning power or tacit faculties which can only be
acquired by apprenticeship and by active participation in scientific research. One of the shortcomings of computer
technology is that it cannot replicate the social dynamics of science. The social dimension of science cannot be
incorporated into science teaching too. The small group activity mentioned above cannot, of course, replicate the
intricate social dynamics of science, since this relies on institutions and social patterns which cannot be
implemented in the classroom. The classroom is not a miniature scientific community and the child is not a little
scientist. Hence, the most important kinds of scientific discovery cannot be replicated either by the computer or by
the child.
< previous page
page_258
next page >
page_259
< previous page
page_259
next page >

Page 259
Notes
1. EE has gained much interest in recent years. The bibliography in Donald Campbell's original paper (1974b)
included nearly 100 references. The bibliography since then has exploded and its later version (Cziko and
Campbell 1990) consists of nearly 1200 references (during the years 19871990, it was doubled).
2. I would like to make a methodological remark with respect to the notion of explication. This notion has been
extensively used in traditional analytic philosophy. For example, the notions of probability and confirmation were
explicated in the logical empiricist tradition by Rudolf Carnap (1950) or Mary Hesse (1974). According to this
conception, we start from an intuitive notion which we believe to be shared by most language speakers with whom
we feel we can communicate. And then we look for a clear definition of the concepts (or embed the concepts in a
formal sytem) which will capture at least some of these intuitions. We do not have to conduct field study in order
to find out with whom we share these intuitions. We rely on our experience as language speakers. It may turn out
that the community with which we share these intuitions is narrow. But it should consist at least of the relevant
professional community we belong to. And since we are not isolated from the rest of society, these intuitions
probably reflect the intuitions shared by some of the general population. Empirical studies must be conducted only
if we want to study the usage of a term in culturally remote societies with whom we have no communication.
The above procedure applies to concepts which appear in ordinary discourse. If one tries to explicate a
metascientific notion such as the notion of discovery as it is used by scientists, then it depends on who is
making the explication. A historian of science, or a philosopher of science, who is not an active scientist, or
has never belonged to the scientific community, should interview scientists or study the history of science in
order to find out about the scientists' intuitions. Someone who is an active scientist may use the above
procedure of explication, since he has had experience in communicating with other scientists. But this is true
for his specific scientific community. The notion may be used differently in other scientific communities. There
is no reason to assume that all metascientific terms are used in the same way by biologists and physicists, for
example, or even by particle physicists and astrophysicists.
In analyzing the notion of discovery in ordinary discourse, I will rely on dictionary definitions, on the writings
of other philosophers of science, as well as on my own intuitions. Since the metascientific notions have evolved
from ordinary discourse, this might help us in understanding the metascientific usages of the term.
< previous page
page_259
next page >
page_260
< previous page
page_260
next page >

Page 260
3. I borrow this term from David Gooding, Thomas Nickles and Simon Schaffer (Gooding et al. 1989). These
authors employ the term with respect to the practices of scientists in using their experimental tools.
4. The numbers within the brackets indicate the masses in Mev.
5. I follow here the description of Edward Neville da Costa Andrade, who worked under Rutherford at the
University of Manchester (Andrade 1964).
6. I do not commit myself to all the interpretations provided by Koestler in this book. I try to rely only on
uncontroversial material.
7. Throughout this chapter, all page numbers appearing in quotations, without further specification, refer to this
book.
8. For further details see I. B. Cohen 1981, 123133.
9. For further details see DeLair and Sarjeant (1975) and Desmond (1975).
10. For further details see Pickering (1984).
11. David Hull (1980) treats biological species as historical entities or as individuals. According to the version of
evolutionary epistemology expounded in this book, science is not "like a particular species." Rather, it is like an
evolutionary line which splits into several branches. But an evolutionary line, according to Hull's interpretation,
can also be viewed as a contingent historical entity, or as an individual.
12. I use here Nicholas Rescher's expression (1986).
< previous page
page_260
next page >
page_261
< previous page
page_261
next page >

Page 261
Bibliography
Amundsen, R. 1989. The traits and tribulations of selectionist explanations. In Hahlweg and Hooker 1989, 413-32.
Andrade, E. N. da C. 1964. Rutherford and the structure of the atom. New York: Doubleday, Anchor Books.
Ayer, A. J. 1952. Language, truth and logic. New York: Dover.
Baggott, J. 1990. Serendipity and scientific progress. New Scientist no. 1706: 67-68.
Bajema, C. J. 1971. Natural selection in human populations. New York: John Wiley.
Barnes, V. E. et al. 1964. Observation of a hyperon with strangeness minus three. Physical Review Letters 12: 2046.
Black, M. 1962. Models and metaphors. Ithaca: Cornell University Press.
Bronowski, J. 1970. New concepts in the evolution of complexity: Stratified stability and unbounded plans. Zygon
5: 18-35.
Buchanan, B. 1982. Mechanizing the search for explanatory hypotheses. In P. Asquith and I. Hacking, eds. PSA,
vol. 2, 129-46. East Lansing, Mich.: Philosophy of Science Association.
Buckley, W. 1972. A systems approach to epistemology. In Trends in general systems theory, G. J. Klir, ed. 188202. New York: John Wiley.
Campbell, D. T. 1974a. Unjustified variation and selective retention in scientific discovery. In Studies in the
philosophy of biology, F. J. Ayala and T. Dobzhansky, eds. 139-61. London: Macmillan.
Campbell, D. T. 1974b. Evolutionary epistemology. In The philosophy of Karl Popper, vol. 1, P. A. Schilpp, ed.
413-63. La Salle: Open Court.
Cannon, W. 1961. Gains from serendipity. In The way of an investigator. New York: Hafner.
< previous page
page_261
next page >
page_262
< previous page
page_262
next page >

Page 262
Carnap, R. 1950. Logical foundations of probability. Chicago: University of Chicago Press.

Chisholm, R. 1982. The foundations of knowing. Brighton: The Harvester Press.
Chomski, N. 1957. Syntactic structures. London: Mouton.
Chomski, N. 1966. Cartesian linguistics. New York: Harper and Row.
Chubin, D. E. and S. Restivo, 1983. In Knorr-Cetina and Mulkay, 53-83.
Cohen, I. B. 1981. Newton's discovery of gravity, Scientific American vol. 244, no. 3: 123-33.
Cohen, L. J. 1981. Can human irrationality be experimentally demonstrated? The Behavioral and Brain Sciences 4:
317-31.
Collins, H. M. 1975. The seven sexes: A study in the sociology of a phenomenon, or the replication of experiment
in physics. Sociology 9, 205-24.
Cushing, J. T. 1990. Theory construction and selection in modern physics: The S matrix. Cambridge: Cambridge
University Press.
Cziko, G. A. and D. T. Campbell. 1990. Comprehensive evolutionary epistemology bibliography. Journal of Social
and Biological Structures 13: 41-82.
Darwin, C. 1958. The autobiography of Charles Darwin. London: Collins.
DeLair, J. B. and W. A. S. Sarjeant. 1975. The earliest discovery of dinosaurs. Isis 66, 5-25.
Descartes, R. 1967. Discourse on the method of rightly conducting the reason, part II. In The philosophical works
of Descartes, vol. I. E. S. Haldane and G. R. T. Ross, eds. Cambridge: Cambridge University Press.
Desmond, A. J. 1975. The hot blooded dinosaurs. London: Blond and Briggs.
Dobzhansky, T., F. J. Ayala, G. L. Stebbing and J. W. Valentine. 1977. Evolution. San Francisco: W. H. Freeman
and Company.
Donovan, A., L. Laudan and R. Laudan. 1988. Scrutinizing science: Empirical studies of scientific change.
Dordrecht: Kluwer Academic Publishers.
Dubos, R. J. 1951. Louis Pasteur: Free lance of science. London: Gollancz.
Edwards, W., H. Lindman and L. Savage. 1963. Bayesian statistical inference for psychological research.
Psychological Review 70: 193-242.
Fermi, E. and C. N. Yang 1949. Are mesons elementary particles? Physical Review 76: 1739-43.
< previous page
page_262
next page >
page_263
< previous page
page_263
next page >

Page 263
Feyerabend, P. 1978. Against method. London: Verso.

Fine, A. 1986. Unnatural attitudes: Realist and instrumentalist attachments to science. Mind 95: 149-79.
Finocchiaro, M. 1988. In Donovan et al. 49-67.
Fodor, J. 1981. Representations: Philosophical essays on the foundations of cognitive science. Cambridge, Mass.:
MIT Press.
Gell-Mann, M. 1953. Isotopic spin and new unstable particles. Physical Review 92: 833-34.
Gell-Mann, M. 1961. The eightfold way: A theory of strong interaction symmetry. Caltech Synchrotron Laboratory
Report CTSL-20.
Gell-Mann, M. 1964. A schematic model of baryons and mesons. Physics Letters 8: 214-15.
Gell-Mann, M. and Y. Ne'eman. 1964. The eightfold way. New York: Benjamin.
Giere, R. 1988. Explaining science. Chicago: University of Chicago Press.
Goldberg, H. and Y. Ne'eman. 1963. Baryon charge and R-inversion in the octet model. Nuclear Physics 27: 1-5.
Gooding, D., T. Pinch and S. Schaffer. 1989. The uses of experiment: Studies in the natural sciences. Cambridge:
Cambridge University Press.
Goodman, N. 1965. Fact, fiction and forecast, Second edition. Indianapolis: Bobbs-Merrill.
Gursey, F., A. Pais and L. A. Radicati. 1964. Spin and unitary spin independence of strong interactions. Physical
Review Letters 13: 299-301.
Hacking, I. 1967. Slightly more realistic personal probability. Philosophy of Science 35: 311-25.
Hahlweg, K. and C. Hooker, eds. 1989. Issues in evolutionary epistemology. Albany: State University of New York
Press.
Hanson, N. R. 1958. Patterns of discovery. Cambridge: Cambridge University Press.
Hempel, C. 1965. Aspects of scientific explanation. New York: Free Press.
Herschel, J. F. W. 1830. A preliminary discourse on the study of natural philosophy. London: Longman.
Hesse, M. 1966. Models and analogies in science. University of Notre Dame Press. Ind.
< previous page
page_263
next page >
page_264
< previous page
page_264
next page >

Page 264
Hesse, M. 1974. The structure of scientific inference. London: Macmillan.

Hofmann, J. R. 1988. In Donovan et al. 201-17.
Hull, D. 1980. Individuality and selection. Annual Review of Ecology and Systematics 11: 311-32.
Hull, D. 1981. Units of evolution: A metaphysical essay. In U. J. Jensen and R. Harre, eds. The philosophy of
evolution. Brighton: The Harvester Press.
Hull, D. 1988. A mechanism and its metaphysics: An evolutionary account of the social and conceptual
development of science. Biology and Philosophy 3: 123-55.
Hurd, D. L. and J. J. Kipling, eds. 1964. The origin and growth of physical science, vol. 2. Harmondsworth,
Middlesex: Penguin Books.
Jacob, F. 1977. Evolution and tinkering. Science 196: 1161-66.
Kantorovich, A. 1978. An ideal model for the growth of knowledge in research programs. Philosophy of Science
45: 250-72.
Kantorovich, A. 1979. Towards a dynamic methodology of science. Erkenntnis 14: 251-73.
Kantorovich, A. 1982. Quarks: An active look at matter. Fundamenta Scientiae 3: 297-319.
Kantorovich, A. 1983. The collective a priori in science. Nature and System 5: 77-96.
Kantorovich, A. 1988. Philosophy of science: From justification to explanation. The British Journal for the
Philosophy of Science 39: 469-94.
Kantorovich, A. 1989. A genotype-phenotype model for the growth of theories and the selection cycle in science.
In Hahlweg and Hooker, 171-84.
Kantorovich, A. 1990. Naturalistic philosophy of science: A socio-evolutionary view. Journal of Social and
Biological Structures 13: 259-77.
Kantorovich, A. and Y. Ne'eman. 1989. Serendipity as a source of evolutionary progress in science. Studies in
History and Philosophy of Science 20: 505-29.
Keynes, M. 1921. A Treatise on probability. London: Macmillan.
Klein, M. J. 1966. Thermodynamics and quanta in Planck's work. Physics Today November.
Knorr-Cetina, K. D. 1981. The manufacture of knowledge. Oxford: Pergamon Press.
Knorr-Cetina, K. D. 1983. In Knorr-Cetina and Mulkay, 115-40.
< previous page
page_264
next page >
page_265
< previous page
page_265
next page >

Page 265
Knorr-Cetina, K. D. and M. Mulkay, eds. 1983. Science observed. Beverly Hills, Calif.: Sage.
Koertge, N. 1980. In Nickles, 139-58.
Koestler, A. 1964. The sleepwalkers. Harmondsworth, Middlesex: Penguin Books.
Kohn, A. 1989. Fortune or failure: Missed opportunities and chance discoveries. Oxford: Basil Blackwell.
Kornblith, H. ed. 1985. Naturalizing epistemology. Cambridge, Mass.: MIT Press.
Krohn, W., E. T. Layton, Jr. and P. Weingart, eds. 1978. The dynamics of science and technology. Dordrecht: D.
Reidel.
Kuhn, T. 1962. The structure of scientific revolutions. Chicago: University of Chicago Press.
Kuhn, T. 1963. The essential tension: Tradition and innovation in scientific research. In C. W. Taylor and F.
Barron, eds. Scientific creativity. 341-45. New York: John Wiley.
Lakatos, I. 1970. Falsification, and the methodology of scientific research programs. In Criticism and the growth of
knowledge. I. Lakatos and A. Musgrave, eds. 91-196. Cambridge: Cambridge University Press.
Lakatos, I. 1971. History of science and its rational reconstructions. In Boston studies in the philosophy of science,
vol. 8, R. C. Buck and R. S. Cohen, eds. 91-135. Dordrecht: Reidel.
Lamb, D. and S. M. Easton. 1984. Multiple discovery: The pattern of scientific progress. Avebury Publishing
Company.
Laszlo, E. 1972. A general system model of the evolution of science. Scientia 107: 379-95.
Latour, B. and S. Woolgar. 1979. Laboratory life. Beverly Hills, Calif.: Sage.
Laudan, L. 1977. Progress and its problems. Berkeley: University of California Press.
Laudan, L. 1978. In Nickels, 173-83.
Laudan, L. 1986. Some problems facing intuitionist meta-methodologies. Synthese 67: 115-29.
Levinson, P. 1988. Mind at large. Greenwich, Conn.: JAI Press.
Levi-Strauss, C. 1962. La pensee sauvage. Paris: Plon.
< previous page
page_265
next page >
page_266
< previous page
page_266
next page >

Page 266
Lewis, W. S. ed. 1960. Horace Walpole's correspondence, vol. 20-IV: 407. New Haven, Conn.: Yale University
Press.
Lorenz, K. 1941. Kant's doctrine of the a priori in the light of contemporary biology. In H. C. Plotkin, ed. 1982,
Learning, development and culture: Essays in evolutionary epistemology. New York: John Wiley.
Lukacs, G. 1975. The young Hegel. R. Livingstone, trans. London: Merlin.
Mayr, E. 1959. Isolation as an evolutionary factor. Proc. Am. Phil. Soc. 103: 221-30. Reprinted in Mayr 1976.
Mayr, E. 1964. Proc. Fed. Am. Soc. Exp. Biology 23: 1231.
Mayr, E. 1976. Evolution and the discovery of life. Cambridge: Harvard University Press.
McMullin, E. 1976. The fertility of theory and the unit of appraisal in science. In Essays in the Memory of Imre
Lakatos, R. S. Cohen et al., eds. 395-432. Reidel: Dordrecht.
Merton, R. K. 1973. The sociology of science. Chicago: University of Chicago Press.
Meyerson, E. 1908. Identit et realit. Paris: Alcan.
Mulkay, M. J. 1972. The social process of innovation. London: Macmillan.
Munz, P. 1989. Watkins's evolutionism between Hume and Kant. In Freedom and rationality: Essays in honour of
John Watkins, Fred D'Agostino and I. C. Jarvie, eds. Dordrecht: Kluwer Academic Publishers.
Musgrave, A. 1988. Is there a logic of scientific discovery? LSE Quarterly 2-3: 205-27.
Nagel, T. 1986. The view from nowhere. New York: Oxford University Press.
Nambu, Y. and G. Jona-Lasinio. 1961. A dynamical model of elementary particles based upon an analogy with
superconductivity. Physical Review I: 122: 345-58, and II: 124: 246-54.
Ne'eman, Y. 1961. Derivation of the strong interactions from a gauge invariance. Nuclear Physics 26: 222-29.
Ne'eman, Y. 1980. Science as evolution and transcendence. Acta Scientifica Venezolana (Ensayo) 31: 1-3.
Ne'eman, Y. 1983. Patterns, structure and then dynamics: Discovering symmetry and conceiving quarks.
Proceedings of the Israel Academy of Sciences and Humanities, Section of Science, no. 21: 1-26.
Ne'eman, Y. 1988. The evolutionary role of research. Metabolic, Pediatric and Systemic Ophthalmology 11: 12-13.
< previous page
page_266
next page >
page_267
< previous page
page_267
next page >

Page 267
Nersessian, N. J., ed. 1987. The process of science. Dordrecht: Martinus Nijhoff.
Nersessian, N. J., ed. 1989. Conceptual change in science and in science education. Synthese, 80, 163-83.
Newton, I. 1962. Mathematical principles of natural philosophy, trans. by A. Mott, revised by F. Cajori. Berkeley:
University of California Press.
Nickles, T. ed. 1978. Scientific discovery, logic and rationality. Dordrecht: Reidel.
Nickles, T. ed. 1980. Scientific discovery: Case studies. Dordrecht: Reidel.
Nickles, T. ed. 1987. 'Twixt method and madness. In Nersessian, 41-67.
Nishijima, K. 1954. Some remarks on the even-odd rule. Progress of Theoretical Physics 12: 107-8.
Okubo, S. 1962. Note on unitary symmetry in strong interactions. Progress of Theoretical Physics 27: 949-66.
Osborn, A. 1963. Applied imagination. New York: Charles Scribner's Sons.
Piaget, J. 1977. The role of action in the development of thinking. In Knowledge and development, vol. 1, W. F.
Overton and J. M. Gallagher, eds. 17-42. New York: Plenum Press.
Piaget, J. and B. Inhelder. 1971. Mental imagery of the child. New York: Basic Books.
Pickering, A. 1984. Constructing quarks. Edinburgh: Edinburgh University Press.
Poincar, H. 1914. Science and method. London: T. Nelson.
Polanyi, M. 1958. Personal knowledge. London: Routledge and Kegan Paul.
Polya, G. 1965. Mathematical discovery, vol. II. New York: John Wiley.
Popper, K. 1959. The logic of scientific discovery. Sixth Impression (1972), London: Hutchinson.
Popper, K. 1969. Truth rationality and the growth of scientific knowledge. In Conjectures and refutations 215-50.
London: Routledge and Kegan Paul.
Popper, K. 1972. Objective Knowledge. Oxford: Clarendon Press.
Pstruzina, K. 1992. New ideas. Psychology of Science Group (e-mail network) PSY-SCI@MTUS5. bitnet. 2 Apr.
Putnam, H. 1973. Mathematics, matter and method: Philosophical papers 1. Cambridge: Cambridge University
Press.
Quine, W. V. 1969. Epistemology naturalized. In Ontological relativity and other essays 69-90. New York:
Columbia University Press.
< previous page
page_267
next page >
page_268
< previous page
page_268
next page >

Page 268
Rawls, J. 1971. A theory of justice. Cambridge: Harvard University Press.

Reichenbach, H. 1938. Experiment and prediction. Chicago: University of Chicago Press.
Rescher, N. 1986. Scientific realism. Dordrecht: Riedel.
Roberts, R. M. 1989. Serendipity: Accidental discoveries in science. New York: John Wiley.
Ruse, M. 1989. The view from somewhere. In Hahlweg and C. Hooker, 185-228.
Sakata, S. 1956. On a composite model for the new particles. Progress of Theoretical Physics 16: 686-88.
Scherk, J. and J. H. Schwarz. 1974. Dual models for non-hadrons. Nuclear Physics B81: 118-44.
Schwarz, J. H. ed. 1985. Superstrings. Singapore: World Scientific.
Shapiro, G. 1986. A skeleton in the darkroom--Stories of serendipity in science. San Francisco: Harper and Row.
Siegel, H. 1980. Justification, discovery and naturalizing of epistemology. Philosophy of Science 47: 297-321.
Simon, H. A. 1987. Is a scientific discovery a topic in the philosophy of science? In Scientific Inquiry in
Philosophical Perspective, N. Rescher, ed. 1-15. Lanham, MD: University Press of America.
Simonton, D. K. 1988. Scientific genius: A psychology of science. Cambridge: Cambridge University Press.
Stein, E. 1990. Getting closer to the truth: realism and the metaphysical and epistemological ramifications of
evolutionary Epistemology. In Evolution, cognition and realism, N. Rescher, ed. Lanham, MD: University Press of
America.
Stich, S. P. and R. E Nisbett. 1980. Justification and the psychology of human reasoning. Philosophy of Science 47:
188-202.
Storer, N. W. 1966. The social system of science. New York: Holt, Rinehart and Winston.
Suppe, F. ed. 1974. The structure of scientific theories. Urbana: University of Illinois Press.
Taton, R. 1957. Reason and chance in scientific discovery. London: Hutchinson.
Thagard, P. 1980. Against evolutionary epistemology. PSA 1980, vol. I: 187-96. East Lansing, Mich.: Philosophy
of Science Association.
< previous page
page_268
next page >
page_269
< previous page
page_269
next page >

Page 269
't Hooft, G. 1971. Renormalizable Lagrangians for massive Yang-Mills fields. Nuclear Physics B 35: 167-88.
Toulmin, S. 1972. Human understanding. Princeton: Princeton University Press.
Toulmin, S. and J. Goodfield. 1962. The architecture of matter. Harmondsworth, Middlesex: Penguin Books.
Vaux, J. 1990. Replicating the expert. New Scientist no. 1706: 55-58.
Waddington, C. H. 1975. The evolution of an evolutionist. Edinburgh: Edinburgh University Press.
Washburn, S. L. and Jane B Lancaster. 1968. Cultural evolution. In International Encyclopedia of the Social
Sciences, vol. 5. New York: Macmillan and The Free Press.
Wheelwright, P. ed. 1966. The Presocratics. New York: Odyssey Press.
Whittaker, E. T. 1951. A history of theories of aether and electricity. London: T. Nelson.
Wigner, E. 1937. On the consequence of the symmetry of the nuclear Hamiltonian on the spectroscopy of nuclei.
Physical Review 51: 106-19.
Winograd, T. A. and F. Flores. 1987. Understanding computers and cognition: A new foundation for design.
Reading Mass.: Addison-Wesley.
Winson, J. 1990. The meaning of dreams. Scientific American vol. 263 no. 5: 86.
Wittgenstein, L. 1968. Philosophical investigation. Oxford: Blackwell.
Yang, C. M. and R. Mills. 1954. Isotopic spin conservation and a generalized gauge invariance. Physical Review
95: 631.
Yoneya, T. 1974. Connection of dual models to electrodynamics and gravidynamics. Progress of Theoretical
Physics 51: 1907-20.
Zahar, E. 1983. Logic of discovery or psychology of invention? British Journal for the Philosophy of Science 34:
243-61.
Zandvoort, H. 1988. In Donovan et al. 337-58.
Ziman, J. 1968. Public knowledge. Cambridge: Cambridge University Press.
Zweig, G. 1964. An SU3 model for strong interaction symmetry and its breaking. CERN preprints 8182/TH401 and
8419/TH412.
< previous page
page_269
next page >
page_271
< previous page
page_271
next page >

Page 271
Index
A
Abduction, 66
Adams, J. C., 94
Adaptability, 200, 206, 216
vs. adaptation, 207-8
and evolutionary progress, 207
and rationality, 193
Adaptiveness, 204
Ampere, A. M., 97, 105
Ampliative inference, 4, 63, 65, 68, 70, 74, 135-6, 152
Amundsen, R., 150
Analogy, reasoning by, 73, 85-6
and tinkering, 227, 230-3
Analytical type, 179
Anderson, P. A., 237
Andrade, E. N. da C., 260
Archimedes, 35, 113, 180
Aristotle, 13, 137, 156, 201
Artificial intelligence (AI), 31, 82, 209
and heuristic search, 30
Association(s), 174, 178-83
chance, 185
infraconscious, 178-9
network of, 182-3
Association-strengths (between mental elements), 178
Avogadro, C., 80
Ayala, F., 150
Ayer, A., 200
page_271
B
BACON, 78-80
Bacon, F., 3, 97
Bajema, C. J., 250
Barnes, V. E., 105, 233
Bayesian theory, 54, 88-91, 104, 125, 132, 177
dynamic assumption of, 89-90
Becquerel, H., 81, 167
Belief, degree of, 90
Bell, J., 238
Berzelius, J. J., 192
Black, M., 84
Blackbody radiation, 29, 77, 163-4
Black-boxing, 93-5, 246, 254
Blind conjectures, 146
Blind variation, 3, 7, 42, 125, 146-57, 174, 184, 186, 218, 223, 243-4
social dimension of, 197-9
Bohm, D., 92, 194
Bohr, N., 194, 258
Bohr-Rutherford model, 26, 32, 83-5, 159
Boltzmann, L., 77, 163-4, 258
Born, M., 194
Boyle, R., 24, 39, 55, 124
Bradshaw, G., 78
Brahe, T., 161
Brainstorming, 257
Bronowski, J., 216
Brown, R., 168
Buchanan, B., 82
Buckley, W., 45
C
Campbell, D. T., 3, 145, 148, 150-1, 174-5, 259
page_271
Cannon, W., 168

Carnap, R., 90, 122, 135, 259
Chance configuration model, 174-6, 178, 185
< previous page
page_271
next page >
page_272
< previous page
page_272
next page >

Page 272
Chance permutation, 174-5, 177-9, 181-6, 194, 197, 217

Chew, G., 192
Chisholm, R., 192
Chomski, N., 201
Chubin, D. E., 241
Coevolution of sensorimotor organs and brain, principle of, 8, 206, 209-12
strategies of discovery derived from the, 221-2
Cognitive apparatus, 18, 21, 94, 145, 147-50, 199-203, 205-6, 208,
210-1, 214, 225, 250, 253
Cohen, I. B., 260
Cohen, L. J., 134, 137, 140
Coherentism, 133
Collective brain, 211-2, 253
Collective mind, 115, 182, 217
preparing the, 219-20
Collins, H. M., 19
Columbus, C., 15, 39, 163
Communication channels with reality, 48
Communication network of scientific community, 191
Community-dependent logics, 75
Competence, collective, 142
Competence-performance distinction, 139-41
Confirmation, 22, 102, 139
degree of, 89-91, 104, 123, 159
dynamic view of, 106
as exposure, 33
of a hypothesis/theory, 12, 22, 66-7, 98, 100, 104, 117, 123, 187
as part of the discovery process, 19, 33, 67-8, 101
Consensus in science, 119, 153, 211
Consequentialism, 97-9
page_272
Conservation laws, 43, 47, 80-1, 231-2

Constructivist epistemology, 45
Context of discovery vs. context of justification, 5, 53, 56, 97-101, 117, 135, 199, 218-9, 226
and description-prescription dichotomy, 99-100
Context of generation, 102, 106-10
Convergent thinking, 179
Cooperation in science, 160, 189-97, 210-1, 240, 254, 257
diachronic, 191, 215
epistemological significance of, 212-5
synchronic, 190-1, 215
Copernicus, N., 59, 162, 165, 213
Coulomb, C., 22, 24
Creation, processes of, 173-88
Creativity
artistic vs. scientific, 186-8
in discovering natural kinds, 70-1
in discovering a theory, 27
and heuristic-guided generation, 152
and non-conformism, 185, 196-7
and opportunism, 223-6, 228-9
vs. reasoning, 61-2
and serendipity, 148
social dimension of, 182, 197, 256
and tinkering, 223-6, 229
and unintentionality, 113
Crick, F., 228
Crookes, W., 15, 72
Cultivation
of the collective mind, 219-20
of creativity, 62, 184-5
of discovery, 3, 5-6, 114-5, 251
of serendipity, 168-71, 221
page_272
of the unintentional, 113-15

Curie, M. and P., 81, 167
Cushing, J., 134, 234
Cuvier, G., 190
Cziko, G. A., 259
D
D'Alembert, J., 156
Dalitz, R., 235
Dalton, J., 80, 101
Darwin, C., 23, 36, 178, 181-2, 190
De Broglie, L., 194
Delair, J. B., 260
Delbruck, M., 228
Democritus, 101, 226
Descartes, R., 59, 63, 65, 69, 190, 194
Descartes' rules of discovery, 50-1
Desmond, A. J., 260
< previous page
page_272
next page >
page_273
< previous page
page_273
next page >

Page 273
Dewey, J., 45
Differential dissemination, 250
Differential reproduction, 249
Dirac, P. A. M., 158, 194
Discoverability, 98
Discovery
blind variation view of, 7, 148
cooperative-historical process of, 5, 52, 78, 113, 142, 183, 189, 197, 203, 219, 233, 256, 258
efficiency of the process of, 55-6
epistemological aspects of, 12-6, 18
as an evolutionary phenomenon, 6-8, 29
by exposure, 4, 29-32, 38, 40, 43-4, 47-9, 64-5, 71, 74, 79-80, 86, 94-5, 101-2, 113-4, 178, 244, 254, 256
by generation, 4, 29, 32-4, 35, 38, 39-43, 47-9, 64, 71, 74, 79-80, 86, 94-5, 101, 113, 114, 122, 178, 254-6
and growth of knowledge, 57
by inference, 4, 30-3, 49, 61-8, 97, 98, 246
vs. invention, 4, 36-9
involuntary process of, 5-6, 35, 102, 110, 113-5, 118, 122, 141-3, 178, 191, 194, 258
mathematical, 34, 55, 62-3
by mathematical calculation, 30
mechanized, 2, 4, 77-80, 115, 142-3, 258
multiple, 185-8
naturalized, 2, 115, 142
object of, 11, 32
by observation, 4, 29
ontological aspects of, 16, 18
and philosophy of science, 1-2
of a problem, 28
process of, 11, 16
product of, 11, 16-29, 32-3
scientific, 13
page_273
by searching, 30, 86
serendipitous, 3, 7, 42, 155, 157, 180-2, 225, 243
as a skill, 5, 92-5, 113, 258
of a theory, as a dynamic process, 26, 32, 80-5
unintentional process of, 5-6, 35, 49, 50, 52, 100, 102, 113-5, 191, 257-8
Discovery-generating argument, 63
Disembodied organisms, 251
Dissemination, the social dimension of, 199
Divergent thinking, 179
D-J distinction, see context of discovery vs. cotext of justification
D-J-D distinction, 199
Dobzhanski, T., 150
Donovan, A., 105
Dormant genes, 165, 204
Dormant ideas, 249
Dubos, R. J., 228
Duhem-Quine thesis, 109, 198
Dynamicism, 106
E
Easton, S. M., 181-2, 190
Edwards, W., 90
Einstein, A., 7, 12, 148, 159, 165-6, 168, 173-4, 177, 189, 192, 196, 258
Electroweak unification, 12, 190, 237-40
Embodied theories, 251
Empedocles, 226
Empirical generalizations, 25, 28, 152
discovery of, 16, 21, 54, 66
vs. laws of nature, 21, 22
Empiricism, 218-20, 246
vs. rationalism, 218
Epicurus, 177
Epistemic cooperation, 3, 184, 189-97, 219, 257
page_273
Epistemic profit, 69, 106, 157-8

Epistemologizing biology, 147
Epistemology
constructivist, 45, 47
evolutionary, see evolutionary epistemology
social, 7, 184, 189, 192, 212
naturalized, 133, 147
transactionalist, 45, 47
< previous page
page_273
next page >
page_274
< previous page
page_274
next page >

Page 274
Euler, L., 156

Eureka event, 5, 35-6, 101, 113, 177, 182, 238
Evidential support, 90
Evolutionary epistemology (EE), 2, 7, 77, 145-8, 199-200, 210, 243-4, 252, 259-60
Evolutionary line, 260
Evolutionary POR, 178, 193, 199, 215, 218-9
Evolutionism, 125, 142
Expectedness, degree of, 89
Experimentation, as generational discovery, 33
Expert systems, 4, 56-7, 92
Explanation
bootstraps kind of, 24
deductive-nomological, 28
discovery of, 49
non-circularity, 24
theoretical, 27-8, 88, 124, 130, 134, 138-9, 212-3
Explication
of basic ideas in a model, 245
of intuitions, 84, 88-90, 103-4, 122-3, 135, 137, 179, 259
Extended sensorimotor organs and brain, 202, 209-12, 225, 253
F
Falsificationism, 14, 83, 125, 248
Faraday, M., 166
Fermat, P., 190
Fermi, E., 234
Feyerabend, P., 109, 156, 213
Feynman, R., 234, 238
Fine, A., 253
Finocchiaro, M., 105
Fleming, A., 15, 113, 154-5, 168-9
page_274
Flores, F., 92
Focal vs. subsidiary awareness, 17, 94, 179
Fodor, J., 140
Freedom of research, 170-1
Fresnel, A., 173
Fuchsian functions, discovery of, 177
G
Galileo, G., 29, 41, 59, 105, 162, 187, 190
Galvani, L., 168
Gauss, K., 178
Geiger, H., 72-3, 222, 227
Gell-Mann, M., 192, 232-3, 235-6, 238
Gene pool, 150, 156, 216, 240, 248, 250
of human species, 204-5
vs. idea pool, 209, 247
Generationism, 98-101
Genetic epistemology, 257
Genius, 52, 174-5, 178, 184
Genotype, 150
Genotype-phenotype (GP) model, 8, 243-52
Giere, R., 45
Glashow, S. L., 236-7
Goals of science, 53, 118-22, 125, 143, 254
Goeranzorn, B., 93
Goldberg, H., 235
Goldstone, J., 237
Goodfield, J., 226
Gooding, D., 94, 260
Goodman, N., 120-3, 126, 135
Goudsmit, S., 258
Gradualism in the growth of knowledge, 156, 171, 180, 217, 225
Grosseteste, R., 58, 71
page_274
Growth by expansion, principle of, 7, 122, 206-9

strategies of discovery derived from the, 220-1
Gursey, F., 233
H
Hacking, I., 89
Hahlweg, K., 145
Hamilton, W. R., 156
Hanson, N. R., 4, 66, 85-8, 91
Hegel, G., 189, 240
Heidegger, M., 92-3
Heisenberg, W., 194, 231, 233
Helmholtz, H. von, 178, 185
Hempel, C., 28f
Hermite, C., 51
Herschel, J., 97-100
Hertz, H., 166
< previous page
page_274
next page >
page_275
< previous page
page_275
next page >

Page 275
Hesse, M., 68, 84, 122-5, 258

Heuristic rules, 75, 149-50, 152-4, 246
domain specific, 81
recursive, 79
Heuristic search, 30, 57
Hilbert, D., 43
History of science, 127
Hofmann, J. R., 105
Hooke, R., 190
Hooker, C. A., 145
Hull, D. L., 210, 215, 247, 260
Hurd, D. L., 167
Huygens, C., 190
Hypothesis-generating argument, 86
Hypothetico-deductive (HD) method, 66-7, 98, 127-8
I
Idea pool, 209, 239, 248-9
Incommensurability, 75, 217, 248
Incubation, process of, 3, 5, 7, 35, 113, 142, 169, 174, 178, 180-2, 194, 217
Induction
elliminative, 30
enumerative, 65-6, 69, 81
justification of, 200
rules of, 149-50
Inference
ampliative, 4, 63, 65, 68, 70, 74, 135-6, 152
deductive, 4, 30-1, 68-74, 81, 86, 102, 136, 151-2, 154, 161-2, 254
inductive, 4, 21, 31-2, 68-9, 86, 121, 154
license, 152
material rules of, 32, 74, 76, 93-5, 246, 254
page_275
scientific, 152-3
Inferential practice, 120-1
Inhelder, B., 46
Innovation-tradition tension, 184-5, 197
Institutions of the scientific community, 142
Instrumentalism, 24, 25, 45
Intellectual migration, 185, 196, 225, 227-9, 241
Internal symmetries of hadrons, 42-3, 230-3
Interpsychic processes, 3, 174, 176, 217, 253
Intervention
cognitive, 19, 21, 26, 45
observational/experimental, 17-20, 40-1, 45, 121, 201, 221
technological, 207
Intrapsychic processes of creation, 3, 113, 173-88, 191, 197, 217, 253
Intuition, 61, 135-9, 142
Intuitionism, metamethodological, 134-5, 137-8
Intuitive type, 178
Invention vs. discovery, 4, 36-9
Inventive arguments, 4, 73, 95
Invisibility
of cognitive apparatus, 17
of premises, 5
of presuppositions, 93-4
Invisible college, 190
Involuntary processes of creation, 3, 5-7, 178, 191, 244, 254
Is-ought fallacy, 134, 138-9
J
Jacob, F., 3, 223-4, 229, 241-2
Jeans, J., 163
Jona-Lasinio, G., 237
Justification
of belief, 13
page_275
consequential, 98, 100

generational, 98, 100
of methodological rules, 120-2, 124-6, 219
of scientific practice, 124-6, 131
Justificatory vs. generational parts of discovery, 62-3, 97-8
K
Kant, I., 44, 149, 251
Kantorovich, A., 3, 39, 82, 148, 210, 243
Kekule, A., 35, 185
Kelvin, Lord, 213-4
Kepler, J., 15, 28, 59, 64, 86, 95, 160-5, 169, 190, 203
< previous page
page_275
next page >
page_276
< previous page
page_276
next page >

Page 276
Kepler's laws, 12, 14, 23, 27, 66, 78, 124, 160-3
Kepler machine, 79
Keynes, J. M., 90
Kipling, J. J., 167
Kirchoff, G. R., 163
Klein, M., 164
Knorr-Cetina, K., 47, 225, 241
Knowledge
endosomatic and exosomatic, 145-6
as justified true belief, 13-4, 145
Koertge, N., 76
Koestler, A., 153, 160-4, 260
Kohn, A., 15, 168
Krohn, W., 92
Kuhn, T., 52, 60, 64, 75, 124, 185, 197, 211, 255, 257
L
Lagrange, J. L., 156
Lakatos, I., 14, 32, 82-3, 109, 127, 132, 159, 244, 246
Lamb, D., 181-2, 190
Lancaster, J. B., 202
Langley, P., 78
Laszlo, E., 216-7
Latour, B., 44
Laudan, L., 98, 105, 134-5, 137
Laudan, R., 105
Lavoasier, A., 12
Laws of nature, 43, 59, 119, 152
vs. empirical generalizations, 21, 22
falsity of, 54
ontological status of, 22
page_276
vs. theories, 16, 22-6

Leibniz, G. W., 51
Lenard, P., 15-6, 72
Leverrier, U., 85-6, 94, 227
Levinson, P., 210
Levi-Strauss, C., 3, 223, 242
Lindman, H., 90
Linneus, C., 159
Logic
community-specific, 4, 95, 74-6, 121, 136
material (content-dependent), 7, 93, 74-6
of ontogeny, 246
of pursuit, 5, 85-92
Logical empiricism, 1, 5, 52, 91, 97, 100-1, 117
Logicism, 6, 118, 125, 131-2, 139, 212
Lorentz, H. A., 158, 189
Lornez, K., 145
Lukacs, G., 190
M
Malthus, T. R., 36, 181-2, 190
Mann, H., 154
Mantell, G. A., 190
Marginality
intellectual, 240
professional, 239, 241
social, 185
Marsden, E., 72-3, 222, 226, 241
Material logics, 7, 74-6, 93
Maxwell, J. C., 23-4, 64, 151, 157-9, 189
Mayr, E., 224, 240
McMullin, E., 83
Mead, G. H., 45
page_276
Mechanized discovery, 2, 4, 77-80, 115, 142-3, 258

data-driven, 79, 82
and the social dimension of discovery, 77
Mendel, G., 182, 204
Mendeleev, D., 39, 159
Mental elements
affinities, 175, 177, 180-1
combinations of, 28, 174
configurations of, 28, 175-6
network of, 182
repertoire of, 185
Mercier, Cardinal, 192
Merton, R. K., 191
Mertonian norms, 191
Mesocosmos, 203-5, 208, 210, 214
Metamethodological theory, 135
Method of discovery
algorithmic, 51, 53, 114, 155, 201, 212
content/domain-specific, 4, 56-7, 70
Descartes', 50-1
functions of, 54-7
generality of, 56
< previous page
page_276
next page >
page_277
< previous page
page_277
next page >

Page 277
origin of, 58-60

postmortem, 4-5, 50
Methodological rules, 120-2, 124-7, 130-1, 139, 157
Methodological theory, 124-6, 130-1, 135
Meyerson, E., 80-1, 168
Mill, J. S., 3
Mills, R. L., 236
Model
behind a research program, 32, 83-5
negative analogy of, 43, 84-5, 250, 252
neutral analogy of, 83-5, 245, 252
positive analogy of, 43, 84-5, 243, 252
Mulkay, M. J., 185-8, 192, 196-7, 228
Multiple discovery, 185
Munz, P., 251
Musgrave, A., 4, 70-1, 73-4, 86
Mutation, blind/quasi-random, see blind variation
N
Nagel, T., 205
Nambu, Y., 237
Nativism, 246
Natural kinds, 17, 21, 32, 70-1, 75, 136, 149, 187
Natural selection, 145-51, 153, 200, 216, 223, 237, 247
mechanized, 77
models of creation and discovery, 2-3, 7, 29, 173-4, 241, 253
models of sociocultural evolution, 42
paradigm of rationality, 147
Naturalism, 6, 142
normative, 134-43
Nature vs. nurture dichotomy, 114, 246
page_277
Necessary truths, 63
Ne'eman, Y., 3, 7, 81, 148, 154, 232, 235, 238-9
Nersessian, N., 258
Newton, I., 13-4, 23-4, 27, 31, 33, 35, 64, 66, 79, 94-5, 124, 128-9, 137, 143, 152, 156, 159, 162-3, 173, 176, 187,
190-1, 201, 203, 214
Newton's Rules of Reasoning in Philosophy, 50-1, 59
Nickles, T., 76, 94, 98, 260
Nicod's Rule, 104
Nisbett, R., 121, 136
Nishijima, K., 232
Normal science, 52, 60, 64, 83, 150-1, 153, 165, 186-7, 197, 215, 217, 243, 248, 252, 255
Normative-descriptive (ND) dichotomy,
and method of discovery, 57
and philosophy of science, 117-34
Novelty-generating argument, 65
O
Ockham's Razor, 59
Oersted, H., 168
Omega-minus, discovery of, 20, 233
Okubo, S., 233
Ontogeny, 8, 243-6, 250-2, 257
Opportunism
in context, 230
in natural selection, 228
in science, 229, 253
Osborn, A., 257
Osiander, A., 165
P
Pais, A., 233
Paradigm of rationality (POR), 6, 124-7, 130-3, 139, 142, 147, 157-8
Paradigm, scientific, 83, 130-1, 153-4, 187-8, 196-7, 203, 211, 217, 248, 250, 256
Particle physics, 12, 19-20
page_277
authorities in, 192

experimenting in, 39-43
observation in, 33, 39-44
strategies of problem-solving in, 221
tinkering in, 8, 229-42
Particularism, 1, 6
Pasteur, L., 6, 113, 181, 228
Pauli, W., 194, 231
Peirce, C. S., 4, 66-8, 87-8
Penicillin, discovery of, 15, 113, 154, 165, 168
Performance errors, 140-1
< previous page
page_277
next page >
page_278
< previous page
page_278
next page >

Page 278
Periodic Table, 39
Philosophy of science
descriptive, 127-30
and discovery, 1-2
explanatory, 6, 110, 117, 123-34
naturalistic, 5-6, 109, 133, 145-6, 212, 255
normative, 126-7, 130, 132-3
Phylogeny, 145, 151, 216, 243-4, 248, 257
Piaget, J., 45-7, 201-2, 251, 257
Pickering, A., 45, 230, 232, 236, 238-9,260
Pinch, T., 92
Planck, M., 36, 72, 78, 163-5, 168, 171, 183, 194, 197, 203, 258
Plato, 226
Platonic solids, 160-1, 169, 226
Plausibility of hypothesis, degree of, 90-1, 104
Poincare, H., 34-5, 36, 61, 68, 86, 88, 99, 102, 110, 114, 177-8, 181, 189
Polanyi, M., 17, 94
Polya, G., 169, 180, 221, 227
Popper, K., 13, 14, 23, 54, 70, 74, 82, 98, 100, 108, 125, 128, 145-6, 148, 198, 209, 227, 244, 248, 255
Preadaptation, 150
Predictability, 88, 91, 104-6, 157, 178
explained, 219
Prepared mind, 113-5, 142, 181-2, 254
and recommendations for cultivating serendipity, 157, 168-71, 221
Probability of a hypothesis
conditional, 89-91
posterior, 89-90
prior, 54, 89-91, 104, 177
subjective, 89-90
Problem-oriented scientists, 228
page_278
Problem solving, 28-30, 151-2, 157, 169-70, 257

by expansion, 207-8, 220-1
in normal science, 150, 153
in research programs, 155, 244
strategies of, 221
Projectible predicates, 69
Proto-theoretical entities, 24, 26
Prout, W., 221-2, 256
Pstruzina, K., 173
Ptolemy, 80, 158, 162, 203, 213
Putnam, H., 102-3
Pythagoras, 65
Pythagoreanism, 59, 160-2, 226
Q
Quantum gravity, theory of, discovery by serendipity, 155, 169
Quantum mechanics, discovery by serendipity, 163
Quine, W. V. O., 133
R
Radicati, L., 233
Radioactivity, discovery of, 165-8
Rationalism, 98
vs. empiricism, 218
Rationality
categorical, 53, 76, 108, 118
as cautiousness, 69
community-specific, 121
domain-specific, 76
ideal theory of, 139-41
instrumental, 53, 118
intuitionistic theory of, 134
naturalistic, 119
normative theory of, 140
page_278
rules of, 6
scientific, 133, 137-8
shallow vs. deep theories of, 6, 134-43
theory of, 120, 122, 134-43
therapist model of, 141
Rawls, J., 121
Rayleigh, J. W., 163
Realism
and collective discovery, 195
constructive, 44-5, 47
convergent, 44, 46
and creative discovery, 36-48
epistemological, 44
and laws of nature, 23
naive, 44
metaphysical, 44
with respect to theories, 45
Reality, shallow vs. deep levels of, 47-8
< previous page
page_278
next page >
page_279
< previous page
page_279
next page >

Page 279
Recombinations, 156
Recommendations for cultivating discovery, 114-5
Reconstruction of the discovery process, 64, 73, 98, 100, 179,
see also Method, postmortem
Recursive procedure, 77-80, 82
Reflective equilibrium, 121-3, 126, 136-7
Refutation of a theory, 12, 100, 109, 117, 198, 244
as a discovery, 19
Regge, T., 234
Regge poles theory, 229, 234
Reichenbach, H., 99, 103, 135
REM sleep, 173
Representation, 45-7
Rescher, N., 260
Research program, 7, 8, 25, 32, 80-5, 130, 152, 154-5, 159, 244-9, 254
degenerative, 83
hard core of, 82-3, 244, 246, 248-9
positive heuristic of, 82-3, 246
progressive, 14, 159
protective belt of, 83
Restivo, S., 241
Retroduction, 4, 66-8
Revolutionary science, 52, 153, 187, 217
Richter, B., 236
Roberts, R. M., 168
Roentgen, W. C., 15, 166-7
Rorty, R., 195
Ruse, M., 204-5
Rutherford, E., 32, 72-4, 83, 86, 95, 159-60, 192, 222, 226-7, 241, 258, 260
S
page_279
Sakata, S., 235

Sakurai, J. J., 236, 238
Salam, A., 237
Sargeant, W. A. S., 260
Savage, L. J., 90
Schaffer, S., 260
Scherk, J., 155
Schroedinger, E., 56, 194
Schwarz, J. H., 155
Schwinger, J., 234, 236
Science education, 8, 255-8
Science of science, 117, 130
Science policy, 8, 170-1
Scientific method, 52-3, 127-8, 137-8
Scientific practice, 124-6, 130-4, 136-9
Selection, 139
vs. generation, 34-5, 68, 86-8, 101
process of, 7
social dimension of, 198-9
Selection cycle in science, 250
Self organization, goal of, 176, 183-4
Sensorimotor organs, extension of, 202, 209, 212, 214, 251
Serendipity, 3, 7, 27, 148-71, 174, 179, 198, 203, 216, 221, 223, 225, 229, 232, 244, 249, 253, 257-8
and adaptability, 208
and cooperation, 154, 160
cultivation of, 168-71, 220
explanation of, 180-2
and freedom of research, 170-1
the principle of, 8, 42, 148-57, 169-70, 217-9, 255
Shapiro, G., 168
Siegel, H., 103
Simon, H., 4, 78, 95
page_279
Simonton, D. K., 3, 7, 36, 169, 173-8, 180-6

Skill of discovery, 5, 92-5, 113, 258
Skill-laden discovery, 94
Social acceptance, 193, 195
Social dimension of science, 7, 36, 49, 115, 119, 133, 160, 189-97, 203, 208, 258
Social epistemology, 7, 184, 189, 192, 212
Social recognition, 183-4, 196
Sociocultural evolution, 42, 147, 203-6, 209, 251
Socio-evolutionary POR, 160, 255
Socio-evolutionary processes of discovery, 219
Sociologism, 6, 125, 142, 219
Speciation, 240
Stahl, G. E., 12
Stein, E., 44
Sommerfeld, A., 258
< previous page
page_279
next page >
page_280
< previous page
page_280
next page >

Page 280
Stich, S., 121, 136

Storer, N., 36
Stratified stability, 156, 171, 216
Superstring theory and serendipity, 155, 169
Suppe, F., 106
Suppressed premises, 5, 64, 94
Szilard, L., 228
T
Tacit Knowledge, 76, 92-3, 136, 220
Taton, R., 34
Thagard, P., 148
Theaetetus, 226
Theory, (scientific), 20-7, 129, 252
dynamic, 25, 27, 82-4, 106, 244-6, 248, 251
vs. law of nature, 16, 22-6
mature (version of), 26-7
ontological claims of, 24-5
plasticity of, 20-7
as a statement, 23-4, 26-7
unified, 229-30
-versions, 32, 83-5
Theory-construction
dynamic, 26, 32, 80-5, 102, 246, 250
as a generational discovery, 33, 148
heuristic-guided, 80, 152, 246
methods of, 150
Theory of science, 126, 139-40
explanatory, 7, 131-2, 138, 157, 253
evolutionary, 7, 133, 139, 147, 157, 173
naturalistic, 134, 247, 253
page_280
Thomson, J. J., 73, 166

't Hooft, G., 12, 190, 238-40
Ting, S., 236
Tinkering, 3, 7, 8, 87, 225-7, 229, 244, 253-5, 257-8
in the context of generation, 226
in current algebra, 229, 238-9
evolutionary, 223-4
with group-theoretical tools, 231-3
in particle physics, 229-42
in quantum chromodynamics (QCD), 230
with quantum electrodynamics (QED), 230, 233-4, 236-7
with quantum field theory, 229, 232-5
with quantum mechanics, 230
with the quark model, 229, 235
and reasoning by analogy, 227
and renormalization procedure, 229, 234, 237-9
in savage thought, 223, 242
and short sightedness, 227
in S-matrix theory, 234-5
in Yang-Mills gauge theory, 236-41
Tomonaga, S., 234
Tool-oriented scientists, 227-9, 235-41
Toulmin, S., 145, 226
Tradition-innovation tension, 184-5, 197
Transactionalist epistemology, 45
Transparency
of cognitive apparatus, 17
of information channels, 95, 254
of observational instruments, 94
of premises, 5
of presuppositions, 18, 95, 246
of the principle of induction, 32
page_280
of rules of inference, 247

and skill, 93-5
of theoretical tools, 94-5
Truth criterion, 192-3, 215
U
Uhlenbeck, G. E., 258
Ultra-violet catastrophe, 29, 163, 165
Uniformity of nature, principle of, 59, 69
Unintentional processes of creation, 3, 6, 27, 36, 113-5, 153, 178, 191, 197, 223, 254
V
Variability and adaptability, 216
of ideas, 217
Variation, 139
unjustified, 148
Vaux, J., 93
Veltman, M., 12, 238-41
Vico, 46
von Neumann, J., 92
< previous page
page_280
next page >
page_281
< previous page
page_281
Page 281
W
Waddington, C. H., 207
Walpole, H., 154
Ward, J. C., 237
Washburn, S. L., 202
Watson, J., 228
Weinberg, S., 237
Whittaker, E. T., 213
Wien, W., 163
Winograd, T., 92
Winson, J., 173
Wittgenstein, L., 93
Woolgar, S., 44
X
X-rays, discovery of, 15, 37, 38-9, 165-8
Y
Yang, C. N., 236
Yoneya, T., 155
Yukawa, H., 234
Z
Zahar, E., 64
Zandvoort, H., 105
Zeitgeist, 185-7
Ziman, J., 119, 211
Zweig, G., 233, 235
Zytkow, J., 78
< previous page
page_281

Aharon Kantorovich-Scientific Discovery PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aharon Kantorovich-Scientific Discovery PDF

Uploaded by

Copyright:

Available Formats

cover

next page >

Scientific Discovery : Logic and Tinkering SUNY Series in

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

SUNY Series in Philosophy and Biology

< previous page

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

Part I In Search for Logic of Discovery

1.1 What is a Discovery?

2.1 The Nature and Function of Method

< previous page

next page >

< previous page

next page >

Chapter 4: Philosophy of Science: From Justification to Explanation

4.1 Normative Philosophy of Science: Justification Relativized

< previous page

next page >

< previous page

next page >

5.1 Evolutionary Epistemology: Taking Natural Selection Seriously

5.4.2 Planck: The Reluctant Revolutionist

6.1 A Psychological Theory of the Creative Process

7.1 Epistemic Cooperation and the Social Dimension of Discovery

< previous page

next page >

< previous page

next page >

8.1 Evolutionary Tinkering in Science

9.1 Non-Creative Discovery: The Genotype-Phenotype Logic of Growth

Epilogue: Implications for Science Education

< previous page

next page >

< previous page

next page >

< previous page

next page >

< previous page

next page >

3.Naturalistic philosophy of science: a socio-evolutionary view, published in Journal of Social and

< previous page

next page >