Full Download The Philosophy of Science: A Companion Anouk Barberousse PDF
Full Download The Philosophy of Science: A Companion Anouk Barberousse PDF
Full Download The Philosophy of Science: A Companion Anouk Barberousse PDF
com
https://ebookmass.com/product/the-philosophy-
of-science-a-companion-anouk-barberousse/
ebookmass.com
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
https://ebookmass.com/product/science-and-humanity-a-humane-
philosophy-of-science-and-religion-1st-edition-andrew-steane/
https://ebookmass.com/product/the-routledge-companion-to-gender-
and-science-fiction-1st-edition-sonja-fritzsche/
https://ebookmass.com/product/bayesian-philosophy-of-science-
variations-on-a-theme-by-the-reverend-thomas-bayes-1st-edition-
stephan-hartmann/
https://ebookmass.com/product/the-oxford-handbook-of-philosophy-
of-political-science-harold-kincaid/
The Philosophy and Science of Language Ryan M. Nefdt
https://ebookmass.com/product/the-philosophy-and-science-of-
language-ryan-m-nefdt/
https://ebookmass.com/product/the-philosophy-of-science-
institute-lectures-st-johns-university-studies/
https://ebookmass.com/product/a-companion-to-the-russian-
revolution-1st-edition-edition-daniel-orlovsky/
https://ebookmass.com/product/a-companion-to-aeschylus-peter-
burian/
https://ebookmass.com/product/language-science-and-structure-a-
journey-into-the-philosophy-of-linguistics-ryan-m-nefdt/
The Philosophy of Science
OXFORD STUDIES IN PHILOSOPHY OF SCIENCE
General Editor:
Paul Humphreys, University of Virginia
Advisory Board
Anouk Barberousse (European Editor)
Robert W. Batterman
Jeremy Butterfield
Peter Galison
Philip Kitcher
Margaret Morrison
James Woodward
A C O M PA N I O N
Edited by
Anouk Barberousse, Denis Bonnay, and Mikaël Cozic
1
1
Oxford University Press is a department of the University of Oxford. It furthers
the University’s objective of excellence in research, scholarship, and education
by publishing worldwide. Oxford is a registered trade mark of Oxford University
Press in the UK and certain other countries.
9 8 7 6 5 4 3 2 1
Printed by Sheridan Books, Inc., United States of America
Contents
Preface vii
Acknowledgments xiii
About the Contributors xv
3. Causality—Max Kistler 95
v
vi Contents
16. Philosophy of Cognitive Science—Daniel Andler 595
Index 727
Preface
General Introduction
Philosophy of science has the aim of answering those questions raised by scientific ac-
tivity that are not directly addressed by science itself. Among such questions, we can
mention: What are the overall goals of science, as well as the specific goals of its var-
ious branches? By what means are these goals pursued? What basic principles does it
put into practice? Philosophy of science also tries to understand the relationships that
exist between the scientific disciplines. To what extent, and in what sense, are they,
and should they be, unified? Also belonging to its domain is the relationship between
science and reality. What, if anything, does science tell us about reality? And to what
extent is it justified in making the claims it does?
Just like the sciences themselves, current philosophy of science is multifaceted and
specialized. A philosopher of science may embark on projects as diverse as the develop-
ment of a formal analysis of the concept of confirmation using probability theory and
the study of the potential contribution neuroscience may bring to our understanding
of consciousness. Thus, it becomes difficult for both students and researchers within
a given domain to be aware of the advances and challenges arising in any specific area
in philosophy of science.
The aim of the present book is to expose the main questions, as well as some of the
answers, being discussed in today’s philosophy of science. We view it as the “missing
link” between introductions and research, and our own goals will have been met if this
book successfully bridges the gap between introductions to the philosophy of science
meant for a general audience on the one hand, and research articles and monographs
vii
viii Preface
on the other. It is therefore primarily intended for the use of advanced undergrad-
uate or graduate students who, after a first introduction to the area, may now wish to
deepen their knowledge. We also hope that The Philosophy of Science: A Companion will
be useful to both junior and senior researchers in philosophy of science wishing to fa-
miliarize themselves with areas outside of their own.
Philosophy of science has become too specialized for this goal to be achieved by any
one person. Thus, our book is a collective effort. We have nevertheless endeavored to
present the basic problems that shape contemporary philosophy of science in a co-
herent way. In contrast with encyclopedias, where contributions tend to simply coexist
and thus lack organic unity, we have tried to maximize complementarity and cross-
referencing between the chapters. Our hope is that this has favored a strong sense of
unity, something that is always hard to attain in such collective undertakings.
The two parts of The Philosophy of Science mirror the traditional distinction between
general philosophy of science and philosophy of the special sciences. General Philosophy
of Science (Part I) deals with generic issues raised by scientific activity, independent
of specific disciplines. General philosophy of science was the very core of philosophy
of science up to the middle of the twentieth century. Philosophy of science itself has
dramatically evolved over the last several decades, becoming increasingly devoted to
issues raised by specific scientific disciplines. The study of general problems never-
theless remains a highly active element of philosophy of science. Moreover, it is our
opinion that the study of these general problems is indispensable to those who focus
on the philosophy of some particular scientific discipline or area, since they represent
a set of tools invaluable to understanding their own, specific objects of study.
The objective of the first part of the book is twofold. We intend to both take stock
of the traditional questions which have shaped analytic philosophy of science and to
introduce certain problems that have been raised more recently. Thus the first two
chapters, bearing upon explanation and confirmation, respectively, tackle issues that
were the subject of intense debate in the middle of the twentieth century—notably
among philosophers of science influenced by logical empiricism—and which, as we
shall see, are still much studied today. With causality, c hapter 3 also focuses on a tra-
ditional concept, though one to which logical empiricism has been rather hostile.
Causality is now at the epicenter of a very vibrant area, straddling the borders of phi-
losophy of science and metaphysics. Metaphysics is also at the heart of c hapter 4, which
deals with scientific realism (an issue that underwent a thorough overhaul during the
1980s) and the metaphysics of science, constituting a topic that is much discussed
today. Chapter 5 addresses the issue of knowing how best to analyze some of science’s
primary products, namely theories and models. Starting from the “received view” of
scientific theories, inherited from logical empiricism, it discusses the objections that
have been raised against this view while also looking at alternative conceptions. Lastly,
chapter 8 deals with issues surrounding the reduction and emergence of properties
Preface ix
and/or theories coming from distinct scientific disciplines. Logical empiricism also
contributed greatly to this research area. We shall see that current reflection on the
matter is closely related to metaphysics, philosophy of knowledge, and sometimes also
to the philosophy of the special sciences (particularly the philosophy of mind).
In our view, these six topics—explanation, confirmation, causality, scientific re-
alism, the nature of theories and models, and reduction—constitute the core of gen
eral philosophy of science, even if they do not exhaust it. This latter consideration
in mind, two further issues are also touched on in Part I. Chapter 6 studies the di-
achronic dimensions of scientific activity, a topic made famous by Kuhn’s much cel-
ebrated book (The Structure of Scientific Revolutions, 1962/1970). Chapter 7 is more
meta-philosophical in character: it reviews the relations between philosophy of science
and other approaches (notably historical and sociological) which share in the aim of
analyzing scientific activity and which are currently referred to as sciences studies.
Although comprehensive, this does not cover all topics having a justifiable claim to the
label of general philosophy of science. For instance, the growing literature on statistics
and statistical reasoning is not represented. But it is our contention that Part I of The
Philosophy of Science will provide the reader with a satisfyingly complete survey of con-
temporary general philosophy of science.
For several decades, philosophers of science have increasingly directed their attention
toward the finer details of scientific activity, in particular to issues exclusive to specific
disciplines. These issues are the object of the philosophy of the special sciences, to
which the second part of The Philosophy of Science is devoted.
Compared with general philosophy of science, philosophy of the special sciences
appears two-sided. Certain problems are essentially instances or applications of
issues belonging to general philosophy of science. In this case, more often than not,
the targeted area of knowledge requires some reconsideration of the issue on the
part of the philosopher. For instance, the issue of justification or confirmation of
theories raises specific problems when one studies, let’s say, economic or mathemat-
ical theories, as opposed to theories from physics, which often serve to illustrate con-
firmation theories. By contrast, certain other issues in the philosophy of the special
sciences are entirely generated by the specific concepts and methods of a given field.
The discussions on the concept of function (in biology) or on the nature of linguistic
universals (in linguistics) are two cases in point. The main objective of the second part
of this volume is to introduce the reader to a representative sample of the issues that
currently structure the philosophy of the special sciences. We have done our best to
respect this two-sided character, i.e., to show how some of the issues are very closely
linked to the “big” issues in general philosophy of science while others are specific to
certain specialized domains of science.
The first two chapters of Part II are devoted to the philosophy of the formal sciences.
More precisely, chapter 9 is concerned with logic and chapter 10 with mathematics. The
x Preface
philosophy of the formal sciences has often been left out of handbooks or textbooks
on the philosophy of science. One of the reasons that implicitly underpins this state of
affairs is that the issues raised by these formal sciences can seem remote from those
raised by bona fide empirical sciences. But there are other reasons that speak in favor
of integrating philosophical discussion on these disciplines. First, there is some inter-
esting convergence between certain issues in the philosophy of the formal sciences
and other issues in general philosophy of science, for example, those related to the
nature of explanation. Second, there are certain other issues which call for a unified
and coordinated answer from both the philosophy of the formal sciences and other
branches within philosophy of science. For example, understanding why mathematics
fits into the physical world so well—an issue that lies at the border between the philos-
ophy of mathematics and the philosophy of physics. Or the problem of understanding
mathematical cognition, which is of interest to both philosophy of mathematics and
cognitive science.
Chapters 11 and 12 are devoted to the philosophy of physics and the philosophy of
biology, respectively. These two areas have a special status in philosophy of science.
Philosophy of physics is considered basic because physics is viewed as the fundamental
scientific discipline. This means at least two things. First, that physics is an area where
scientific reasoning is supposed to reach its zenith, and thus, in particular, that it is
indispensable to be at least minimally familiar with it if one wishes to gain an under-
standing of scientific reasoning in general. And, second, that it is crucial to clarify the
picture of the world as it is depicted by the physical sciences. Philosophy of biology
has become an extremely active field, such that there is probably no other area in the
philosophy of the special sciences whose importance has grown more over the last two
decades.
An entire chapter is devoted to the philosophy of medicine. Our main reason for this
is that philosophy of medicine is an area where philosophy of science overlaps with
normative and practical philosophy. This reveals itself with respect to the question of
whether the concepts of health and illness have an essential normative dimension,
and also as regards the study of clinical reasoning. In both cases, the discussion goes
beyond the purely epistemic point of view dominant in the philosophy of the natural
sciences.
Another particular feature of Part II is the space we have devoted to philosophy
of the human and social sciences (chapters 14 to 17). Interestingly, in these areas
the philosopher’s stance and corresponding expectations may differ from those that
are generally endorsed in the philosophy of the natural sciences. In the former area,
philosophers often assume that there is nothing wrong with the way science is done
and thus refrain from making recommendations to scientists or from criticizing their
methods. Not so in the latter case, and this is to be expected, since there are far more
methodological uncertainties, debates, and disagreements involved in the human and
social sciences.
Chapters 14 and 15 broach the social sciences. Chapter 14 deals with general issues
in the philosophy of the social sciences, for example, methodological individualism
Preface xi
and the relations between social sciences and cognitive sciences. Chapter 15 focuses
on one specific social science, economics. This emphasis is to be welcomed, in light
of the scientific and social impact of economics, and all the more so since it currently
constitutes a particularly active field of study for philosophers.
The last two chapters are organized in a similar way. Both are devoted to disciplines
that study human cognition. Chapter 16 is a general presentation of the issues raised
by cognitive science from the point of view of philosophy of science. Chapter 17, on the
other hand, bears on one specific discipline—linguistics. While philosophy of language
is a well-structured and well-known area in philosophy, there are relatively few phil-
osophical discussions on linguistics as a science. Both for this reason and for the fact
that the philosophy of cognitive science focuses more on disciplines like psychology
and neuroscience, we deemed it fitting to devote a whole chapter to linguistics.
Acknowledgments
We are grateful to the contributors to this volume, to Thierry Martin (the editor of
the series in which an earlier version of The Philosophy of Science was published, under
the title “Précis de Philosophie des Sciences” in 2011), to Daniel Andler who brought
financial support to this initial version through his Senior Fellowship of the Institut
Universitaire de France and to Christopher Robertson, who translated many of the
chapters’ earlier versions. The current version has benefited from the comments of two
anonymous referees. It was notably supported by the Institute of Cognitive Studies at
Ecole Normale Supérieure (Paris) under grant ANR-10-L ABEX-0087 IEC and by Mikaël
Cozic’s Junior Fellowship of the Institut Universitaire de France. Lastly, we wish to
thank the Institut d’histoire et de philosophie des sciences et des techniques (UMR
8590, Paris I—ENS Ulm—CNRS), which has provided us with a highly stimulating sci-
entific environment for 15 years.
Anouk Barberousse, Denis Bonnay, and Mikaël Cozic, Paris, January 2018.
xiii
About the Contributors
xv
xvi About the Contributors
formal theories of rationality. His current research concerns the relationship between
cognitive science and positive and normative economics, as well as several issues
in Bayesian epistemology, including the revision of one’s beliefs upon learning the
opinion of others.
Jacques Dubucs is a senior scientist at the Centre National de la Recherche Scientifique
and the head of the Social Sciences and Humanities Department at the French Ministry
of Higher Education, Research, and Innovation. His scientific work deals with logic
and philosophy of science.
Paul Égré (born 1975; PhD, 2004) is directeur de recherche at Institut Jean-Nicod
(CNRS) and an associate professor in the Philosophy Department of Ecole Normale
Supérieure in Paris. Besides work in formal semantics and on the epistemology of lin-
guistic theory, a large part of Paul Egré’s work over the last decade has been on the
topic of vagueness in language and in perception, dealing with semantic, logical, and
psychological aspects of the phenomenon. Since 2012, Egré is also the editor-in-chief
of the Review of Philosophy and Psychology.
Jon Elster is the Robert K. Merton Professor of Social Science at Columbia University.
He is the author or editor of more than thirty-five books translated into more than
seventeen languages on the philosophy of social sciences, the theory of rational choice,
political psychology, deliberative democracy, and the history of political thought (Marx
and Tocqueville), to name a few of their subjects. He is currently working on a compar-
ative study of the Federal Convention (1787) and the first French constituent assembly
(1789–1791).
Michael Esfeld is full professor of science at the University of Lausanne. His research
is in the metaphysics of science, the philosophy of physics, and the philosophy of
mind. His latest book publication is A Minimalist Ontology of the Natural World, with
Dirk-André Deckert (New York: Routledge, 2017).
Élodie Giroux is an assistant professor at Jean Moulin Lyon 3 University, where
she teaches philosophy of science and philosophy of medicine. She is director of the
master’s in “Culture and Health.” Her main research interests are the history and
epistemology of “risk factor epidemiology”; causation in medicine and public health;
and risk, health, and disease concepts. She is currently working on precision medi-
cine. Besides several papers on modern epidemiology, she published Après Canguilhem,
définir la santé et la maladie (Paris: PUF, 2010) and Naturalism in the Philosophy of Health
(Cham: Springer, 2016), and she edited a special issue on the history of risk factor ep-
idemiology in Revue d’Histoire des Sciences (2011) and on precision medicine in Lato
Sensu (2018).
Max Kistler is professor at the Department of Philosophy at University Paris 1
Panthéon–Sorbonne and head of IHPST (Institut d’Histoire et de Philosophie des
Sciences et des Techniques). His research topics include causation, powers and
dispositions, laws of nature, natural kinds, and reduction. He is the author of Causation
About the Contributors xvii
and Laws of Nature (Routledge, 2006), L’esprit matériel. Réduction et émergence (Ithaque,
2016), and coeditor (with B. Gnassounou) of Dispositions and Causal Powers (Ashgate,
2007).
Hélène Landemore is an associate professor of political science at Yale University. She
is a political theorist interested in democratic theory, theories of justice, Enlightenment
thinkers, and the philosophy of social sciences. Her book Democratic Reason (Princteon,
NJ: Princeton University Press, 2013) was awarded the 2015 David and Elaine Spitz
Prize for best book in liberal and/or democratic theory published two years earlier. She
is currently writing a new book on postrepresentative or “open” democracy.
Maël Lemoine is a professor at the University of Bordeaux, France, where he teaches
philosophy of medical science. He published an introductory essay in the philosophy
of medical science in 2017 and has recently published various articles on biological re-
search in psychiatry, animal models, and precision medicine.
Pascal Ludwig is an associate professor in the Department of Philosophy, Sorbonne
Université, Paris. He has coauthered several books on the philosophy of science and
the philosophy of the mind.
Thomas Pradeu is a CNRS senior investigator in philosophy of science (permanent
position) at ImmunoConcept (CNRS and University of Bordeaux), and associated
member at the Institut d’Histoire et des Philosophie des Sciences et des Techniques
(CNRS and University Pantheon–Sorbonne). His research focuses on biological indi-
viduality, immunology, the microbiota, and the interactions between philosophy and
science.
Philippe de Rouilhan is a senior researcher emeritus at the CNRS and a member of
the Institut d’Histoire et de Philosophie des Sciences et des Techniques (CNRS and
Université Panthéon–Sorbonne), of which he was the director for a long time. His work
pertains to logic lato sensu or, more specifically, to formal ontology, formal semantics,
philosophy of logic, philosophy of mathematics, and philosophy of language. He is cur-
rently preparing a book on truth, logical consequence, and logical universalism.
Marion Vorms is a lecturer (maître de conférences) in philosophy at University Paris 1
Panthéon–Sorbonne and a Marie Curie fellow at Birkbeck College, London, psychology
department. Her past work in philosophy of science concerns the nature of scientific
theories and representations. Her new project, which is at the crossroads of episte-
mology and the psychology of reasoning, bears on the notion of reasonable doubt; she
is particularly interested in judicial reasoning and decision-making.
1
Why is Nicolas angry? Because he thinks Dominique wanted to play a nasty trick on
him. Why was Gomorrah destroyed? Because God wanted to punish its inhabitants.
Why did the dinosaurs disappear? Because a giant asteroid crashed into the earth.
In asking the question “why?” we bring a real or reputed fact—Nicolas’s anger, the
destruction of Gomorrah, dinosaur extinction—to the attention of our interlocutor,
and we ask for an explanation of that fact. These explanations may rely on simple eve-
ryday knowledge: it is well known that people do not like having nasty tricks played on
them. Explanations can be of the religious sort: the biblical account tells not only of
Gomorrah’s existence but also of the sins of its people, going on to explain the destruc-
tion of the city by an act of divine retribution. And then there are the explanations
offered to us by science: thus, the extinction of the dinosaurs being one of the enigmas
that paleontology faces, an asteroid strike is one of the explanations put forward.1
More than just a simple side issue of scientific activity, explanation takes its place as
one of the specific goals of science. Of course, as we have just seen, it is not just science
that claims to offer explanations. And, conversely, science certainly has goals other
than explanation too. Science enables us to describe and classify phenomena, as well as
1
I thank Anouk Barberousse, Mikaël Cozic, Henri Galinon, Marion Vorms, and Kenneth Waters for var-
ious discussions, comments, and re-readings, which were of help to me. I also wish to thank Christopher
Robertson, who translated the French version. This work received funding from the ANR (The IHPST’s
Logiscience program) and from the Institut de Recherches Philosophiques (Université Paris Nanterre).
The survey on theories of explanation is also obviously indebted to some other, similar enterprises, in
particular the surveys by Salmon (1989) and Woodward (2009).
3
4 The Philosophy of Science
enabling us to predict and control them. Nevertheless, one of the motivations, be they
individual or collective, to “do science” in the first place seems to be to find explanations
that cannot be found elsewhere—for example, research on electricity and magnetism,
and also work on the electromagnetic theory, that is developed to explain a group of
mysterious phenomena such as static electricity, the properties of Magnesia stones,
or lightning and its effects. In contrast, it is not easy to imagine what sort of thing a
scientific theory that explained nothing would be. A strict typology, say a botanical clas-
sification of different plant species according to their phenotype for example, doesn’t
strike us as being a bona fide scientific theory, insofar as it lacks any explanatory power.
Not lacking, however, are opponents to the idea that the aim of science is to pro-
vide explanations. Pierre Duhem, in The Aim and Structure of Physical Theory, opposes the
idea that the object of a scientific theory is to explain a set of observable regularities, an
opinion shared by other physicists of his time such as Ernst Mach. But this refusal is pri-
marily grounded in Duhem’s own concept of explanation. To explain would be “to strip
reality of the appearances that envelop it like a veil, in order to see the bare reality itself”
(Duhem, 1908); Duhem considers that attaching an explanatory ambition to science
makes it subservient to metaphysics, the only domain to claim possession of the keys
to the ultimate essence of things.2 The approach that we will follow here is not quite the
same. In determining whether science provides explanations or not, we will not start out
with some overly demanding concept of explanation. We will set out from the intuition
that science provides explanations, and we will try to identify a concept of explanation
such that this concept would enable us to account for the explanatory power of science.
What can be expected from this line of enquiry? What goals are we pursuing? In a
good concept of explanation, we expect first of all that it be adequate; that is, that it will
allow us to understand which elements provided by science constitute explanations
and by what virtue they come to possess their explanatory power. For example, if
an explanation has some epistemological virtue, in that it allows us to “understand
what is happening,” then a good concept of explanation must tell us how scientific
explanations allow us to “understand what is happening.” We would hope then, off
the back of this, to be in a position to evaluate explanations, that is to say, to have the
capacity to distinguish between good and bad explanations. An analysis of the concept
of explanation will obviously not tell us if the explanation is right, in the sense of its
expressing truth, but it should be able to tell, or at least indicate to us, whether some
explanation would be a good explanation, presuming that it does express the truth.
And lastly, we would like some insight regarding the relationship between the explan-
atory aim of science and its other aims—prediction, control, and so on.
We will begin by looking in detail, during the first section, at the theory of scientific
explanation proposed by Hempel and Oppenheim known as the deductive-nomological
model (DN). The importance of place we give it here is justified conceptually by the
rigor of the analysis it proposes and historically by the role of cardinal reference it
2
On the question of realism—does science give us access to the very nature of things or not?—and on
the metaphysical scope of science, see chapter 4 of the present volume.
Scientific Explanation 5
continues to play in contemporary debates on explanation, despite its no longer being
the dominant model. In the second section, and in light of the DN model, we will re-
visit the general properties of explanation, discussing the link between explanation
and prediction, the temporal conditions that weigh, or do not weigh, on explanation,
as well as the characterization of the laws of nature. The third section is devoted to an
examination of the classic objections brought against the DN model, these taking the
form of a list of counter-examples. The main rival theories that have emerged to resolve
these problematic examples in the DN model’s stead—causal theory and unificationist
theory—are presented and discussed in the fourth section. In the closing section, we
will sketch out some other approaches toward contemporary reflection on explanation.
Let us begin then by looking at the inaugural example given by Hempel and Oppenheim
(1948). A mercury thermometer is rapidly immersed in a basin of hot water. The level
of the mercury column falls slightly at first before rising swiftly. Why? Here we have
a little puzzle to solve—we were expecting that the level of the mercury would simply
rise, though this is not exactly what has happened. In fact the explanation is quite
simple. The rise in temperature, at first, affects only the standard quality glass tube
which contains the mercury. Expanding, the tube leaves more room for the mercury,
whose level promptly drops. Then, rapidly, the heat spreads out and the mercury
expands in turn. As its coefficient of expansion is much higher than that of glass, the
mercury level rises and exceeds its own initial level.
Analyzing this example makes the distinction between the explanandum, what is
to be explained, namely the slight decrease followed by rapid rise in the level of the
mercury, and the explanans, which does the explaining, immediately clear. Under ex-
planans we see, first, the initial conditions, the particular facts reported in the expla-
nation, such as the set-up involved—the glass tube, the mercury column, the bowl
of hot water—and the act of immersing the tube in hot water itself. Then too, the
general laws come into effect, such as the laws governing the thermal expansion of
glass and mercury, and a statement regarding the relatively low thermal conductivity
of glass. The explanandum is subsumed under the general laws, in the sense that it can
be deduced from these laws and the initial conditions.
Hempel and Oppenheim’s theory is that the full generality of scientific explanation
can be read in this particular case. To explain, one need not do anything other than de-
duce the phenomenon to be explained by using general laws and the initial conditions,
which justifies the labeling of their model as the deductive-nomological (DN) model
of explanation. Thus, the general form for scientific explanation that we draw from
Hempel and Oppenheim is as follows:3
The double-lined bar ==== indicates that the statement below follows on logically from those statements
3
above it.
6 The Philosophy of Science
For there to be explanation, certain conditions must be met by the explanans and
by the explanandum (the explanandum is a statement describing the phenomenon to
be explained, the explanans is a set of statements describing the initial conditions and
the laws involved):
The logical conditions of adequacy are purely formal. They specify the properties of
the explanans and of the explanandum, which do not depend on the actual state of the
world. This is not the case with the condition of empirical adequacy, which states that
a supposed explanation is not truly an explanation unless one additional condition is
satisfied: the statements contained in the explanans must be true. (R1) and (R4) to-
gether imply that the statement, which is the explanandum, is also true.
Condition (R1) carries the full weight of the analysis. When we are given the expla-
nation of a phenomenon, we understand why this phenomenon occurred, in the sense
that we have an argument that shows that it was to be expected that the phenom-
enon would occur (see Hempel, 1965b, p. 337). Salmon (1989) summarizes this point
by saying that the essence of scientific explanation, according to Hempel, lies in nomic
expectability.4 The initial conditions being in place, the phenomenon could only but
occur, since it follows on logically from the initial conditions using general laws.
Note that Hempel’s model does not leave room for the common idea that to ex-
plain is to explain surprising or unfamiliar phenomena by reducing them to facts and
principles with which we are already familiar (Hempel, 1966). To explain is to bring
everything back to laws. If these laws are familiar, then the explanation will equal re-
duction to the familiar, but this is not necessarily the case. An example of the first sort
of explanation would be the kinetic theory of gases: the behavior of the molecules of
a gas, with which we are not familiar, is explained by subsumption under laws that
also apply to the movements of things with which we are familiar, such as billiard
balls. But science is overflowing with examples of the second sort. Very often, familiar
In this context, nomological simply means “relative to the laws of nature.”
4
Scientific Explanation 7
phenomena are explained by less familiar things, such as when we explain the range
of colors of the rainbow, with which we are very familiar, using the laws of reflection
and refraction of light, with which we are certainly less familiar. That the proposed
model of what a scientific explanation is does not imply that these explanations work
by reduction to the familiar is a good thing if it is simply not true that all scientific
explanations work by reduction to the familiar.
Condition (R2) enables the distinction of scientific explanations from pseudo-
explanations. Carnap (1966) explores the example of the vitalist theories of German
biologist and philosopher Hans Driesch. Driesch proposed explaining the various phe-
nomena of life by means of the notion of entelechy. The entelechy is “some specific force
that makes living beings behave in the way they behave.” The various levels of complexity
in organisms correspond to various types of entelechies. What we call the spirit of a
human being is nothing other than a part of its entelechy. It is this same entelechy, the
vital force, that explains, for example, that skin heals over after an injury. To those who
criticize the mysterious nature of the concept of entelechy, Driesch replies that it is no
more mysterious than the concept of force used in physical theory. Entelechies are not
visible to the naked eye, but electromagnetic force is no more observable—in both cases,
we see only the effects. But, as Carnap highlights, there is a crucial difference between
Driesch’s entelechies and the forces of physics. The concept of force used by physical
theories is called on from within a set of laws, whether this be the general laws of motion
and the law of gravitation in regards to gravitational force, or Coulomb’s law when re-
garding electrical force. If the concept of force has explanatory virtue, in the sense that
it can be included in scientific explanations, such as the explanation of an eclipse based
on the antecedent position of celestial bodies, the laws of motion, and the law of gravita-
tion, then it is precisely because it plays a crucial part in the formulation of these general
laws. No such thing occurs in the case of the entelechy: there are no laws of the entelechy.
Driesch offers many zoological laws that are indeed bona fide laws, but the concept of the
entelechy is nowhere to be seen, it appears at the end as something of a deus ex machina
expected to explain away the mystery of life. For Carnap this firmly establishes that en-
telechy explanations are mere pseudo-explanations, so that a virtue of Hempel’s analysis
of scientific explanation is precisely that it allows us to establish this.
Condition (R3) means that the statements in the explanans can be tested, at least
in principle. It is redundant if the explanandum is indeed an empirical fact, since in
that case the very fact that the explanandum is a consequence of the explanans enables
it to be tested. Its inclusion alongside (R1) and (R2) is no doubt a sign of Hempel and
Oppenheim’s resolutely empiricist mindset.
Condition (R4) makes the concept of explanation an objective one. Without (R4), the
concept of explanation is relative to a theoretical framework. The flaming of a match
can be deduced from the presence of phlogiston5 and the law dictating that phlogiston
5
In the chemical theory preceding Lavoisier’s modern theory, phlogiston was a hypothetical substance
supposedly found in all flammable materials and would dissipate into the air during combustion, thus
explaining the decrease in mass observed subsequent to combustion.
8 The Philosophy of Science
is released under certain circumstances, causing the phenomenon of combustion. The
modern theory of combustion, which explains the same phenomenon from the recom-
bination of various elements with oxygen, provides another explanation. In a relativ-
istic perspective, we would say that these are two explanations for one and the same
phenomenon: two explanations existing in two distinct theoretical frameworks, one
where the laws of combustion grant pride of place to phlogiston, and another where
the laws of combustion accord this honor to oxygen. But if what we want from the
concept of explanation is that it be an objective one, then this is clearly not satisfac-
tory. The explanation proposed by Lavoisier is not merely some other explanation for
combustion, rather it replaces the phlogistic explanation, the latter no longer to be
considered a genuine explanation. Subscribing to this way of seeing things, which is
undoubtedly the way of seeing things that would come naturally to scientists, implies
having an objective concept of explanation. It is just such a concept that the addition
of condition (R4) provides.
The deductive-nomological model is generalized out in two directions. First, the ex-
planandum need not necessarily be a particular event, it can also be a law, explained by
means of more general laws from which it is derived. This possibility is brought about
by the characterization given by Hempel and Oppenheim, since, although the inclu-
sion of initial conditions in the explanans may not be strictly required, the inclusion of
laws is. The canonical example of this kind of explanation is the derivation of Kepler’s
laws of planetary motion from the general laws of motion and the law of universal
gravitation. A thorough examination of this kind of explanation nevertheless uncovers
a set of problems of its own, hidden in the requirement that the laws contained in the
explanans be more general than the law to be explained.6 Note that, as before, this
explanation clearly shows us that it was to be expected that the planets would move
according to the laws set forth by Kepler, since these laws are in fact a consequence of
the law of gravitation, by way of the general laws of motion.
Second, certain scientific laws liable to arise within explanations are statistical laws,7
which do not enable us to deduce a particular phenomenon with absolute certainty,
6
Hempel and Oppenheim (1948, note 28) make the following remark. From the conjunction K & B of
Kepler’s laws and Boyle’s law, one can derive both Kepler’s laws K and Boyle’s law B. But this derivation
is not explanatory. Subsuming K and B under the simple conjunction K & B does not in any way con-
stitute an advancement in regards to explanation, as opposed to the derivation of Kepler’s laws from
Newtonian principles. The formulation of the unificationist theory of explanation given in section 4.2
aims, among other things, at resolving this problem.
7
A statistical law does not tell us that an event will always occur under certain conditions but that under
certain conditions an event has a certain probability of occurring. For example, the law that the nucleus
of a tritium atom has a three in four chance of disintegrating after 24.6 years is a statistical law. A proba-
bilistic explanation is the explanation of a phenomenon that is based on the probability that is ascribed
to this phenomenon.
Scientific Explanation 9
but simply enable us to ascribe it a high probability. Here is an example taken from
Salmon (1989). The ratio of carbon 14 to other carbon isotopes in a piece of wood found
on an excavation site is equal to half the same ratio in the atmosphere. Why? Because
this piece of wood comes from a tree that was cut down about 5730 years ago and the
half-life of carbon 14 is 5730 years. The proportion of carbon 14 in the atmosphere re-
mains constant due to cosmic radiation. The tree absorbs carbon from the atmosphere
while it is alive, but the chopped timber does not, and so the percentage of carbon-14
decreases due to radioactive decay. The general form of this kind of explanation is as
follows:
where the laws L1, . . ., Ll (notably, in our example, the law establishing the half-life
of carbon-14) and the initial conditions C1, . . ., Ck (notably, in our example, the date
on which the wood was cut) enable us to infer E (in our example, that the ratio of
carbon-14 isotopes in the wood sample is equal to half the atmospheric ratio) with
probability r which must be high. Note that here the probability is assigned to the
inductive inference, and not to the explanandum. What is explained is that the ratio
has been halved, which is neither probable nor improbable—it is quite simply true.
The explanation given is a statistical explanation insofar as the phenomenon to be
explained is not a logical consequence of the explanans, it doesn’t “definitely” result
from it, but only with a certain probability. It seems natural to demand that this prob-
ability be high since, otherwise, the explanans wouldn’t provide us reason to expect
that things should have occurred as they did; that is to say that it wouldn’t have pro-
vided us reason to expect that the explanandum be true. Based on this, it is tempting
to modify the conditions of adequacy for the deductive-nomological explanation to
the explanation Hempel calls inductive-statistical (IS) in the following manner:
Any rationally acceptable answer to the question ‘Why did event X occur?’ must
offer information which shows that X was to be expected—if not definitely, as
in the case of DN explanation, then at least with reasonable probability. Thus
the explanatory information must provide good grounds for believing that X did
in fact occur; otherwise that information would give us no adequate reason for
saying, “That explains it—that does show why X occurred.” (1965b, pp. 367–368)
where S stands for ‘suffering from a strep infection,’ P for ‘treated with penicillin,’ a for
‘John Jones,’ and G for ‘get better.’ P(G|S and P) is a conditional probability; it’s the
probability of G knowing that S and P (thus, in this instance, the probability of getting
better knowing that the patient is suffering from a strep infection and is being treated
with penicillin). Now, here’s the problem. Certain strains of streptococcus are resistant
to penicillin; in these cases the probability of getting better if treated with penicillin is
very low. So if the specific strain that has made John Jones ill is a resistant strain, we
can explain that John Jones doesn’t get better in the following manner:
Hempel introduces what he calls the requirement of maximal specificity (RMS),8 which
can be stated in the following manner. Let S be the set of statements contained in the
explanans and K the set of statements accepted at the time of the explanation,
If the conjunction of S and K implies that b belongs to a certain class F1 and that
F1 is a subclass of F, then the conjunction of S and K must also imply a statement
specifying the statistical probability of G in F1, say
In inductive logic, Carnap (1950) introduced the requirement of total evidence according to which, “in the
8
application of inductive logic to a given knowledge situation, the total evidence available must be taken
as a basis for determining the degree of confirmation” (Carnap, 1950, p. 211).
12 The Philosophy of Science
P (G|F1 )=r1
here r1 must equal r, unless the probability statement just cited is simply a the-
orem of mathematical probability theory. (Hempel, 1965b, p. 400)
If r1 does not equal r, this means that available and relevant information was not
taken into account, since it is from here that the even more precise characterization of
b being an F1 ensues, a characterization that alters the situation regarding the proba-
bility of G’s occurring. Conversely, when the requirement of maximal specificity is met,
we know that all the available and relevant information has been taken into account,
since the deployment of all our background knowledge S can tell us no more about the
probability of b’s being G.
We obtain the conditions of adequacy for IS explanations by adding a condition of
empirical adequacy to the conditions (R′1) to (R′4) we already have:9
(R′5) The statistical law contained in the explanans satisfies the requirement of
maximal specificity.
Coming back to the example of John Jones and the strep infection, “P(G|S and
P) = 0.95” can be contained in the explanans only if we do not know that Jones is carrying
a resistant strain. Indeed, since P(G|S and P) and P(G|S and P and R) are, for empirical
reasons, completely different values, the requirement of maximal specificity is violated
if the statements that we accept imply that Jones belongs to the subclass “S and P and
R” of “S and P.” Note that P(G|S and P and G) = 1—this is an elementary law of proba-
bility calculation. So in the case where we know that Jones got better, without knowing
that he was carrying a resistant strain, the requirement of maximal specificity would
nevertheless risk not being satisfied since “S and P and G” is a subclass of “S and P” and
P(G|S and P) and P(G|S and P and G) have different values. The function of the final
clause, “unless the probability statement just cited is simply a theorem of mathematical
probability theory,” is precisely to eliminate trivial counter-examples of this sort.
Finally, note also that the addition of the condition of adequacy (R′5), in which
the set K of statements accepted at the time of explanation appears as a parameter,
introduces an important difference between DN explanation and IS explanation. While
DN explanation is purely objective—the conditions of adequacy make no reference to
our knowledge state—IS explanation has an irreducibly subjective element—since the
fact that the explanans satisfies or doesn’t satisfy the requirement of maximal speci-
ficity depends on what we know. In this regard Hempel speaks of an epistemic rela-
tivity of statistical explanation.
9
This condition of adequacy is genuinely empirical, since it depends on our knowledge state, and thus on
the state of the world insofar as the fact that our knowing or not knowing something is, in a broad sense,
a fact of the world. To highlight that the only facts on which that condition depends are facts about what
we know, we could speak, as Salmon does (1989), about an epistemic condition of adequacy.
Scientific Explanation 13
We can sum up all of the above by drawing out the four types of explanations
identified by Hempel in the following table, once again from Salmon (1989, p. 9):
TABLE 1
Types of explanations
The Weber-Fechner law, the formulation of which is contemporary to the birth of psychophysics, is itself
10
a law whose validity is considered as being only approximate. It is generalized by Stevens’s law, according
to which sensation is related to stimulation by a power law.
Scientific Explanation 15
Price
Marginal cost
Average cost
Pm
Pc
Average revenue
Marginal revenue
Xm Xc Quantity
FIGURE 1 Price determination in a monopoly and in a competitive market1
1
At equilibrium, the price Pm in a monopoly situation is higher than the price Pc in a competitive situation,
and the quantity produced Xm in a monopoly situation is lower than the quantity produced in a competitive
situation. The shaded surface represents profit.
Source: Wikipedia, License Creative Commons Attribution ShareAlike 3.0
where the hypothesis that companies seek to maximize their profits comes into play in
determining the equilibrium: the quantity of goods produced by the monopoly is the
quantity at the intersection of the curves of marginal revenue and marginal cost, since
any other level of production would lead to reduced profits, and the company wants
to maximize its profits. This is quite clearly a teleological explanation. The explana-
tion is teleological because the principle of profit maximization informs us on what
the economic agents want to do. And it is indeed an explanation because this prin-
ciple is used as a law that enables, along with other laws, the derivation of a phenom-
enon to be explained, in this instance the negative effect monopolies have on price and
production.
The DN model is a general model for scientific explanation based on, as we have seen, the
idea of nomic expectability. A phenomenon is explained in so far as it has been shown
that it was to be expected that it occur. This brings us to a second important property of
the DN model, the symmetry between explanation and prediction. There is symmetry
to the extent that the difference between explanation and prediction appears as being
purely relative to our epistemic state. If a fact F is already known, its derivation from
particular laws and circumstances is an explanation. If a fact F is not known, but the par-
ticular laws and circumstances are, the same derivation is a prediction. This symmetry
leads to what Hempel calls the thesis of structural identity (Hempel et Oppenheim, 1948,
Hempel, 1965b) which can be presented as two sub-theses. On the one hand, every ade-
quate explanation is potentially a prediction, and on the other, every adequate prediction
is potentially an explanation.
16 The Philosophy of Science
Hempel (1965b) discusses an objection Scriven (1962) brings against the thesis of
structural identity, an objection which more specifically targets the first sub-thesis.11
Scriven considers the example of a metal bridge which collapses. The collapse could
have been brought about by overloading, by external damage, or by metal fatigue.
The load weighing on the bridge at the moment of its collapse was normal, and a
meticulous inspection revealed that no external damage had been caused to the
bridge’s structure. The investigators reached a conclusion of fracture by fatigue. Yet
even though metal fatigue explained the collapse of the bridge, it couldn’t have been
used to predict this collapse. By assumption, there is no other sign of the excessive
weakening of the metal than the collapsing of the bridge. When, as is the case here,
the only reason we have to subscribe to one of the elements of the explanans resides
in our acceptance of the explanandum, an adequate explanation does not, Scriven
explains, have any value for potential prediction. Hempel’s response is simple and, it
seems to us, convincing. An adequate explanation is a good prediction only when cer-
tain epistemic conditions are satisfied—that is, when the statements in the explanans
are known and the explanandum is not. In Scriven’s bridge scenario, these conditions
are far from being met, since one of the statements in the explanans cannot be known
unless the statement making up the explanandum is. The thesis of structural identity
has the following counterfactual consequence: had we known, independently, that
the metal had been weakened to breaking point, then we would have been in a po-
sition to predict that the bridge was going to collapse. However, this counterfactual
conditional is indeed true, to the extent that, by assumption, laws of physics assure
us that excessive metal fatigue is sufficient to cause the collapse of the bridge. So
Scriven’s example is not in fact a counter-example to the thesis of structural identity.
This response is illuminating in that it brings precision to the relationships between
explanation and confirmation.12 Explanation and confirmation do not generally go in
the same direction. The function of explanation is not to assure us of what is to be
explained: the phenomenon to be explained is supposed to be known. Very often the
explanandum can, on the contrary, contribute to confirming the elements contained
in the explanans, particularly the general laws. Scriven’s bridge scenario is simply a
borderline case where an element of the explanans—in this instance a specific cir-
cumstance, the fatigue in the metal the bridge is made of—has only the explanandum
as empirical support.
11
The second subthesis is only correct if every prediction is based on a law, which is not entirely ev-
ident. We can predict that the sixth egg out of the box will turn out to be rotten if the first five
were ruined without it seeming necessary to call on a law and without that prediction potentially
constituting an explanation for why the sixth egg is rotten. Hempel (1965b) suggests that, for cases
such as this, the prediction is correct only if we can present statistical laws that would validate the
probabilistic inference that the sixth egg is rotten. Otherwise, Hempel concedes the problematic
nature of the second subthesis, which is not, contrary to the first, inseparable from his theory of
explanation.
12
The next chapter of the present volume is dedicated precisely to an analysis of the concept of
confirmation.
Scientific Explanation 17
2.3 The Temporality of Explanation
Whether we consider our general discussion of the criteria of adequacy or the more
focused discussion on the difference between explanation and prediction, the issue of
temporal conditions was never brought to bear. That might seem strange. When a cer-
tain phenomenon has occurred, we can try to explain why it has occurred. Conversely,
we can try to predict that a phenomenon which has not yet occurred is going to
occur. A prominent difference between explanation and prediction thus seems to be
of a purely temporal nature. In Hempel’s model this difference is not primitive, it is
uniquely the result of an epistemic parameter. When we explain, we explain something
we know to be true, and, in the majority of cases, we know this thing to be true because
we have seen it happening in the past. Conversely, we predict things that we do not yet
know, and our ignorance is quite often related to future events. But nothing prevents
our predicting that a certain event of which we have no direct knowledge must have
happened in the past, on the basis of other facts. Another potentially relevant tem-
poral condition concerns not the chronological relationships between the particular
fact that is the explanandum (in cases where the explanandum is indeed a particular
fact) and the time of the explanation, but rather the chronological relationships be-
tween the particular fact that is the explanandum and the particular facts contained in
the explanans. In the example of the column of mercury thrust into a basin of boiling
water, the prominent particular facts of the explanans are prior to the phenomenon
to be explained: a certain set-up is described (the column of mercury in a glass tube, at
a certain temperature, the water in the basin at a certain temperature) and what will
happen next is explained on the basis of these antecedent conditions. The anteriority
of the explanans is a natural candidate for the title of condition of adequacy of the
explanation. And so, Hempel and Oppenheim (1948, §3) do indeed speak, regarding
statements describing the particular facts of the explanans, of statements “stating
specific antecedent conditions” (the emphasis is ours). All the same, the anteriority of
the explanans is not explicitly mentioned in the conditions of adequacy.
What must be made of this situation? Two remarks to start off with. First, we can
distinguish, as Hempel does, between laws of succession, which describe the evolu-
tion of a system, and laws of coexistence, which describe the state of a system. The
law of universal gravitation and the laws of movement can be used to describe the
evolution of the solar system (the movements of the planets). Boyle’s law, which
relates the pressure, volume and temperature of a real gas, describes the state of
a gaseous system. Boyle’s law can be used to explain the volume of a gas using its
temperature and its pressure. In this particular case, and in all cases where laws of
coexistence are used, the particular circumstances contained in the explanans are
not strictly prior to the explanandum, they are concomitant to it. Second, it is some-
times possible to use laws of succession “backwards,” when the processes described
are reversible. The particular facts described by the statements C1, . . . , Ck take place
at instants t1, . . . , tk which are posterior to the instant t when the particular fact
F took place and which we derive from laws and also from C1, . . . , Ck. For example,
18 The Philosophy of Science
we can deduce the position of the planets at an instant t using the laws of celestial
mechanics and the position of the planets at a time t’>t. The deductive-nomological
structure is the same as for the explanations or the “genuine” predictions for which
the anteriority of the particular circumstances described in the explanans is con-
firmed. Hempel (1962, p. 116) speaks of “retrodiction” to name the counterpart of a
prediction where the explanans is prior to the time of the explanation. But the intro-
duction of the term does not resolve the problem. If we have retrodiction when the
epistemic situation is one of prediction (F was not known ahead of time), is there,
yes or no, explanation, admittedly of quite a particular type, the retrodictive type,
when the epistemic situation is one of explanation (F was already known)? Here is
Hempel’s response:
If the full weight of the analysis is carried by the concept of laws, the analysis will
only be complete when that concept itself is clear and precise. Following on from
Hempel, let us begin by distinguishing laws and nomological statements, a nomolog-
ical statement being a statement that is a law provided that it be true. It is not for us to
decide which nomological statements are true—it is to science itself that it falls to say
which nomological statements are confirmed to a high enough degree and are to be ac-
cepted as true. Our task, in completing Hempel’s analysis, is then to characterize nom-
ological statements, which account for the nomic expectability of the explanandum in
the DN model.
Nomological statements are typically universal, conditional statements, such as “all
metals are conductors” (Hempel and Oppenheim, 1948, §6, entitled “Problems of the
concept of general law”). The general form of nomological statements, in logical nota-
tion, is ∀x (φ(x) → ψ(x));13 that is, for every x, if x is a φ then x is a ψ. The putative law
thus establishes the relationship between the fact of being φ (for example, the fact of
being a metal) and the fact of being ψ (for example, the fact of being a conductor of
electricity). By contrast, a particular statement, such as “certain metals are present
in nature in a non-oxidized state” clearly doesn’t claim the status of general law, and
thus does not constitute a nomological statement. A universal statement whose scope
is artificially restricted will not count as a nomological statement either. Saying that,
on earth, the bodies of all living organisms contain carbon is not stating a general law
about living organisms.14 There is still another way in which a nomological statement
is general: it must not make reference to specific individuals. The general unrestricted
universal statement, “all of Napoleon’s brothers-in-law became kings” is not a candi-
date to be a law, because it makes reference to a very specific individual, Napoleon.
Neither should the generality of the statement be compromised by reference, implicit
or explicit, to specific times or places. The statement, “all boats which navigate beyond
13
Hempel and Oppenheim point out that in reality only the universal form is necessary since, syntacti-
cally speaking, the conditional statements can be transformed into equivalent statements that are not
conditional. For example, the universal conditional statement, “all metals are conductors,” is logically
equivalent to the statement, “all things are not metals or are conductors,” which is universal but not
conditional. Nevertheless, it is possible to make the same remark regarding universal quantification,
since “all metals are conductors” is equivalent to “it is false that some metals are not conductors.” It is
thus necessary to provide a definition of the concept of universal statement that is not purely syntactic
(see 1948, §7).
14
The exclusion of restrictions on scope poses its own problems. Many laws apply ceteris paribus. For ex-
ample, the law establishing the thermal expansion coefficient of a metal only applies all other things
being equal: the length of a heated metal bar will not increase by the proportions predicted by the law if
somebody hammers at one of the ends of that bar (Lange, 1993). For a discussion of ceteris paribus laws
in relation to economics, see chapter 15 in this volume.
20 The Philosophy of Science
the 75th degree of northern latitude risk being trapped in the ice” is universal, unre-
stricted, and doesn’t make reference to individuals. Its generality is nevertheless lim-
ited by reference to a particular location (the 75th degree of northern latitude) so that
it cannot claim to be a law either.15 Having reached the end of the analysis, it appears
that a nomological statement must be a universal statement, without restriction of
scope and containing no purely qualitative terms. Are these necessary conditions also
sufficient?16 Consider the following statements:
(1) All signals travel at speed less than or equal to the speed of light.
(2) All solid spheres of gold have a diameter of less than one mile.
(3) All solid spheres of uranium-235 have a diameter of less than one mile.
(1), (2), and (3) satisfy the conditions we have just set forth. However, only (1) and
(3) are nomological statements. (1) is one of the fundamental principles of the theory
of general relativity, and (3) comes from the laws which govern nuclear fission. The
critical mass of uranium-235, the mass beyond which a chain reaction of nuclear fission
spontaneously occurs, is well below the mass of a one mile sphere of that isotope. Even
if (2) is probably just as true as (1) and (3), it is still not a law of nature. That there is
not a gigantic golden sphere in the universe is merely an accidental generalization.
Correlatively, (2) does not seem to have any explanatory power. Saying that some me-
tallic sphere has a diameter of less than one mile because it is made of gold does not in
any way seem to constitute a good explanation. On the contrary, we could explain that
the speed of a given signal transmission is inferior or equal to the speed of light by ref-
erence to (1).17 Further, there is no difference between (2) and (3) in terms of the logical
form of the statement or in terms of the nature of the expressions contained therein,
so that it seems pointless to try and separate them by recourse to conditions like the
necessary conditions which have been given thus far.
We can nevertheless point out the differences between (2) and (3). A first difference
concerns what happens when certain fictional situations are envisaged. Consider the
following counterfactual statements:
(4) If this sphere were made of gold, its diameter would be less than one mile.
(5) If this sphere were made of uranium, its diameter would be less than
one mile.
15
We omit the difficulties relative to the ideas of unrestricted scope and purely qualitative terms. Only a
certain number of them are discussed by Hempel and Oppenheim (1948).
16
This short introduction to the problem of characterizing laws of nature follows the classics van Fraassen
(1989, part 1) and Salmon (1989, pp. 14–19). See Carroll (2012) for a more thorough survey.
17
That the distinction between nomological statements and accidental generalizations seems to intui-
tively overlap with the distinction between universal statements having explanatory power, and uni-
versal statements not having explanatory power, corroborates the importance the DN model ascribes
to the laws of nature.
Scientific Explanation 21
Let’s imagine that (4) and (5) are stated in front of an enormous bronze sphere
which could well have a diameter of more than one mile. Intuitively, in that context,
(4) is false. If the bronze sphere has a diameter of more than one mile, had it been
made of gold, it would still have a diameter of more than one mile. Intuitively, in the
same context but also in all other contexts, (5) remains true. Had the sphere been
made of uranium, then it couldn’t have had a diameter of more than one mile since
it would have exploded before reaching that mass. Nomological statements support
counterfactuals—they remain true when they are reworded counterfactually, like when
(3) becomes (5) —while accidental generalizations do not support counterfactuals: (2)
may well be true, (4) certainly is not.
Another similar difference is related to modal contexts.18 So, let’s compare the
following:
(6) Necessarily, all solid spheres of gold have a diameter of less than one mile.
(7) Necessarily, all solid spheres of uranium-235 have a diameter of less than
one mile.
(6) is true to the extent that the existence of such a sphere would defy the laws
of physics which apply in all possible worlds (or at least in all the physically possible
worlds, were we to posit the existence of logically possible but physically impossible
worlds). By contrast, (7) is certainly not true: an enormous solid gold sphere, patiently
put together by generations of goldsmiths or present in a natural state thanks to some
exceptional conditions, and having a diameter of more than one mile could well exist.
Nomological statements have modal import—(6), which is the modalized version of
(2), is true—while accidental generalizations have no modal import: (7), the modalized
version of (3), is not true, even if (3) is true.
Perhaps we will hold on to these conditions, adding them to the previous ones to
characterize nomological statements in a necessary and sufficient manner. A nomo-
logical statement would then be defined as a universal statement without restriction
of scope, containing only qualitative expressions, that support counterfactuals and
have modal import. Less the adequacy of this characterization, it is rather its analyt-
ical virtue which is now problematic. We can give account for the notion of nomolog-
ical statements in either modal or counterfactual terms. But the fact of having modal
import or of supporting counterfactuals seems at least as mysterious as the fact of
being able to claim the status of a law. It could even be tempting to turn the order of
the analysis around and to say that (2), for example, supports counterfactuals because
(2) is a law and not simply an accidental generalization. In the same way, it could be
tempting to clarify the notion of necessity by saying that anything is possible that
doesn’t defy the laws of nature. Problems of conceptual priority like this arise with any
By “modal context” we mean a subclause taking on the role of a modal operator, such as “necessarily,” “it
18
Language: English
BY
J. G. FRAZER, D.C.L., LL.D., Litt.D.
FELLOW OF TRINITY COLLEGE, CAMBRIDGE
PROFESSOR OF SOCIAL ANTHROPOLOGY IN THE UNIVERSITY OF
LIVERPOOL
PSYCHE’S TASK
I. Introduction
The dark and the bright side of Superstition: a plea for the accused:
four propositions to be proved by the defence 3-5
II. Government
IV. Marriage
VI. Conclusion
INDEX 177-186
ENDNOTES
PSYCHE’S TASK
I.
INTRODUCTION