. 1
( 7)


This page intentionally left blank
A Realist Philosophy of Social Science

This introduction to the philosophy of social science provides an orig-
inal conception of the task and nature of social inquiry. Peter Manicas
discusses the role of causality seen in the physical sciences and offers a
reassessment of the problem of explanation from a realist perspective.
He argues that the fundamental goal of theory in both the natural and
social sciences is not, contrary to widespread opinion, prediction and
control, or the explanation of events (including behavior). Instead, the-
ory aims to provide an understanding of the processes which, together,
produce the contingent outcomes of experience. Offering a host of con-
crete illustrations and examples of critical ideas and issues, this accessi-
ble book will be of interest to students of the philosophy of social science,
and social scientists from a range of disciplines.

°   .   ® ©    is Director of Interdisciplinary Studies at the
University of Hawai™i at M¯ noa.
A Realist Philosophy of
Social Science
Explanation and Understanding

Peter T. Manicas
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo

Cambridge University Press
The Edinburgh Building, Cambridge cb2 2ru, UK
Published in the United States of America by Cambridge University Press, New York
Information on this title: www.cambridge.org/9780521861403

© Peter T. Manicas 2006

This publication is in copyright. Subject to statutory exception and to the provision of
relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.

First published in print format 2006

isbn-13 978-0-511-22010-4 eBook (EBL)
isbn-10 0-511-22010-3 eBook (EBL)

isbn-13 978-0-521-86140-3 hardback
isbn-10 0-521-86140-3 hardback

isbn-13 978-0-521-67858-2 paperback
isbn-10 0-521-67858-7 paperback

Cambridge University Press has no responsibility for the persistence or accuracy of urls
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
To my wife

Acknowledgements page viii

Introduction 1
1 Explanation and understanding 7
2 Theory, experiment and the metaphysics of Laplace 26
3 Explanation and understanding in the social sciences 42
4 Agents and generative social mechanisms 75
5 Social science and history 103
6 Markets as social mechanisms 126

Appendix A The limits of multiple regression 151
Appendix B Comparison, Mill™s methods and narrative 157
Appendix C Rational choice theory and historical
sociology 171
Appendix D The neo-classical model 186

References 200
Index 217


The ideas in this volume have been germinating for some time. In that
time, I have accrued many debts. My education in the social sciences
began perhaps with my PhD mentor at Buffalo, Marvin Farber, who,
along with Bill Parry, started me on the path I pursue in this book.
Farber also let me write a dissertation in the Philosophy Department
which would not have been possible at most universities. My instruction
in history and the social sciences was advanced when I became Director of
the Program in Contemporary Civilization at Queens, a genuinely inter-
disciplinary program required of all students. While Farber and Parry had
convinced me that an understanding of social science was essential “ we
read Dewey™s Logic along with Schutz and standard works in the philos-
ophy of science “ Rom Harr´ and Paul Secord, authors of the important
(if for me, mis-titled) volume, The Explanation of Social Behavior (1973)
were critical in disabusing me of my lingering and unacknowledged log-
ical empiricism. Paul and I team-taught when we were both members of
the Queens College faculty, but I remember well my failure to see the con-
nection between the main messages of their book and what I had brought
from Buffalo. Some of these ideas were tried out at Queens College at
the Monday lunch group, a remarkable assembly of social scientists “
too many to list here “ who met regularly, lunched, heard papers and
had marvelous discussions. I also regularly team-taught with many of
this group in an undergraduate social science honors course “ a luxury of
inef¬ciency not much tolerated these days. The furious debates between
the members of that ¬‚oating group “ Tito Gerassi, Ray Franklin, Carl
Riskin, Mike Harrington, Bill Tabb, Paul Avrich, Mike Wreszin, Saul
Resnick, Mike Brown, Burton Zweibach and Lenny Markowitz to name
only a few “ were remarkable learning experiences, even if we sometimes
terri¬ed the undergraduates with our passion. The Journal for the Theory
of Social Behavior, initiated by Harr´ and Secord, but edited with style
and insight for the past twenty years by Charles Smith, not only provided
opportunity for me to test some of these ideas but was where I was intro-
duced to the work of Roy Bhaskar. His presence, along with that of Rom

Acknowledgements ix

Harr´ , was large in my History and Philosophy of the Social Sciences, my
¬rst major attempt to set out what was amiss in our understanding of the
social sciences.
I have been lucky also to have held two posts at the University of Hawai™i
at M¯ noa. After taking an early retirement at Queens, I became a member
of the Sociology Department and Director of Interdisciplinary Studies at
Hawai™i. My two long-standing colleagues in that wonderful unit, Jaishree
Odin and Emanuel Drechsel, both powerful interdisciplinary thinkers,
have been a constant source of knowledge and support. Prior to that I
had taught off and on in the Department of Political Science, where I
picked up some further debts, but especially to my good friend Man-
fred Henningsen. Each of these roles has been a delight. They certainly
enabled me to continue my interdisciplinary interests. My large lecture
section in sociology 100 was meant to enlarge the idea of introductory
social science. While everyone talks about cultivating “the sociological
imagination,” it is hard to see how one can do this with the awful standard
disciplinary textbooks and multiple choice exams. Teaching the sequence
of required courses in sociological theory to both graduates and under-
graduates forced me to re-think ideas that once seemed clear to me. I am
indebted also to my close colleague and friend in the Sociology Depart-
ment, Herb Barringer.
Most of my debts remain unacknowledged, but I must mention, at
least, those who have read and made suggestions on the present work. In
addition to some of those already mentioned, these include: Sam Pooley,
Gregory Maskarinec and Manfred Steger.

This volume reassesses the problem of explanation in social science from
what remains a marginalized, realist perspective. Because the problem of
explanation is central to inquiry in social science, the volume also pro-
vides a systematic philosophy of social science. It begins with the idea
that the fundamental goal of theory in both the natural and social sci-
ences is not, contrary to widespread opinion, prediction and control, or
the explanation of events (including “behavior”). Rather, more modestly,
theory (at least in one of its clear senses) aims to provide an understand-
ing of the processes which jointly produce the contingent outcomes of
experience. We understand why the planets move in ellipses, why mate-
rials burn, and why salt dissolves in water (if and when it does) when we
have a physical theory that provides a causal mechanism. By providing the
principles detailing the nature of molecules, the atomic structure of salt
and water, the principles of their action, and so on, we can understand
combustion and solubility “ and other chemical processes. Indeed, while
the theoretical work of physical scientists often begins with the effort to
understand patterns, they are not interested in, nor generally capable of,
providing either “explanations” or “predictions” of particular events. For
example, the trajectory of a boulder splintering as it rolls down a hill is
fully understood in terms of physical principles, but neither the trajec-
tory nor the ¬nal positions of the splintered parts can be explained or
predicted. But an adequate understanding of the outcome is easily avail-
able. The foregoing does not seem either surprising or novel. But, for
good historical reasons, reigning assumptions both in the philosophy of
social science and in much current social scienti¬c practice violate what
thus seems commonsensical.
It seems hardly deniable that understanding such natural processes
as splintering, oxidizing, dissolving, fertilizing and dying requires one
to understand the causal mechanisms at work “ physical, chemical and
biological, some available in direct experience, some not. No one will ever
see a photon but they are among the important non-observables posited
in a physical theory that enables us to understand a range of phenomena.

2 A Realist Philosophy of Social Science

The argument thus joins “realist” criticisms of empiricist conceptions of
theory and Humean notions of causality. Once this is in place, it is easy
to see why fairly long-standing objections to both the dominant view of
theory and the still dominant covering law model of explanation are fatal.
But by developing the ideas of agents as causes and of social mecha-
nisms as agent-generated causal mechanisms, the book extends, in a novel
way, the argument to the social sciences. Here we join old debates over so-
called “methodological individualism,” and the critical role of hermeneu-
tics, and recent debates in the philosophy of social science regarding
the ontology of society as provoked by Giddens, Bhaskar, Bourdieu and
others. Thus, the ontological status of “social structure” is clari¬ed and
resolved. Understanding in social science is achieved when, as in the
physical sciences, we have a causal mechanism, but unlike the physical
sciences, minded persons working with materials at hand will be consti-
tutive of social causal mechanisms.
Because these themes are interconnected, the volume introduces a
philosophy, or meta-theory, for social science. Uncritically in¬‚uenced
by long outdated doctrines in the philosophy of science, the volume
argues that, among both philosophers and social scientists, there remains
a widespread set of misconceptions about the tasks and limits of social sci-
ence. We need to understand that there are important differences between
the scienti¬c study of nature and the scienti¬c study of society, but we
need ¬rst to be clear about the nature and goals of science more generally.
By drawing on and integrating recent developments in the philosophy of
science, this volume aims to do this.
The structure of the argument is as follows: Chapter 1, “Explanation
and understanding,” begins with a close examination of the so-called
“Deductive-Nomological” (D-N) or “covering law” model of explana-
tion. It is of considerable interest to note that while the critical literature
of this model is now of long standing, and that while many writers, both
in philosophy and the social sciences, have rejected the epistemology of
empiricist (“positivist”) theory of science, many of these same writers
fail to see that a powerful alternative to the D-N model of explanation is
available. Once it is shown that understanding is the primary goal of the
sciences, the whole edi¬ce of science™s empiricist philosophy crumbles “
from its metaphysically implausible event ontology, including its contra-
empirical constant conjunction conception of causality, to its conception
of theory. We show then that understanding requires appeal to causal
mechanisms properly conceived as productive powers. The chapter pro-
vides both illustration and argument for these ideas.
Chapter 2, “Theory, experiment and the metaphysics of Laplace,”
argues against what is sometimes termed “deductivism,” the idea that
Introduction 3

theories in the physical sciences can be fully expressed as a deductive
system, with axioms and deductions therefrom. Rather, following the too
often neglected work of Rom Harr´ (1970), it is argued that theories iden-
tify how “things” “ molecules and atoms, for example “ are structured,
and how they interact. Theories, of course, are representations, but they
are meant to represent reality, as it is in-itself. We look then at anti-realist
criticisms of this view of theory, provide an account of experiment as it
actually functions in science, and offer a post-positivist (post-Kuhnian)
account of theory acceptance. The chapter concludes with a criticism of
the Laplacean metaphysics assumed by empiricist theories of science. In
our world, most events “ birth, growth, rain, ¬res, earthquakes, depres-
sions, revolutions “ are the products of a complex nexus of causes of many
different kinds, conjunctively at work. Indeed, it is for this reason that
the natural sciences, instead of seeking to explain concrete events, more
modestly seek to understand the mechanisms and processes of nature.
This means that while everything is caused, there is radical contingency
in both natural and human history. The implications of this are critical
for a human science, as Chapter 3 shows.
On the basis of the foregoing account of science, Chapter 3, “Expla-
nation and understanding in the social sciences,” offers a philosophy of
social science, making clear the critical points of difference in the subject
matter of the natural and social world and the implications for inquiry.
After setting out and rejecting, by way of summary, the key prevailing
assumptions regarding science, an account of “persons” is developed.
The view of causality already set out is critical here. Once we notice that
a host of causal mechanisms, biological, psychological and social, are epi-
genetically implicated in the constitution of a human being “ and of their
concrete actions “ we can see that “nature” and “nurture” are inextricably
involved and that, in consequence, there is no reason to believe that any
one science, psychological or social, could improve on the way we ordinar-
ily explain and predict behavior. As with the natural sciences, the task of
the social sciences is understanding how social mechanisms “structure,”
but do not determine, outcomes. We turn then to an account of how this
is to be conceived, drawing on the key distinction between “brute facts,”
or facts about features of the world that exist independently of us, and
“institutional facts,” or facts about features of the world which require
human institutions for their existence (Searle, 1995). The usual “subjec-
tive / objective” dichotomy is fruitfully undermined. Following Giddens
(1984), then, social structure is conceptualized as “real,” incarnate in the
activities of persons, but, accordingly, having no independent existence.
If so, versus stronger forms of the idea of social structure, it cannot, like
a magnetic ¬eld, for example, be causal.
4 A Realist Philosophy of Social Science

Chapter 4, “Agents and generative social mechanisms,” applies the
causal mechanism analogy to physical science. In the physical sciences,
regression to more fundamental mechanisms is sometimes possible. So
quantum theory offers a generative mechanism of processes in molecu-
lar chemistry. But in social science, since persons are the critical causes
of everything that occurs in the social world, the generative mechanisms
are the actions of persons “working with materials at hand,” and no fur-
ther reduction is either possible or necessary. Drawing on the argument
of agent / structure duality, a systematic account of the construction of
models of social mechanisms is offered. The chapter offers a range of
illustrative examples drawn from writers including Marx, Willis, Goff-
man, Tilly, Ogbu, Burawoy and others. For example, following Willis,
a social mechanism can be theorized which gives us an understanding
of why working-class kids get working-class jobs. Typically this involves
identifying their place in society, their beliefs about their “world” “ some
true and some false “ typical behavior predicated on these beliefs, and the
mostly unintended consequences of this behavior. The argument shows
that an ethnographic (and hermeneutic) moment is essential to grasping
a social mechanism, but as Weber had long since noted, it was but the
¬rst step in social scienti¬c inquiry. That is, while we need to understand
the social world as its members understand it, we need to go beyond this
and to consider the adequacy of their understanding of their world. Since
social process is the product of our activity, and since members may well
misunderstand their world, social science is potentially emancipatory.
Chapter 5, “Social science and history,” is very much in¬‚uenced by the
work of Max Weber. It looks critically at the question of history and its
relation to sociology, beginning with the century-old debate over the dis-
tinction between two kinds of science, “nomothetic” and “idiographic,”
and the attending argument that explanation in the nomothetic sciences
proceeds by appeal to “general laws,” while explanation in the human
sciences requires verstehen and a narrative rhetorical form. The idea of
a historical sociology gives us direct access to current versions of the
pertinent issues, including the role of comparative analysis in identifying
causes. Disagreements over the nature of a historical sociology can be
resolved with a proper understanding of the nature and goals of social
science. Brie¬‚y, if the goal is understanding, for example, why working-
class kids get working-class jobs (Willis) or why in “total institutions”
(Goffman) outcomes are inconsistent with their explicit goals, one does
not require history, even if, as Weber insisted, our interest remains the
historically concrete. That is, unlike the natural sciences where there
are “general theories” of generative mechanisms, in the social sciences,
the generative social mechanisms are always historically situated. Thus,
Introduction 5

while the generative mechanisms of, for example, oxidization are the
same everywhere, the mechanisms that explain why working-class kids
get working-class jobs need to be concretely theorized. Social science
very often goes beyond the effort to understand a social process. Unlike
the “abstract” natural sciences, it seeks to explain concrete events and
episodes, for example, the collapse of a regime, a depression, a dramatic
rise in divorces. To achieve this goal, in addition to an understanding of
the pertinent concrete generative mechanisms, one also needs history “
as Weber rightly contended. In these cases, explanation takes the form of
a narrative that identi¬es the critical social mechanisms and links them
sequentially with the contingent but causally pertinent acts of persons.
Chapter 6 looks critically at one of the most in¬‚uential and thoroughly
theorized social mechanisms in the social scienti¬c literature: the mar-
ket model of neo-classical economics. This tradition was quite correct
in what it sought to do, and its dif¬culties do not stem from its attempt
to offer explanations in terms of actors. The problem is not that mar-
kets are not social mechanisms which can give us an understanding of
outcomes by appeal to the actions of persons “ the bogey of methodolog-
ical individualism “ but that the basic model makes assumptions about
explanation, and very strong assumptions about the actors, their condi-
tions and their behavior, which simply are not sustained, except perhaps
in the remotest of cases. Mainstream neo-classical theory accepts the cov-
ering law model of explanation and a deductivist conception of theory.
If this idea of science is misconceived, however, then these models are,
on their face, poor grounds for thinking that economics is an advanced
social science. Moreover, in order to carry out its deductivist program, the
assumptions of the theory bear little relation to reality. Put succinctly, on
the mainstream view, persons are conceived as atomized, and as histori-
cally indifferent “rational beings” with approximately similar motivations.
Even more importantly, they are conceived as having approximately equal
powers and capacities. But CEOs of corporations, mom and pop Chinese
restaurateurs, heart surgeons, immigrant farm workers, non-unionized
plumbers, unionized auto workers, part-time female sales clerks, public
school teachers and drug dealers “ one could go on “ do not have similar
beliefs or capacities, either as producers or consumers. Drawing on famil-
iar criticisms, the chapter examines critically the neo-classical model and
argues that it has been burdened by a spurious (positivist) theory of social
science. Markets are important social mechanisms, but, drawing on the
account of the preceding chapters, a sociologically richer model is shown
to be both possible and necessary.
There are four appendices. They are included as appendices because
they address the critical literature and provide supplementary materials
6 A Realist Philosophy of Social Science

not essential to the central argument of the volume. Appendix A treats
the limits of multiple regression and similar techniques, given a proper
understanding of causality and explanation. Appendices B and C pick
up on arguments in the current literature that are highly relevant to the
arguments of the volume. Appendix B considers the dispute between
Theda Skocpol and William Sewell regarding narrative and causal
analysis. A very recent defense of the use of Mill™s methods in historical
sociology is examined critically. The goal of comparative work is further
clari¬ed. Appendix C considers the lively debate in The American Jour-
nal of Sociology over the pertinence of rational choice theory in historical
sociology. The effort is made to clarify the argument and to resolve it.
Finally, appendix D offers some additional explication and criticism of
neo-classical theory.
1 Explanation and understanding

Despite some contentiousness between both working social scientists and
many philosophers, ideas about explanation in social science are remark-
ably taken for granted. Worse, when examined in the clear light of day,
there is good reason to say that these taken-for-granted ideas are down-
right wrong. Most social scientists have been socialized to carry on inquiry
as de¬ned by their disciplines, they have well-de¬ned research projects
and, perhaps quite reasonably, they are content to leave the philosophical
problems to the philosophers. No one presses them to wonder whether,
indeed, key assumptions unre¬‚exively absorbed are problematic. Some
very good work is done that cannot be squared with their implicit or,
sometimes, explicit background assumptions. Not only is it not always
easy to tell others exactly what we are doing, but we can be mistaken
about what we are doing. In his 1933 Herbert Spencer lecture at Oxford,
Einstein, often ahead of most people, offered pertinent advice: “If you
want to ¬nd out anything from the theoretical physicists about the meth-
ods they use, I advise you to stick to one principle: Don™t listen to their
words, ¬x your attention on their deeds.”1
A good deal of the responsibility for the taken-for-granted ideas about
explanation among social scientists owes directly to philosophers in the
latter half of the twentieth century, although the antecedents are found
as early as Comte in the early nineteenth century. Comte, inventor of the
term “positivism,” had argued that “the explanation of facts is simply
the establishment of a connection between single phenomena and some
general facts,” or in other words, a scienti¬c explanation was a deduction
from general laws. His reasons for this are also pertinent. He was much
concerned to put science on a secure empirical foundation, to expunge
“¬ctitious ideas,” both metaphysical and religious, from scienti¬c expla-
nation. These concerns and ideas were powerfully reinforced by a host

1 Quoted from Holton 1970 in Manicas, 1987: 242.

8 A Realist Philosophy of Social Science

of philosopher / physicists in the later quarter of the nineteenth century.
The list is impressive and included G. R. Kirchoff, Wilhelm Ostwald,
Ernst Mach, Ludwig Boltzman, Karl Pearson, Henri Poincar´ , Pierre e
Duhem and William Thompson (Lord Kelvin). The philosophers of
the so-called “Vienna Circle” picked up on these ideas in the 1920s and
developed what came to be the dominating theory of science, “logical pos-
itivism” (or “logical empiricism”). Central to these doctrines was what
came to be called the “Deductive-Nomological” (D-N) or “covering law”
model of explanation.3
The majority of social scientists working today are not particularly
aware of this history or of their indebtedness to these ideas. But they
appear in seemingly innocent phrases like “the search for laws is the
goal of science,” “science aims at prediction and control,” “a theory is
a deductively organized set of propositions and law-like statements,” “a
good theory predicts.” The relatively few methodologically oriented dis-
cussions by social scientists paying special attention to the social sciences
have taken the D-N account for their point of departure, either to show its
pertinence to their domain,4 or to argue that if this is the correct model of
scienti¬c explanations, then the human sciences cannot provide them.5

The covering law model of explanation
While in some quarters at least, the critique of the covering law model will
be old news, if we are to make sense of explanation, both in the natural
and social sciences, we need to be clear about the model and its failings.
Consider ¬rst the classic formulation as put forth by Carl Hempel.6 For

2 See Manicas, 1987 and for an excellent fuller treatment, John Passmore, 1957: chapter
3 An excellent history of views of causality and explanation from the Greeks to the logical
empiricists and their critics is found in Wallace, 1974. While the covering law model is
a de¬ning attribute of “empiricist” (positivist, neo-positivist) understandings of science,
there is now a substantial critical literature which has subjected this assumption to fatal
criticisms. See, among others, Scriven, 1959, 1962; Harr´ , 1970, 1986; Dretske, 1977;
Bhaskar, 1975; Salmon, 1978, 1984; Achinstein, 1981; Aronson, 1984; Woodward, 1984;
Lewis, 1987; Kim, 1987; Manicas, 1987, 1989a. In what follows, I draw on some of the
main lines of such criticism.
4 Outstanding examples include Friedman, 1968 and Merton, 1957. More recently, see
Turner, 1987 and Alexander, 1987. While Turner defends a neo-positivist theory of
science, Alexander is explicitly “post positivist,” endorsing the developments following
Kuhn™s The Structure of Scienti¬c Revolution. But as with many others who would consider
themselves “post-positivist,” Alexander remains committed to the covering law model
and thus to the idea that it is the goal of social science to “search for laws.”
5 This is the route of so-called “interpretative sociology.” See below and chapter 3.
6 His important papers are gathered together in the volume, Aspects of Scienti¬c Explanation
Explanation and understanding 9

him a scienti¬c explanation takes the form of a deductive argument, with
premises and a conclusion:
C 1 , C2 , . . . Ck
L1 , L 2 , . . . L T
The “explanans,” C1 , C2 , . . . Ck , are statements describing the par-
ticular facts invoked, sometimes called “the initial conditions,” and L1 ,
L2 , . . . Lr are general laws. The event to be explained (the “explanan-
dum”), E, is a logical consequence of the premise set. As he said: “The
kind of explanation thus characterized I will call deductive-nomological;
for it amounts to a deductive subsumption of the explanandum under
principles which have the character of general laws.” This is helpfully
termed an epistemic conception of explanation since the relation between
explanans and explanandum is logical.7 The simplest case takes the form
of a syllogism:
If a, then b (the form of a general law)
a (the relevant “conditions”)
b (the event to be explained)
Of course, this will count as an explanation only if the premises are true.
Hempel subsequently enlarged his model to include “probabilistic
explanation” or “inductive-statistical” (I-S), where the “laws” are not
strictly universal, as in the deductive model. Instead of the premises
entailing the explanadum, the event to be explained is but probable on
the strength of the premises. So roughly,
The probability of b, given a, is very high.
probably b.
Moreover, Hempel went on to argue that nomological explanations,
deductive and inductive, could be found in historical writing, where the
“relevant generalizations” are sometimes suppressed, and in two special
cases of explanation in history, what he termed “genetic explanations”
and “explanation by motivating reasons.” It was assumed, to be sure, that
the models applied also to all explanation in the social sciences.
In the 1950s, a hardly noticed critical literature of what came to be
called “the standard view” began to develop. By now there are a number
7 Epistemology is inquiry into the grounds of knowledge (Greek: episteme, Latin: scientia)
and thus includes, critically, logical analysis. Our alternative account is termed “ontic.”
Ontology is inquiry into the nature of the “real,” which, after Kant, became scienti¬cally
10 A Realist Philosophy of Social Science

of fatal objections to the model, but before we get to these, notice ¬rst
that there is currently no consensus among philosophers of science for an
alternative account. This chapter attempts to provide at least the sketch
of an alternative. Secondly, and as important, the critique of the covering
law model has not yet ¬ltered into the disciplines of the social sciences.8
Hempel™s overall conclusion is also important. He insisted that his claims
did not
imply a mechanistic view of man, of society, and of historical processes; nor,
of course, do they deny the importance of ideas and ideals for human decision
and action. What the preceding considerations do suggest is, rather, that the
nature of understanding, in the sense in which explanation is meant to give us
an understanding of empirical phenomena is basically the same in all areas of
scienti¬c inquiry. (1965: 41)
For most inquirers, this was reassuring, which contributed to the account
becoming conventional wisdom. To be sure, not everyone agreed with
Hempel on these matters, often dubbed “naturalism” in the philoso-
phy of the human sciences. A variety of writers, called “anti-naturalists,”
could not see how, given any of Hempel™s models, one could escape a
“mechanistic view of man, society, and historical process.” This was usu-
ally joined to the claim that getting an understanding of human action
in society and history was not at all “basically the same” as getting an
understanding of nature, that a very different idea of explanation was
required. On this view, any sort of causal explanation in the human sci-
ences was wrongheaded. The alternative, then, was the idea that human
action could only be explained in terms of the meanings of actors; hence
the appeal to verstehen (understanding) or what is sometimes called “inter-
pretative sociology.” Weber, of course, had insisted, rightly on the present
view, that there was no opposition between verstehen and causal explana-
tion (erkl¨ ren) and that, indeed, both were required in the human sci-
In chapter 3, we need to consider carefully these objections. In some
ways they go to the heart of the problem of a human science. But the prob-
lem we need to address ¬rst is not whether there are important analogies
8 Some evidence for this assertion may be found in chapter 6 and appendix C below. See
also Tilly, 2001: 25. See also, of course, the standard textbooks for the entry-level courses
in the social sciences.
9 Originally, “hermeneutics” referred to the effort to understand and interpret religious
texts. In opposition to the Comtean view that there were laws of history, Droysen
argued that we needed to understand mind (Geist) differently than nature. Thus, ver-
stehen gives humans access to meanings. Dilthey developed this idea. His work motivated
Weber™s effort to resolve the opposition between understanding, understood as verstehen,
and causal explanation. This became part of the important Methodenstreit. See below,
chapter 5.
Explanation and understanding 11

in achieving explanations in the physical and human sciences. That is to
say, it is not merely that the D-N model does not work as regards expla-
nation in the human sciences; rather, the problem is that it does not work
at all. Much of what follows in this chapter demonstrates this.
Notice ¬rst that Hempel™s account took the explanation of events as
the primary task. Events can be conveniently understood as being space /
time locatable, concrete and particular. There are those which typically
¬gure as conclusions in exemplary covering law explanations, for example
(referring to a robin™s egg in my hand now), “This egg is a robin™s egg:
that is why it is greenish blue,” or “At time T, the body in free fall fell
sixteen feet,” or “At time T, the salt dissolved in water” or “At time T,
the moon was at coordinates x, y, z.” They may be more earthy and
refer to “happenings”: “Kilauea erupted on July 5, 1978”; “Tornados
hit Nebraska in June of 2003.” Or they involve the actions of persons:
“On January 11, 1998, Sam robbed the convenience store at 16th Street
and Broadway,” or more dramatically: “The two towers of the World
Trade Center collapsed on September 11, 2001.” They may also involve
complexes which are suf¬ciently related to allow us an adequate term of
denotation, such as the expiration of the woolly mammoth, the passing
of the Paleolithic era, the French Revolution, the Great Crash of 1929.
More might be said here regarding the idea of an event, but for present
purposes this will suf¬ce.
Perhaps surprisingly, in most of the well-established natural sciences,
attempts to explain particular events are rare. Where they occur, they
occur almost entirely in those natural sciences that are historical, such as
meteorology or geology, or where, as in astronomy, for some purposes at
least, time is irrelevant, for example, in giving an account of the position
of the earth™s moon on any given day. There are good reasons for both
their absence in what Max Weber usefully called “the abstract sciences,”
physics, chemistry, biochemistry and general biology, and for their pres-
ence where they do occur, as we shall see.10 In what follows, we argue
that, following both common usage and the practices of philosophers, the
idea of explanation is most at home in contexts where the explanation of
an event is what is aimed at, and that, by contrast, physicists, chemists
and biochemists aim at understanding “ a notion not at all easy to clar-
ify, but certainly not to be confused with understanding in the sense of
verstehen as that is generally understood.
But the foregoing suggests another observation: on the covering law
model, explanation and prediction are symmetrical. If you are in a position to
explain some event b, then b could have been predicted “ and conversely.

10 See his seldom studied Roscher und Kneis (1975) and Manicas, 1987: 127“140.
12 A Realist Philosophy of Social Science

The logic is the same. If we have a law of the form “If a, then b,” then
given that a has occurred, we explain b. Similarly, by means of the law, we
can predict b, if and when a occurs. This has some prima facie plausibility
especially if we have as our paradigms those events which are typically
instances of laws in the D-N model: if you put salt into water, it dissolves.
But if we consider events like the collapse of the Twin Towers, the devas-
tation of hurricane Iniki to Kaua™i in the Hawaiian Islands in 1992 or the
Great Depression of 1929, which seem, at least, to involve a battery of
laws, and in which sequences seem critical, the covering law model seems
far less plausible. Once something happens, we can begin the search for
an adequate explanation. But could these events have been predicted?
How one answers this question depends on one™s ontology, as we shall
The view that explanation and prediction are symmetrical leads to the
idea that the goal of science is prediction, indeed, that an explanation
is not scienti¬c if it does not enable prediction. As Hobbs (1993: 177)
writes: “Ex post facto explanations, those which are given only after the
event to be explained has occurred, have long been considered epistemo-
logically suspect, along with the theories that sanction them.” I do not
know if Hobbs is correct in saying that the symmetry thesis is no longer
assumed by most philosophers, but the idea remains pervasive (and per-
nicious) in the social sciences where, paradoxically, the failure to provide
“good predictions” is taken to be the major fault of the social sciences.
This, too, we need to challenge. We can begin with an effort to clarify the
idea of explanation and look also at its important relative, understand-
ing, an idea much used when writing about science, but almost never
considered by philosophers of science.

Clarifying “understanding” and “explanation”
Both “understanding” and “explanation” have many uses and, indeed, in
some contexts at least, are interchangeable. Our interest here is scienti¬c
explanation and in the understanding offered by science. We consider
¬rst “understanding.” R. L. Franklin has rightly noted:
“Understand” is a word we understand as well as any, but we do not understand
philosophically what it is to understand. The word catches some notion important
enough to appear in many of our book titles, yet in an age of linguistic analysis it
has virtually escaped investigation in English-speaking philosophy. (1983: 307)

This is perhaps paradoxical. But defenders of the D-N model, like most
philosophers of science, suggest that the problem is not theirs, since con-
cepts like understanding and intelligibility are psychological and prag-
matic. Michael Friedman is an important exception. He notes, rightly,
Explanation and understanding 13

that “the entailment relation puts a constraint on the explanation relation,
but it does not by itself tell us what it is about the explanation relation
that gives us understanding of the explained phenomenon, that makes
the world more intelligible” (1974: 7). This is indeed the problem: how
do we go about achieving an understanding of the phenomena? What
makes the world “more intelligible”?
We need, of course, to be wary of an account of understanding which
is subjective in the sense of it being entirely arbitrary what will count as
generating understanding. If we are interested in the understanding that
science can provide, then not just any process of generating understanding
will do. While we aim for an account that is objective, the account will
necessarily be pragmatic and psychological in that it will involve exam-
ining both our aims and interests, and the contexts in which scienti¬c
understanding is demanded and achieved.
Nobody has so far provided a rigorous and satisfactory account of
understanding and you should not suppose that you will get one here.11
Nevertheless, it is fairly easy to understand how scienti¬c understanding
is achieved. This requires some radical revision of conventional wisdom
in the philosophy of science, especially regarding the critical concept of
causality, central on the present view to both understanding and expla-
Friedman, who noticed that explanations of events were rare in the
physical sciences, offered instead what he rightly took to be typical
questions for science and the typical answers to these questions. These
included the following questions: Why does water turn to steam when
heated? Why do the planets obey Kepler™s laws? Why is light refracted by
a prism? The answer to the ¬rst question is:
Water is made of tiny molecules in a state of constant motion. Between these
molecules are intermolecular forces, which, at normal temperatures, are suf¬-
cient to hold them together. If the water is heated, however, the energy, and
consequently, the motion, of the molecules acquire enough energy to overcome
the intermolecular forces “ they ¬‚y apart and escape into the atmosphere. Thus,
the water gives off steam. (Friedman, 1974: 5)

This account, while informal, does give us understanding; it does make
the world more intelligible. It is not an explanation of an event, but
may be construed (as by Friedman), as the explanation of a pattern, or
regularity “ or, very loosely, a “law.” The understanding comes from see-
ing that water “ like everything else “ is composed of molecules and that

11 See Friedman, 1974, and the responses it generated: Kitcher, 1976, 1981; Gemes, 1994;
Hintikka and Halonen, 1995. All of these are more or less efforts at “rigor” and all confess
that they leave much unanswered. On the present view, all omit consideration of causality
and fail to offer an appropriate analysis of the ontology of science.
14 A Realist Philosophy of Social Science

there is a great deal that can be said about how they behave. Indeed,
this story gives us an understanding of not only the changing of water
to steam, but of an extraordinary list of experienced phenomena: the
dissolving of salt in water, the rusting of iron, the nutritional capaci-
ties of broccoli . . . the list is nearly endless.12 Indeed, as Peter Atkins
(2003: 135) notes in his lovely book, Galileo™s Finger, “chemistry is the
bridge between the perceived world of substances and the imagined world
of atoms.” While patterns are the material for questions and questions
demand answers which, in science at least, call for theory, the real goal
of science is neither the explanation of events nor the explanation of pat-
terns, though this idea catches some of the truth of the matter. Rather,
it has as its goal an understanding of the fundamental processes of nature
(Harr´ 1970: 260“266; Bhaskar 1975: 17, 66). Once these are under-
stood, all sorts of phenomena can be made intelligible, comprehensible,
Friedman™s example, of course, employs critical terms in the theoretical
discourse of science: molecule, intermolecular forces, energy. The depth
of understanding attained depends on the knowledge of the audience.
What after all is a molecule? What are molecular forces? What is energy?
Most of us, perhaps sadly, lack any sort of understanding of these ideas “ a
regrettable failure of science education in the USA (and elsewhere). This
is doubly sad since not only does this ignorance contribute to reinforcing
misunderstandings about science “ for example, that its goal is prediction
and control, or that theory is a deductive system “ but these ideas can be
made available without extensive socialization in a science or an extensive
background in sophisticated mathematics.
By examining how energy is dispersed in a steam engine, Atkins
gives a highly informative, non-mathematical account of still more

12 Friedman says that “scienti¬c explanations do not confer intelligibility on individual phe-
nomena” by showing them to be somehow “natural, necessary, familiar, or inevitable,”
and that we need to pay heed to the “global features of explanation,” the idea that “our
total picture of nature is simpli¬ed via a reduction in the number of independent phe-
nomena that we have to accept as ultimate” (1974: 18). This is “uni¬cation” in one
clear enough and important sense, even if it is not easy to provide a rigorous formal
account of the notion. As Aronson, 1984 argues, the uni¬cation missed by Friedman
regards showing that otherwise disparate phenomena have a common ontology. There
is a sense, contra Friedman, that “natural,” “necessary” and “inevitable” “ though not
“familiar” “ are involved in understanding. See below.
13 Including, perhaps especially, patterns which are anomalous. Two characteristics of
black-body radiation which were identi¬ed in the nineteenth century and became known
as Wien™s and Stefan™s laws could not be explained in terms of classical physics. The
problem provoked Lord Kelvin and then Max Planck to propose what came to be quan-
tum theory; providing an enormously improved understanding of “the deep structure of
reality” (Atkins, 2003: 204).
Explanation and understanding 15

fundamental processes involved in the changing of water to steam. He
Let™s suppose that the fuel is oil, a mixture of hydrocarbons (compounds built up
from carbon and hydrogen only) . . . [A hexadecane molecule is easily represented
in a graphic model, ¬g. 4.11 in his book, as a chain of sixteen carbon atoms to
which are attached thirty-four hydrogen atoms.] This is the molecule typical of
fuel oil and diesel fuel; it is also closely related to the molecules of fat that are
present in meat and which help to lubricate the muscle ¬bers as well as acting
as an insulating layer and a reserve of fuel. That we eat foodstuffs closely related
to diesel fuel, some more than others, is no accident, but the thought is a little
When the oil burns, molecules like the one in the illustration are attacked
by oxygen molecules of the air. Under the onslaught of the attack, the carbon
chain breaks up and the hydrogen molecules are stripped from it. The carbon
atoms are carried away as carbon dioxide molecules and the hydrogen molecules
are carried away as water molecules. A great deal of heat is produced because
the new formed bonds between the atoms are stronger than the original bonds
in the fuel and in the oxygen, so energy is released when the weak old bonds
are replaced by strong new ones and the atoms settle into energetically more
favorable arrangements. And why does the hydrocarbon burn? Because in doing
so there is a huge decrease in disorder and therefore in entropy. There are two
principal contributions to this increase in entropy. One is the release of energy,
which disperses into the surroundings and raises their entropy. The other is the
dispersion of matter, as long, orderly chains of atoms broken up and the individual
atoms spread away from the site of combustion as little gaseous molecules. The
combustion is portraying the content of the Second Law [of Thermodynamics].
(Atkins 2003: 128“129)
This account, unlike the example of the boiling water, deepens our under-
standing by showing how molecules break up and recombine into new
molecules. Combined with oxygen, the hexadecane molecule becomes
carbon dioxide and water; and entropy explains the release of energy in
this process. Now, of course, we need an account of the forces which
explain the “bonds,” we need an account of energy “ an absolutely per-
vasive feature of all processes in the universe “ and we need to understand
entropy, “the spring of all change.” As Atkins writes, with an understand-
ing of entropy at hand, “we shall come to understand the simple events
of everyday life, such as the cooling of hot coffee, and we shall see at least
the ankle of the explanation of the most complex events of everyday life,
such as birth, growth and death” (Atkins 2003: 109).
Atkins™s way of putting the situation is just right. “We get an under-
standing but only an ankle of an explanation of the most complex events
of everyday life.” The explanation of events presupposes understanding,
and we get only “an ankle of explanation” because we need a good deal
more than an understanding of entropy. But the foregoing account does
16 A Realist Philosophy of Social Science

not even touch on questions of the forces which explain inter- and intra-
molecular processes, nor have we mentioned the level below atoms, the
world of electrons, s- and p-orbitals, quarks, waves and particles: essen-
tial for a still deeper understanding of not only all chemical processes,
but of matter itself.
Reaching an adequate understanding at this level will not be easy “
but it need not be that demanding either. Questions asked are pragmatic
and aim at serving some purpose; hence also the answer given. A small
child asking why the egg in my hand is greenish-blue may be satis¬ed to
be told that it is a robin™s egg and that all robins™ eggs are greenish-blue.
Indeed, even that answer offers some understanding in that it offers an
order to the experience. But most of us will want more. And surely a
scienti¬c answer to this question would require a good deal more. This
requires an account of causality and how it functions in science.

Understanding and causality
Such examples give a general picture of how understanding proceeds
in physical science, but we need now to connect these examples to an
account of causality including, critically, the idea of a causal mechanism,
which is an important feature of both understanding and the explana-
tion of events. On the present view, the aim of science is to provide an
understanding of the fundamental processes of nature and this requires
identifying the causal mechanisms which are, willy nilly, at work in the
14 My account is in¬‚uenced by the writings of Harr´ , 1970 and Bhaskar, 1975. But see
also Cartwright, 1989 and, more recently, the accounts of Glennan, 1996; Machamer,
Darden and Craver, 2000, and Bunge, 2003. There are differences, terminological and
substantive, between these writers, which are not developed here. One critical difference
regards the understanding and pertinence of “laws.” The account of Machamer et al.
(2000: 3) seems closest to the present account. They de¬ne mechanisms as “entities
and activities organized such that they are productive of regular changes from start or
set-up to ¬nish or termination conditions.” Bunge objects, insisting that this de¬nition
is incomplete and “misses the concept of a concrete system “ one of the categories
sadly absent from mainstream ontology, along with those of matter, energy, state and
emergence” (Machamer et al., 2000: 3). But the reference to a “concrete system” may
raise more problems than it solves. See chapter 4, below. Machamer et al. may be correct
in saying that “there is no adequate analysis of what mechanisms are and how they work
in science” (2000: 2). Of course, “adequate” is a pragmatic term. But all these writers
would seem to agree that causal mechanisms are critical to understanding and that this is
the goal of science. As Machamer et al. put the matter: “In many ¬elds of science what is
taken to be a satisfactory explanation requires providing a description of a mechanism”
(2000: 1). Bunge notes, using “understanding” in the sense of the account in this chapter,
that “the relevance of mechanism to understanding is such that it is not uncommon to
¬nd in the scienti¬c literature apologies of the form, ˜Unfortunately, no mechanism is
known to underlay the fact [or equation] in question™ ” (Bunge 2003: 186). We noted
earlier the often interchangeable uses of the terms “explanation” and “understanding.”
Explanation and understanding 17

As before, let us begin with the dominating, though mistaken, view of
causality. This view comes from David Hume (1711“76), who argued that
all we can know of a causal relation is that there is an observable constant
conjunction between two events. Thus “a is the cause of b” means nothing
more than “if a, then b.” The idea that a cause is a productive power was
dismissed as a metaphysical idea since, on this view, there is nothing in
experience which says that a produces b. The Humean view of the matter
is not the common-sense view of the matter “ nor is it empirical, if that
means known by experience. On this view, when we push a door open, our
action was the cause, and it produced the outcome. Indeed, as Harr´ ande
Madden write: “Can anyone seriously deny that we sometimes veridically
perceive the waves eating away the shore, the axe splitting the wood, and
the avalanche destroying the countryside” (1975: 49). Terms like “eating
away,” “splitting” and “destroying” are clearly causal concepts, and it is
likely that our ordinary understanding of causality comes directly from
our experience, especially the experience of our own actions as causes in
contexts like pushing open a door.
Nor, according to the Humean conception, can we impute necessity
to the relation: as empirically established, the connection of a to b is
purely contingent. As Hume put the matter: “If we have really no idea
of a power or ef¬cacy in any object, or of any real connection betwixt
causes and effects, it will be to little purpose that an ef¬cacy is neces-
sary in all operations” (Hume, 2000: Part III, Section xiv). But as shown
by the active verbs (above), which function as causal terms, there is no
problem in seeing a “real connection,” even if we will need to say more
about “power and ef¬cacy” in the scienti¬c application of causality. The
presumed absurdity that one explains the drowsiness which comes from
taking opium by saying that it has a “sopori¬c power” (virtus dormitiva),
is the pertinent example here. Avoiding such “absurdity” powerfully moti-
vated the Humean idea that causal laws had to be analyzed as invariant, if
contingent, relations, as Comte put it, “of association and resemblance.”
On this view, according to the dominating strand in philosophy of sci-
ence, to say that opium is a “sopori¬c power” is to say only, “if one takes
opium, one becomes sleepy,” where, as above, this is an invariable relation
of association.
There were many reasons why the Humean view became conventional
wisdom. But foremost were the empiricist prohibitions which motivated
Hume and Comte. “Science” had to ¬ght off the metaphysical philoso-
phers and the theologians if it was to establish its independent authority,
and that meant that experience and experiment would be the anchors
of its claims. Indeed, Pierre Duhem (1861“1916) went even further and
insisted that since we could not disconnect explanation from causality,
18 A Realist Philosophy of Social Science

science does not seek to explain. He wrote: “A physical theory is not an
explanation. It is a system of mathematical principles, which aim to
represent as simply, as completely, and as exactly as possible a set of
experimental laws” (1954: 19). Bertrand Russell (1872“1970), explicitly
rejecting the common-sense idea as pertinent to modern science, noted
that causality was “a product of a bygone age” and suggested that it be
expunged as unnecessary to science. As this also suggests, part of the
reason for abandoning the commonsensical notice of causality was the
capacity to employ mathematics to express relations and make deductions
from these. Newton™s systemization of celestial mechanics was very much
the background to this. Thus, at the end of the nineteenth century, Ernst
Mach (1838“1916) argued that mathematical functions of theory were
“abridged descriptions.” The compendious representation of the actual,
necessarily involves as a consequence “the elimination of all super¬‚uous
assumptions that are metaphysical in Kant™s sense” (1959: 210), that is,
as exceeding the bounds of experience.15
This, then, takes us back to the covering law model. Indeed, it was
the assumed Humean notion of causality which gave it much, if not all,
of its power. To anticipate, if causal necessity had to be expunged from
science, then logical necessity might well serve in its place. This involved,
more fundamentally, an epistemic conception of explanation instead of
the commonsensical ontological conception “ which, presumably, had no
place in metaphysically cleansed science. That is, instead of construing
explanation in terms of causes as productive powers, it was construed in
terms of rational argument.
Thus in a much used and highly regarded textbook, Research Methods in
the Social Sciences, Frankfort-Nachmias and Nachmias (1992: 10) write:
“Ever since David Hume . . . an application of the term explanation has
been considered a matter of relating the phenomenon to be explained
with other phenomena by means of general laws.”
Let us go back to the simplest form of the D-N model and see both
how the Humean conception ¬t neatly into the by then well-conscribed
logical analysis of science, how it avoided the pragmatics and psychology

15 An important nineteenth-century exception was Hermann Helmholz. In the twentieth
century the exceptions include Albert Einstein, Max Planck and David Bohm. In a
letter to Schlick, Einstein pertinently insisted: “In general your presentation fails to
correspond to my conceptual style insofar as I ¬nd your whole orientation so to speak
too positivistic . . . I tell you straight out: Physics is the attempt at the conceptual
reconstruction of a model of the real world and its lawful structure . . In short, I suffer
under the unsharp separation of Reality of Experience and Reality of Being . . . You will
be astonished about the ˜metaphysicist™ Einstein. But every four- and two-legged animal
is de facto in this sense a metaphysicist” (quoted by Holton, 1970: 188).
Explanation and understanding 19

of understanding and how, by illicit con¬‚ation, it pretended to do what
was needed.
If salt is put in water, then it dissolves.
The salt was put in water.
Therefore, the salt dissolved.
As a deduction, the explanandum follows logically. If P, then necessarily Q.
If the premises are true, we have provided suf¬cient grounds for believing
that the explanandum is true. So, presumably, this fact is explained. But
it is easy to construct counter-examples of D-N explanations with true
premises which are just plain silly. They not only do not explain, but
they do not even provide grounds for believing that the explanandum is
Anyone who takes birth control pills regularly will not become pregnant.
John took his wife™s pills regularly.
Hence, John did not become pregnant.
When a woman takes a birth control pill, there is a causal mechanism
at work which prevents pregnancy. This would explain why Joan did not
get pregnant. But this is plainly not what is at issue as regards John. Given
that he is male, it is biologically impossible for him to become pregnant.
(For a Humean, the only impossibility is logical.)
It is easy to construct D-N “explanations” with true premises in which
there is not even the suggestion of causality. Suppose, for the sake of argu-
ment, that there is perfect correlation between the price of eggs in China
and the behavior of Microsoft stock the next day on the New York Stock
Exchange. We can construct a D-N explanation which explains the price
movement of Microsoft by appealing to the price movements of eggs in
China! As I argue subsequently, strong correlations are most useful for
prediction though not, symmetrically, for explanation “ as nearly every-
one would agree.16 That is, if it is true that there is a strong correlation
between these two variables, I can act pro¬tably on the NYSE by know-
ing the price of eggs in China today, whatever is the correct explanation
for price movements on the stock exchange tomorrow.
The point here is that even logical necessity between premise set and
conclusion does not make the argument an explanation. This is the
wrong relation. We need necessity but not logical necessity. Laws (like the

16 Both Wien™s and Stefan™s laws (above, note 9) offered perfect correlations which, func-
tioning as D-N explanations, manifestly fail. Indeed, while every textbook in quantitative
methods warns students not to confuse correlation and causation, it is usually left mys-
terious how it does this. Indeed, it is left to one™s common sense and intuition. That is,
one must suspect that there is or is not a mechanism at work. See appendix A.
20 A Realist Philosophy of Social Science

foregoing) which subsume instances (still less mere generalizations) can-
not explain since “entails” is the wrong relationship. As Dretske (1977)
writes: “The fact that every F is G fails to explain why any F is G.”
Consider again our salt example. This is plausibly an explanation “
even if the understanding conveyed is minimal. It is surely pre-scienti¬c;
the “law-like” major premise is a generalization “ and there are thousands
like it, familiar for centuries, and indispensable to ordinary life.
We can certainly acknowledge that this particular salt would not have
dissolved in that particular water unless someone put it in the water. So,
on the common-sense way of thinking, this was also the cause that brought
about the outcome. Indeed, the “if . . . then” statement even looks like
a causal law, as analyzed by a Humean. This perhaps explains some of
the confusion. If putting salt in water is necessary for the outcome, we
think we have an explanation and in some contexts, at least, perhaps this
will suf¬ce. But if it does, it is also because we take for granted that there
is something about both salt and water such that when one puts salt in
water, it dissolves. Salt is water-soluble. This is surely causal, but it is not
part of the explanation that was offered.
Worse, as stated, the law-like major premise, “if salt is put in water, it
dissolves,” is not even true: when it is put in water, salt doesn™t always
dissolve, for a variety of reasons. One might patch this up, of course, and
say that it usually does, so we have here not a D-N explanation but an
“inductive statistical” explanation. Consider what this does.
While at least a true universal maintains the hold on the individual case,
anything less loses all sense of explanation. Perhaps 67 percent of people
exposed to herpes contract it; Sam and Harry were both exposed, but why
did Sam contract the disease and why didn™t Harry? Similarly, suppose
we contend that most Texans are Republicans and that Jones is a Texan.
It follows that probably Jones is a Republican. But suppose that most
philosophers are not Republicans and that Jones is a philosopher. It fol-
lows that he is probably not a Republican. Or consider the explanation
that people who have colds will probably get over them in a week if they
drink plenty of Coca Cola. Jones did this, and he got over his cold. But not
only do most colds last about a week, but we know of no mechanism which
would link this behavior with this outcome. In these sorts of cases, there
is no explanation because there is no “real connection” between drinking
Coke and getting over the cold in one week. In the D-N case, we could
be more easily misled because at least there is logical necessity between
the explanans and the explanandum. Explanation, like understanding,
requires that there is a “real connection,” a generative mechanism or
causal nexus that produced or brought about the event (or pattern) to be
Explanation and understanding 21

A causal relation presupposes a nomic and necessary connection. We
need not balk at this. Indeed, Jaegwon Kim is prepared to say that “most
philosophers will now agree that an idea of causation devoid of some
notion of necessitation is not our idea of causation “ perhaps not an idea
of causation at all” (1987: 234). The basic idea is clear enough. Causes
bring about their effects, either as events which initiate a change in cir-
cumstances, e.g., the match which lights the ¬re, or as mechanisms with
causal powers, the combustible material which burns (and doesn™t evap-
orate or become vinegar).
We alluded earlier to “causal powers” and to “causal (or generative)
mechanisms” “ the core of an alternative realist account of causality.
Consider the following improved explanation.
If salt is put in water, then because salt is water-soluble, it dissolves.
The salt was put in water.
Therefore, the salt dissolved.
On the Humean reading, the clause, “because salt is water-soluble” is
redundant since it can only mean, “If salt is put in water, it dissolves.”
While this empiricist analysis of dispositional concepts has won consider-
able favor in contemporary accounts “ especially in the social sciences “
solubility cannot be unpacked in terms of if-then clauses. If salt is water-
soluble then there is something about it such that if it is put in water, it
dissolves. “Water soluble” is a promissory note to be ¬lled in by providing
a causal mechanism. The same is true regarding the appeal to virtus dor-
mitiva ascribed to opium. This attribution is hardly satisfying, but it does
give us the promise of better explanation because it directs us to look for
the relevant generative mechanism. Still, as Harr´ notes, even having only
the promise is an improvement, simply because the explanation no longer
supposes that the outcome results merely from the fact that it was put in
water. Nor will we understand the outcome better if we repeat the exper-
iment a hundred times. Rather, we now are directed to consider what
it is about salt and water “ or opium “ such that salt dissolves in water,
and opium induces sleep. This shifts the question from the presumption
that any law-like regularity from which one can deduce the event to be
explained counts as a scienti¬c explanation. Critically, it also shifts the
problem of explanation to the question of the nature of salt and of water “
a theoretical question for science, and as we have noted, the answer to
which can provide genuine understanding.
The physicist David Bohm wrote: “Clearly . . . the concept of a causal
relationship implies more than just regular association, in which one set
of events precedes another in time. What is implied in addition is that
(abstracted from contingencies of course) the future effects come out of
22 A Realist Philosophy of Social Science

past cause through a process satisfying necessary relationships” (1984:
5f.). Moreover, the concept that is needed and overlooked entirely by
concentrating on events as causes, is the idea that things have causal prop-
erties by virtue of their nature. It is here that necessity can be located. Bohm
Thus, the qualitative causal relationship that water becomes ice when cooled
and steam when heated is a basic part of the essential properties of the liquid
without which it could not be water. Similarly, the chemical law that hydrogen
and oxygen combine to form water is a basic property of the gases hydrogen and
oxygen . . . Likewise, the general mathematical laws of motion satis¬ed by bodies
moving through empty space (or under other conditions) are essential properties
of such bodies, without which they could be bodies as we have known them.
Examples of this kind could be multiplied without limit. (1984: 14)

We need some new language to get at what is going on here. We can
think of a causal property of a thing as an ascription of a power or ten-
dency, true of it because of what it is.17 Accordingly, causal laws are not
universal conditionals of the form, “if X, then Y.” Rather, causal laws
look more like: “By virtue of its intrinsic structure S, C phi™s when C is
triggered,” where phi-ing refers to the activity of the mechanism C. It is
important to notice here that the outcome of C™s phi-ing will be a function
of what other causal mechanisms are also at work. If salt is put in water,
it dissolves; if iron is put in water, it will rust.
Moreover, salt usually dissolves in water, but if it does not, then while
the causal mechanism that explains this may be triggered, there are other
causes at work. Something has happened, but not what we expected to
happen. Finally, ordinary, concrete salt is what we experience. NaCl is
a theoretical object arrived at by abstraction from the concrete. It exists, but
perhaps only or usually in a less than pure form. While if our theory is true,
NaCl must dissolve in H2 O, if on some occasion, the salt did not dissolve
when put in water, we would almost certainly assume that what we put in
water was not salt or that there was something about either the salt or the
water which prevented it dissolving. If suf¬ciently sophisticated, we might
suggest that the solution was super-saturated, perhaps because it was too
cold. Indeed, in our world there is contingency, but in a world where there
was only contingency, there would be no stability. Salt does not explode
when one puts it in water; nor does it change the water into gin, etc. In

17 The concept of powers does not ¬gure in the discourse of science. Nor need there be
reference to causes in this discourse, even if it is replete with terms which imply causality.
But indeed, there is frequent reference to “mechanisms.” “Powers” and causality are
terms employed by philosophers of science in the effort to get a better understanding of
how science proceeds.
Explanation and understanding 23

our world, there is both necessity and contingency, the upshot of which
is, to use John Dewey™s language, both stability and precariousness.
Thus to add to Bohm™s potentially limitless examples, consider the
apparently non-causal concept, “copper.” Scientists think of copper as
having a host of properties, including malleability, fusibility, ductility and
electric conductivity. These properties assign powers and liabilities “ what
copper will do or undergo given certain conditions. They are promissory
notes, which may be analyzed as dispositions understood as permanent
(or relatively permanent) capacities or liabilities which exist whether or
not they are exercised and whether or not, when exercised, they are ful-
¬lled. Science does not simply ascribe causal properties to things: it ful¬lls
the promissory note by explaining them in the sense of offering an account
of the causal mechanisms which give them these properties. To be sure,
the components and modes of operation of such causal mechanisms will
differ depending upon the phenomena they explain. More importantly,
they will not, in general, be like the “mechanisms” typical of the inner
workings of an old-fashioned clock. The “mechanism” of electron trans-
fer is different from the “mechanism” by which, according to the second
law of thermodynamics, waste heat is discarded.18
Thus, theory tells us that water and salt molecules are composed of
atoms, which in turn are composed of electrons, neutrons, and below that,
of quarks and photons. At each level, the theory provides an account of the
generative mechanisms that account for the causal properties at the next
level, why in other words, they have the powers they have. As Harr´ notes,
“explanatory mechanisms become a new subject for scienti¬c study and
the explanation of their principles of operation calls for the hypothesis of
further explanatory mechanisms, new model building and so on” (1970:
262). The theory provides an understanding not only of why water turns
to steam, and why salt dissolves in it, but an understanding of all the
possible behaviors of these molecules in interaction.
As Atkins says, for chemists, the periodic table is “their single most
important concept. It summarizes the properties of the elements “ the
variation in their physical properties, such as the number and type of
bonds they form to other atoms . . . At a glance we can see whether
an element has the properties characteristic of a metal (iron), a non-
metal (sulfur), or something in between (silicon)” (Atkins, 2003: 159).
Mendeleev compiled the table empirically, that is, in accordance with the
observational properties of the elements. But, continues Atkins, “he knew

18 The point is suf¬ciently important to suggest that we abandon the term “mechanism” in
these contexts. But I can ¬nd no preferable alternative. Causal processes require causal
mechanisms. See also Machamer, Darden and Craver, 2000; Bunge, 2003.
24 A Realist Philosophy of Social Science

nothing of the structures of atoms and could have had no conception of
the underlying foundation of the table. We have that understanding. The
periodic table, as we now know, is a portrayal of the rhythms of the ¬lling
of the energy levels of atoms” (Atkins, 2003: 160). Thus, to round out
his superb account,

for hydrogen, with its single electron, all the orbitals of a given shell have exactly
the same energy. For atoms other than hydrogen . . . each shell contains orbitals
of progressively higher energy. In all cases, p-orbitals ¬rst become available in the
second shell, d-orbitals become available in the third shell and f-orbitals become
available in the fourth shell (Figure 5.7).
With two simple ideas “ that electrons organize themselves so as to achieve the
lowest possible energy, and that no more than two electrons can occupy any given
orbital “ the pattern of matter becomes understandable. (Atkins, 2003: 161)

Obviously, armed with such a powerful understanding, it will hardly
be a complicated task for the chemist to explain why salt dissolves in
water “ at whatever level of understanding one demands.
But if grasping the nature of salt is within the competence of any-
body who knows some chemistry, there are a host of other very stable
patterns which require a more complicated account. Consider our earlier
example, that robins™ eggs are greenish-blue. As stated, this is at best mis-
leading, but it can easily be rewritten to be more scienti¬cally accurate.
Thus, “Under normal conditions, a robin™s egg will appear greenish-blue
to normal percipients.” There are, plainly, a number of generative mech-
anisms at work here which, taken together, explain the generalization.
These include the well-understood biological properties which explain
the inherited traits of organisms (why robins lay eggs which produce only
robins), and the chemical and optical properties of material surfaces.
There are also the less well-understood neurophysiological properties of
the human perceptual system. But ¬nally, there are the not at all under-
stood processes which give normal percipients the experience of seeing
greenish-blue when looking at a robin™s egg. Such generalizations pre-
suppose the natural necessities of “things,” and, indeed, it is these which
make acting on our generalizations rational. If, in normal circumstances,
an object which was identi¬ed as a robin™s egg appeared white, we would
rightly be suspicious. Perhaps, after all, it was a small chicken™s egg, or
a genetic ¬‚uke? Perhaps it is only that there is an unnoticed light effect?
Since scientists are not, in general, interested in explaining patterns of
this sort “ requiring as they do an understanding of mechanisms theo-
rized in very different disciplines “ you will not ¬nd this example in any
book by a biologist, a physicist or a chemist.
Explanation and understanding 25

To round out this part of the argument: what are best termed the
“abstract sciences” aim at an understanding of the fundamental processes
of nature. Such inquiry may be motivated by discerning a pattern, but
not all patterns will be of concern. Indeed, patterns which emerge from
experimentally generated data, e.g., the results of Lavoisier™s painstaking
use of the chemical balance, are of high importance. Finally, their interest
in concrete events is also at best marginal, pretty much restricted, as we
shall subsequently suggest, to events which can provide an especially
potent test of theory. This raises a new nest of problems to consider.
2 Theory, experiment and the metaphysics
of Laplace

Chapter 1 noted that understanding came when we had a well-con¬rmed
theory about a generative mechanism. In this chapter we consider the
essentials of construction and con¬rmation of theory, including the role
of experiment in those sciences where experiment is possible. While the
literature on theory is both vast and contentious, we can here be relatively
brief. Our aim is to focus on what is absolutely essential for the purposes
of a philosophy of the social sciences. More important for us is the effort
in this chapter to undermine the bad metaphysics of what is too often
taken for granted in talk about theory and the goals of science. A key
theme will be to show that celestial mechanics is a very poor example for
the sciences, social and physical.

What a theory is
Despite much talk to the contrary, no real theory in the physical sciences
can be fully expressed as a deductive system, with axioms and deductions
therefrom. The idea has a long legacy dating at least from Descartes, from
Newton™s great work, and from the still older idea that mathematics is
the ideal of knowledge (Harr´ , 1970: 8). As Harr´ says:
e e

In fact, in actual science, deductive systems are quite rare: fragments of such
systems can be found in physics, but mostly scientists come up with descriptions of
structures, attributions of powers and laws of change, related by having a common
object, not being then and there deducible from a common set of axioms. (Harr´ , e
1970: 10)

For some theories “ though surely not all “ mathematics will play a crit-
ical role, especially in developing the abstractions of system dynamics.
But expressing laws and descriptions of objects in mathematical style
does not make them mathematical propositions: their meaning remains

Theory, experiment and the metaphysics of Laplace 27

non-mathematical “ even where we resist relating the mathematics to
visualizable models. Indeed, Atkins concludes, reluctantly, that “any
¬nal theory, if there is one, is likely to be a purely abstract account
of the fundamental structure of the world, an account that we might
possess but not comprehend” (2003: 358). This is, he notes, “probably
too extreme a view.” He continues: “Humans are adept at interpret-
ing mathematics, particularly the mathematics used to support physics,
in homely terms, aware all the time that their interpretation is fraught
with danger and incompleteness, but interpreting nonetheless” (Atkins,
2003: 358).
Harr´ ™s de¬nition of theory acknowledges the necessity of
interpretation: “A theory consists of a representation of the structure
of the enduring system in which those events occur which as phenom-
ena are its subject matter, and by which they are generated” (1970:
14). As already insisted, theories supply “an account of the constitution
and behavior of those things whose interactions with each other are
responsible for the manifested patterns of behavior” (Harr´ , 1970: 35).
They identify “things” “ molecules and atoms, for example, how they are
structured, and how they interact. They are, of course, representations,
but they are meant to represent reality “ as it is in-itself.
Following Harr´ (1970), it is convenient to think of this representation
as involving hypotheses of several types, (1) existential: “atoms exist;”
(2) a model description: “molecules are in random motion;” (3) causal
hypotheses: “pressure is caused by the impact of molecules;” (4) modal
transforms: “temperature is another way of conceiving of mean kinetic
energy.” The examples taken from Atkins (above) were meant to give a
hint of this (see also Machamer et al., 2000).

Realist versus instrumentalist conceptions
of theoretical terms
As a consequence of empiricist commitments, beginning at least with
Mach but extending through the heyday of logical empiricism, there has
been much debate among philosophers about the function of theoretical
terms. For the dominating empiricist view, in contrast to the realist view
brie¬‚y summarized above, theoretical terms can function without making
“existential” commitments. Thus, the meaning and application of terms
like “electron” are fully given by means of “reduction sentences,” of which
“operational de¬nitions” are a key variety, or by means of “correspon-
dence rules,” which more indirectly link the theoretical term (T-term) to
terms in the observation language (O-terms). Thus:
28 A Realist Philosophy of Social Science

“X has theoretical property T,” means “if X is placed under test conditions C,
then the test yields observable results O.”1

There is hardly a textbook in quantitative methods in the social sciences
that does not repeat some version of this. Compare the example in the
previous chapter from Frankfort-Nachmias and Nachmias (1992). They
Often the empirical attributes or events that are represented by concepts cannot
be observed directly . . . In such cases, the empirical existence of a concept (sic)
has to be inferred. Inferences of this kind are made with operational de¬nitions.
(1992: 31)

When we refer to T, we mean: “If C, then O.” So “T” has been
“reduced”: for all practical purposes, it has been eliminated. By 1958,
Hempel saw that there was a serious problem with an account of theory
in which no existential commitments were made. He remarked: “The use
of theoretical terms in science gives rise to a perplexing problem: Why
should science resort to the assumption of hypothetical entities when it is
interested in establishing predictive and explanatory connections among
observables” (Hempel 1965: 179). Indeed, if this is their purpose, the
problem can be expressed as a dilemma:
1. Either theoretical terms serve their purpose or they do not.
2. If they serve their purpose, since they establish predictive and explana-
tory connections among observables, they are unnecessary.
3. If they do not serve this purpose, they are surely unnecessary.
4. Hence, theoretical terms are unnecessary.
Indeed, if their meaning and application can be given by sentences in the
O-language, they are but handy place-markers for organizing experimen-
tal data. Hence 2. But perhaps this is not at all their purpose? Hempel
came to see that theoretical terms serve another and more critical pur-
pose. As he said:
When a scientist introduces theoretical entities such as electric currents, magnetic
¬elds, chemical valences, or subconscious mechanisms, he intends them to serve
as explanatory factors which have an existence independent of the observable
symptoms by which they manifest themselves. (Hempel 1965: 205)

Indeed! As argued in chapter 1, appeal to these ideas can explain exactly
because they are taken to represent the generative mechanisms which

1 As in chapter 1, this is also the standard empiricist explication of dispositional terms, like
water-soluble. So “X is water-soluble” means “if X is put in water, X dissolves.” For the
realist, roughly, “X is water- soluble” means “there is something about X and water, such
that if X is put in water, it dissolves.” For the realist, water-soluble refers to the powers
and tendencies of X and of water. And theory provides an account of these.
Theory, experiment and the metaphysics of Laplace 29

produce the pertinent observables. All useful theories make inescapable
ontological commitments. But doesn™t this introduce metaphysical spec-
ulation into science? How does it remain empirical?

Post-Kuhnian grounds for establishing
scienti¬c consensus
Thomas S. Kuhn™s The Structure of Scienti¬c Revolutions (1962) caused
an enormous stir. Many came to the conclusion that science was not the
rational enterprise it was thought to be. But this was the wrong conclu-
sion to be drawn from his work. Rather, along with several others,2 he
showed that it was not rational in the sense that the logical empiricists
had held it to be rational. Today, there would be agreement that there
is no “theory-neutral” observation language that could serve as a foun-
dation for truth claims, that the logic of con¬rmation or of falsi¬cation
fails, and the historical and social environment of scienti¬c practice is
the key to understanding scienti¬c success. It was in terms of these fac-
tors that at some point a consensus regarding a theory emerged in the
scienti¬c community. To be sure, this opened the door for the idea that
science was no more rational than any other practice, for it raised the
question of what brings the scienti¬c community to this consensus “ of
agreeing that hypotheses, existential, descriptive and causal, are true? It
is safe to say, perhaps, that while few writers remain holdouts to what
is sometimes termed empiricist “foundationalism,”3 few writers would
now argue that science is just another practice that gives us no special
access to knowledge about the world. In what follows, a third alternative
is sketched.
The problem begins by acknowledging that all we can do as human
inquirers is to represent the world. As would now also be generally
acknowledged, since there is no “theory-” or “concept-neutral” way to
do this, we can never be sure that our representations truly represent the
world the way it is in-itself.4 The problem begins with our ordinary expe-
rience and runs straight through to sophisticated scienti¬c theory. Put in
2 Included here are Quine, 1961; Hanson, 1958; Toulmin, 1953, 1961; Sellars, 1963;
Feyerabend, 1975. For an excellent collection of essays, see Suppe, 1977. A still very
useful account is Brown, 1977.
3 Empiricist foundationalism assumes a “theory neutral” observation language which
anchors all theory. It is in terms of it that theory is tested, con¬rmed or falsi¬ed. See
4 In its modern form, this is the problem raised by the philosopher Immanuel Kant who
distinguished between “things as experienced” (phenomena) and “things-in-themselves.”
Kant argued that the latter were unknowable, but rescued scienti¬c objectivity by hold-
ing that the categories of mind were universal, the same in all “rational beings.” For a
perceptive view of the history of modern epistemology, see Rorty, 1981.
30 A Realist Philosophy of Social Science

other terms, we cannot step out of our history and have a God™s-eye view
of the world.5
This entails an intractable fallibilism: no truth claim is certain and all
are revisable in the light of new experience and new theories. But we need
not abandon the idea of truth. The products of scienti¬c practice vindicate
the physical sciences as the preferred means of ¬xing belief about the
world.6 Like juries, scienti¬c communities come to agreement, but each
member of the community is constrained by historically generated values,
goals and practices which, as a community, they accept. Let us not forget,
the world, as it is in-itself, remains the most critical constraint. No set of
beliefs will allow humans to ¬‚y like birds, to stay healthy on an exclusive
diet of cheeseburgers, to build a perpetual motion machine, to pollute
the air and the earth, and maintain, inde¬nitely, an environment suitable
for human life.
Consider a parable. It is possible that a society could come to believe
that women should be forbidden from eating bananas because they believe
that doing so will undermine their reproductive capacities. It also may
be that this belief is well supported by convictions arising from their
creation story and other ongoing practices in their everyday life. We say
that bananas will not have ill-effects on female reproduction. Indeed, we
say that they are quite nutritious. Who is right? Or perhaps both are right?
In our culture we accept the idea of science “ even if we are not always
clear why we should. We might say to our new-found friend that he should
allow females to eat bananas and see who is right. Likely, he will not, but
even if he did and the female who consumed bananas continued to be
fertile, he could, we can be sure, explain this outcome in terms of inter-
vention by one of his well-respected gods. And he could insist, consistent
5 Kuhn™s concern was wholly the representations of scienti¬c theory, but his line of argu-
ment conjoined with other so-called, post-modern epistemologies. Thus it is now argued
that, for example, women, the colonized or indigenous people have a distinct perspective
or framework of meaning and experience, and that these are privileged, or at least that
some scienti¬c perspective is not privileged. We will return to this idea in chapter 3.
6 The expression, “scienti¬c practice” “ like “scienti¬c method” “ is both highly abstract
and crude. Both are learned from those skilled in the practice of the sciences, as appren-
tices, neither may be articulated in any sort of clear fashion and, as the text from Einstein
quoted when we began says, when articulated “ especially by philosophers and textbook
authors “ they are all too often distortions of real practice. Since Kuhn™s book, a host of
literature shows this to be the case. See, for example, Latour et al., 1979, 1987; Pickering
(ed.), 1992; Knorr-Cetina, 1981, 1999; Hacking, 1983, 1992, 2000. For a defense of the
so-called “strong programme” in the sociology of knowledge against the charge that it
is guilty of vicious epistemological relativism, see Manicas and Rosenberg, 1985, 1988.
It is also the case that when good scienti¬c practice is violated we get “shoddy science”,
“reckless science” and “dirty science.” The best account remains Jerome Ravetz, 1971.
Indeed, given the development of “big science,” it has been increasingly dif¬cult to sus-
tain “good scienti¬c practice.” See, for example, Richard Lewontin™s review (2004) of
two pertinent recent accounts.
Theory, experiment and the metaphysics of Laplace 31

with our logic, that there remain scores of yet untested cases. How do we
know what will happen in these? We feel con¬dent that bananas are nutri-
tious and good for both females and males, in part because we think that
his beliefs about creation and all the kapus which are legitimated by it are
wrong “ interesting perhaps, but not plausible. Moreover, and much more
important epistemologically, although our human practices are socially
constructed, bananas are not “ even if the meanings attached to them in
social interaction are. Bananas exist and they exist independently of our
beliefs about them. We know that they are wrong about them because we
understand why bananas are nutritious. We can produce well-established
theories about reproduction, health and the bio-chemical properties of
Of course, our kinsmen may not be convinced. Moreover, one might
argue that their belief system is to be preferred. Perhaps it has allowed
them to reproduce a pleasant and just life. (Their women might not
agree!) So we might not wish to interfere with our “scienti¬c” ideas and
we may hope that they might be able to preserve the way of life that they
prefer. Indeed, it is a huge error to suppose that just because we believe
that some claim is true, coercive intervention can be justi¬ed.7 This is all
quite consistent with saying that they are wrong about bananas. Indeed,
we can now offer technologies based on this knowledge which can prevent
ill-health, extend mortality and increase fertility. This is, of course, the
second obvious argument in favor of believing the claims of science. We
are today surrounded by technologies made possible only through the
understanding provided by the physical sciences “ for good or for ill.
In the foregoing, the idea that we could test the hypothesis that bananas
were good for females as well as for males was appealed to uncritically. We
need, straightaway, to reject two claims about scienti¬c method that are
dead ends. One is the inductivist assumption that one con¬rms a hypoth-
esis by piling up cases. This has plausibility as regards hypotheses like the
one in our example “ even if, as we noted, our kinsman has our logic on
his side. Not only is it the case that for any ¬nite number of instances,
there will always be more instances not yet tested, but we need to be con-
¬dent that the sample is apt “ an enormous problem which should not
be underestimated. For example, inferring ancestry from gene markers
of sampled populations is fraught with dif¬culty (chapter 3). The other
error, promoted by Karl Popper, is the idea that (versus the inductivists),
since we cannot positively con¬rm hypotheses, we can falsify them. No

7 Which is not to say either that coercive intervention is never justi¬ed. For some discussion
of a very large and dif¬cult topic, see, for example, essays by Seyla Benhabib, Martha
Nussbaum and Jonathan Glover in Nussbaum and Glover, 1995.
32 A Realist Philosophy of Social Science

fallacy will be committed here. Hypotheses which resist falsi¬cation then
are accepted.
Popper was interested in demarcating science from non-science and
his main interest was in showing that some hypotheses resist falsi¬cation
in principle, for example, that God is good, and hence, that they cannot
be scienti¬c. This remains a viable way to distinguish science from non-
science. But the idea that one can escape fallacy by seeking to falsify
hypotheses will not do either: the effort to falsify any hypotheses always
involves auxiliary hypotheses and thus, as a matter of formal logic, the
test is no more conclusive when it is negative then when it is positive. If T
is the theory, A are auxiliary hypotheses needed to make the test, and O
are observations, this is the logic of the situation: The argument on the

Con¬rmation Falsi¬cation

If T (and A) then O1 If T (and A), then O2

. 1
( 7)