scientific explanation and the causal structure of the world - wesley c. salmon(cut)

8/18/2019 Scientific Explanation and the Causal Structure of the World - Wesley C. Salmon(Cut)

http://slidepdf.com/reader/full/scientific-explanation-and-the-causal-structure-of-the-world-wesley-c-salmoncut 1/35

Scie tific

Explanatiand theCausal Structure

Princeton Univ ersity PressPrinceton New Jersey



Copyright © 1984 by Princeton University PressPublished by Princeton University Press. 41 William Street ,Princeton, New Jersey 08540In the United KinRdom: Princeton University Press,Guildford, Surrey

ll Rights Reserved

Library of Congress Cataloging in Publi ca tion Data ,,·ill be found on thelast printed page of this book

ISBN cloth) 0-691-07293-0LPE) 0 -691-10170-1

This book has been co mposed in Linotron Times Roman

Clothbound editions of Princeton University Press books are printedon acid -free paper. and binding materia ls are chosen fnr strength

and durability. Paperbacks, although satisfactory for personalcollections, are not usually suitable for librmy rehinding

Printed n the United States of Americaby Princeton University Press, Princeton, New J ersey



To

Car G. HempelPhilosophical Master

aka

Peter Hempel

Colleague and Friend



ontents

PREF CE ix

CKNOWLEDGMENTS X I l l

I Scientific Explanation: Three General Conceptions 3

Explanation versus Description 4Other Types of Explanation 9Laplacian ExplanationThree Basic Conceptions 15

An Outline of Strategy 2

2 Statistical Explanation and Its Models 24

Inductive-Statistical Explanation 27

The Statistical-Relevance Approach

3 Objective Homogeneity 48

Epistemic Relativization 48

Randomnes s 55

Homogeneity 60

Some Philosophical Applications 72Some Philosophical Reflections 76

4 The Three Conceptions Revisited 84

The Epistemic Conception 84

The Modal Conception I l l

The Ontic Conception 121

How These Concertions Answer a Fundamental Question 124

Conclusions 133

5 Causal Con nec: tion s 135

Basic Problems 135

Two Basic Concepts 137

Processes 139The 'At -At' Theory of Causal Propagation 147



viii J Contents

6 Causal Forks and ommon a u s e ~

Conjunctive ForksInteractive ForksRelations between Conjunctive and Interactive PorksPerfect ForksThe Causal Structure of the WorldConcluding Remarks

7 . Probabilistic Causality

The Sufficiency /Necessity ViewStatistical Relevance and Probabilisti c CausalityCausality and Positive RelevanceCausal Processes and Propensities

8. Theoretical Explanation

Causal Connections and Common CausesExplanatory versus Inferential PrinciplesThe Common Cause Principle and Molecular RealityThe Explanatory Power of TheoriesEmpiricism and Realism

9 The Mechanical PhilosophyLogic versus MechanismsExplanation in Quantum MechanicsExplanation and UnderstandingThe Causal/Mechanical ModelThe Final Contrast

BIBLIOGRAPHY

INDEXi

158

158168174177178182

184

185190192202

206207211213227229

239239242259267276

280

295



Preface

ALTHOUGH there is no doubt in my mind that scientific understanding ofthe world and what transpires within it represents a marvelous intellectualachievement, the chief purpose of this book is not to laud scientific un-derstanding, but, rather. to try to get clear on what it is. Our aim is tounderstand scientific understanding. We sec ure scientific understanding byproviding scientific explanations ; thus our main concern will be with thenature of scientific explanation.

The attempt to find an adequate characterizati0n of scientific explanationis a problem that has perplexed scie ntists and philosophers for severalmillennia. As we shall see in chapter I . its roots go back at least to Greekantiquity, and it continues to be an object of considerable attention today.There are, it seems to me, three basic conceptions that have figured significantly in the discussions for twenty -five hundred years. We shall examine them in some detail.

One particular view has played a dominant role, at least since the middleof the twentieth century, in the philosophical literature on scientific explanation. t has often been called omewhat inappropriately I thinkthe covering law model. Its crucial tenet i s that explanations are arguments. Another view, which has enjoyed some philosophical support,has a good deal of common-sense appeal. It takes scientific explanationsto be essentially causal. A major aim of our investigations will be to draw

a sharp contrast between these two conceptions . A third conception, whichinseparably associates explanation with necessitation , will be judged scientifically anachronistic .

t will be my contention that the ' received view' (the inferential conception) is infected with severe diniculties that. in the end, render ituntenable . The causal conception also suffers from serious problemsnamely, philosophical perplexities c0nceming the nature of causality - hutl believe they can be overcome. This can be done only if we are able to

provide an adequate philosophical treatment of fundamental causal concepts, as [ attempt to do in chapters 5 · 7.

The distinction between these two concep tions of scientific explanationis by no means trivial or merely verbal. As l indicate in the final sectionof the last chapter , the transition from the one conception to the otherdemands nothing less than a radical gestalt switch in our thought about



x I Preface

scien tific explanation. I shall, of course, attempt to provide stron g philil lljillira rllrlil fllinn frrr r ing lllf s hifl .

Alth ough much modern work on scientific explauatlon has been ratherformal and technical--often treating various quasi -forma l 'models' in greatdetail- sha ll dwell ex ten sively upon less formal considerations. Thereare two reasons for this emphasis. In the first plac e, I have been convincedfor some tim e that many recent philosophical discussion s of scientificexp lanat ion suffer from a lack of what Rudolf Carnap called ''clarificationof the explicandum.'' As Camap has viv idly shown, precise philosophicalexp lication s of important concepts can egregiously miss the mark if wedo not have a sound prior informal grasp of the concept we are endeavoringto explicate. It is impossible, I think, to read chapters 1, 2, and 4 of

Camap 's clats ic , Logical Foundations of Probability (1950), without seeingthe profound importance of such preliminary clarifications . Because of my ·stro ng conviction that the concept of scien tific explanation needs similare lucidation, chapters I 4 and 9 of thi s book arc large ly devoted to thistask.

In the second place, I have earnestly attempted to make the discussionaccessible to a wide group of readers-philosophers, scientists, and otherindividuals who have serious interests in science or philosophy . Everyeffort has been made to illustrate the philosophi ca l points with examplesthat have either his tori ca l or contemporary significance from a wide rangeof scientific disciplines . At the same time, a number of examples havedeliberately been drawn from such sources as Scientific American andScience for I hope this book will be intelligible to those who find suchliterature rewarding. No particular specialized knowledge of any scientificdi scip line is presupposed . In addition, I have tried to supply enough philosophical background to obviate any need for prior familiarity with thephilosophical literature on scie ntific explanation . It is my special hope thatsome parts of th is book - f or example, the discussions of ca usality-willbe useful to philo sop hers who work in areas outside of philosophy ofscience .

I do not mean to suggest that this book is easy; it is not. For the mostpart, however, it shou ld be comprehensible to inte lligent readers, more orless regardless of backgrou nd , who are willing t o expe nd some effort uponit. Unavoidably, some parts are rather technical - in particular, the detailsof the S-R basis in chap ter 2 and the treatm ent of randomness and objectivehomogeneity in chapte r 3. These can be skimmed or omitted on first readingby thos e who are not interested in such tec hn ica li ties for their own sake.

Th is book had its real beginnings when , around 1 9 7 ~ I began to recognize the serious shor tcomings and limit ations of the statis tical -re leva nce



Prerace 1 x1

model or sc ientific ex planation a s propounded in (Sa lmon et al . , 1971 ) -

and the need to come to grips with the explanatory force of scientificthcoric i. At kind in vitation, I p r o m i ~ e dt con tribu te a paper entitled"Theoretica l Explanation" to a sy mpo sium on explanation organized by

Stephan Komer at Bristol; the proceedings are published in (Korner. 197 5).As sometimes happen s when the title is written before the paper , the contentof the paper did not fit the title very we ; all of my efforts to deal withtheories got bogged down in problem s of causality. As chapters 5 R ofthe present book wi ll show, I now believe that causa l explanation is thekey to theoretical explanation. No adeq uate account of theoretical explanation could be developed withou t ext ensive work on causality itself .

The actual writing of the book began during a visit to the Departmentof History and Philosophy of Science at the University of Melbourne ,where was invited to give a seminar on scientific explanation in 1978 .Suffice it to say that every chapter bear s the marks. explicitly or implicitly,of stimulating interactions with Australian philosophers . I am ind eed grateful for the personal and intellectual ho spitality accorded me durin g a splendid three -month visit.

There are severa l individuals to whom 1 owe special thanks for substantive help in developing the ideas in this book. Nancy Cartwright pointedout a serious flaw in an earlier attempt to explicate the propagation ofcausal influence , and stimulated me to improve my account. Alberto Coffamade me see the crucial role of epistcmic relativization in Car G. Hempel 'stheory of inductive explanation, and made me realize the indispensablecharacter of objective homogen eity for the type of theory I was attemptingto develop. Clark Glymour called to my attention the epoch-making workof Jean Perrin on Avogadro's number and directed me to Mary Jo Nye'ssuperb account of that work (1972). Paul Humphre ys offered sub stantia laid in dealing with objective homo ge neity, probabilistic causality, andpropensities . Patrick Maher and Ri chard Otte noticed a fatal flaw in anearlier explication of causal interacti on and forced me to correct that account. Otte also provid ed va luabl e cri tiqu es of various aspects of probabilistic causality. Hu gh Melior, more than anyone else, led me to appreciatethe importance of clarifying the explicandum in dis cussion s of scientificexplanation . Bas van Fraasse n made me realize that there must be twokind s of causal forks, co njun ct ive and interactive. Philip van Bretzel firstcalled to my attention the direct relationship between causa l forks andcausal inter actions .

1 cannot beg in to detail the ways in which my ideas on scientific explanation have been influen ced by the wri tings of Bertrand Russcll especially The Ana ysis of Matter ( 1927), Mysticism and Logic (1929). andHuman K n o w l e d ~ eIts Scope anrl Limits (1948)--as well as Hans Rei -



xii I Preface

chenbach ' s posthumous work, The Direction o Time (1956). Althoughneither of these authors dealt explicitly with scientific explanation at anylength, both provided gold mines of material directly relevant to the subject.

My main intellectual debt- l ike that of anyone who has worked onphilosophical theories of scientific explanation in recent years-is to CarG. Hempel, who has taught all of us how to think clearly and deeply aboutthe subject. f we have failed to learn the lesson , th e fault is not his.Although I disagree on many points with his views, his method of philo sophizing seems to me peerless, and I can think of no more worthy goalthan to try to emulate it

Many of the ideas discussed in this book were published in preliminaryfom1 in var iou s artic les that have appeared since 1975. Specific citationswill be found under Acknow ledgments.



Acknowledgments

I AM GRATEFUL to Scientific American for permission to use the excerptsfrom Quantum Theory and Reality'' by Bemard d 'Esp agnat, Copyright

© 1979 by Scientific American, Inc . These excerpts appear in chapters 1and 9; specific page references are given at the end of each quotation .

I shou ld lik e to express my sincere appreciation to the following publishers for permission to adapt material from my previously publish edarticles for use in this book:

Portions of the material in chapters 1 and 4 were included , in a much

abbreviated fonn, in Comets, Pollen, and Dreams: Some Reflection s onScientific Explanation,' ' in Robert McLaughlin, ed., What? Where? When?Why? pp . 155 - 178 . Copyright© 1982 by D Rei del Publishing Company,

Dordrecht, Holland .The section Epistemic Relativization in chapter 3 is adapted from

Indeterminism and Epistemic Relativization, Philosophy o Science 44pp . I99-202. Copyrig ht © J977 by the Philosophy of Science Association.East Lansing, Michigan.

The sections Randomness and Homogenei ty in chapter 3 are drastically revised versions of material that appeared in Objectively Homo

geneous Reference Classes, Synthese 36, pp. 399--414. Copyright© 1977by D. Rei del Publishing Company, Dordrecht, Holland.

The section ' 'The Epistemic Conception'' in chapter 4 includes materialadapted from' ' A Third Dogma of Empiricism.'' in Robert Butts and Jaakko

Hintikka, eds., Basic Problems n Methodology and Linguistics pp. 149 -166. Copyright© 1977 by D. Reidel Publishing Company, Dorclrecht. t

also contains material from Hempel 's Co nception of Inducti ve Inferencein Inductive -Statis tical Explanation,' ' Philosophy n Science 44, pp. 180-

185. Copyright © 1977 by the Philosophy of Science Association. EastLansing, Michigan.

Some of the basic ideas in chapters 5 and 6 were presented in a preliminary and much-condensed version in Causa lity : Production and Propagation, in Peter Asquith and Ronalct N. Giere, eels., PSA 1980 pp. 49 -69. Copyright © 1982 by the Philosophy of Science Association, EastLansing, Michigan. The section The 'At-At' Theory of Causal Propaga tion in chapter 5 is an adaptation of An 'At-At' Theory of Causal



· xiv Acknowledgments

Influence, Philosophy of Science 44, pp. 215- 219 . Copyright© 1977by the Philosophy of Science Association, East Lansing, Michigan .

Most of the material in chapter 7 was drawn from Proba bilistic Causality, Pa c ific Phil os ophical Quarterly 61, pp. 50-74. Copyright© 1980

by the University of Southern California . •Some of the material in chapters 8 and 9 was anticipated in a much lesscomplete fonn in Further Reflections, in Robe1t McL1ughlin ed., What?

Where? When? Why ? pp. 231 --280 (especially pp. 260- 278) . Copyright© 1982 by D . Rei del Publishing Company, Dordrecht, Holland .

should like to express spec ial thanks to the University of Melbournefor providing me with a research appointment in September through November 1978, and to the National Science Foundation for support of my

research on causality and scientific explanation.This material is based upon work supported by the National Science

Foundation under Grant No. GS-42056 and Grant No . SOC-7809146.Any opinions, findings, and conclusions or recommendations expressed

in this publication are those of the author and do not necessarily reflectthe views of the National Science Foundation.



Scientific Explanation andthe Causal Structure of the World



Scientific Explanation:Three General Conceptions

MoDERN SCIENCE provides us with extensive knowledge about the worldin which we live . We know that our universe originated in a hig bang·severa l thousand million years ago, and that it has been expanding ever

since. We know that th ere was a fairly severe drought in northeast ern

Arizona during the last few years of the fourteenth century A D. , and thatlarge settlements were abandoned at that time . We know that Halley s

comet moves in a roughly elliptical orbit, and that it will return to perihelionin 1986. We know that-barring nuclear holocaust-the human populationof the earth will continue to increa se for some time to come. We knowthat the planet Uranus has rings, that E coli liv e in the norn1al humanintestinal tract, that copper is an electrical conductor, and th at the surfacetemperature of Venus is hi g h . As these examples show, science provides

knowledge of what has happened in the past, what will happen in thefuture, and what is happening now in regions that we are not observingat the present moment. t encompasses knowledge of both particular factsand general regularities. In none of these instances, of course, is our

scientific knowledge to be regarded as certain or incorrigible; neverthele ss,the physical, biological, and social sciences furnish impressive bodies of

knowledge about what goes on in the world, and we have every reaso11 tobelieve that these sciences will continue to grow at a prodigious rate.

Such knowledge - valuable as it may be for intellectual or practicalpurpose s i s not fully satisfy ing . Not only do we desire to know wh t

happens; we also want to understand why Moreover, it is widely acknowledged today that sc ien ce can provide explanations of natural phenomena:indeed, to many philosophers and scient ists, this is the primary goal of

scie ntific activity . Scientific explanations can be given for such particularoccurrences as the appearance of Hallcy s cornet in 1759 or the crash of

aDC-10 jet

airlinerin

Chicagoin

1979, as wellas

such general featuresof the world as the nearly el lipti cal orbits of planets or the electricalconductivity of copper. The chief aim of this book is to try to discoverjust what scientific understanding of this sort con sists in.

Before undertaking the t ~ kat hand. it may be u ~ e f u lto make a remarkabout the basic strategy. During the last thirty -six years since I 94R . a



4 Scientific Exp lanation•

numb er of qua si-forma l 'models' o scie ntific explanation have appearedin the philosophical litera ture , and a good deal of attention has been devotedto the in ves tigation of their formal properties. Quite a few are such familiarfrie nds that philosophers often refer to them m ere ly by their initials . Wesha ll meet many of tt;em in the ensuing chapters: deductive-nomological(D-N), deductive-statistical (D-S), inductive -s tatisti ca l (J-S), statisticalre levanc e (S-R), d e c l u c t i v e n o m o l o g i c < ~ l p r o b a b i l i s t i c(D -N-P) , expectedinform ation (E- 1 , ls well as several others for which such initializeddesignations hav e no t bee n generally adopted.

Although shall di sc uss in some detail various technical aspects of themodels, my initial co nce rn will be with what C a r n < ~ p( 1950, sec. 2) called

clarification of the explicanclum. Many philosophical studies, includingthe one to which this book is devoted, aim at providing reasonably preciseexplications of fund amentnl concepts. but unless we take preliminary stepsto give some unde rstanding of the concept we are trying to explicate - theexplicandum- ·any attempt to formulate an exact explication is apt to bewide of the mark. J am firmly convinced that such terms as explanation, ''scientific ex planation,·' ''scientific understanding,' ' nnd a host o closelyrelated terms arc far from being well understood even in a preliminary andimprecise way . Consequently, in an effort to improve that situation, shalldevote the llrst chapter to a survey of three general conceptions of scientificexplanation that have ven erable histories and contemporary importance.In the fourth chapter, these three general conceptions will be reconsideredin detail in the light of certain pervasive features of explanations that makesome essential appeal to statistical laws . As a result of these efforts atpreliminary clarification of the explicandum, we shalt find important guidelin es for the evaluation of the variou s 'models' that have been proposed

in recent years.

EX PLAN ATJON VERSUS DESCRIPTION

The i d e ~that there are two kinds of scientific knowledge - knowledgeof what add knowledge of why- i s not new. In the Posterior Analytics71 b 18-25), Aristotle distinguishes syllogisms that provide scientific u n ~

derstanding from those that do not. In The Art of Thinking ( Port-Royal

Logic, first published in 1662) , Antoine Arnauld di stinguishes d emonstra tion s that merely convince the mind from those that also enlighten it .He co ntinues , ' 'En lightenment ought to be the principal fruit of true knowledge. Our minds are unsatisfied unless they know not only that a thing isbut why it is'' (1964, p. 330).

The notion th at explanation is a major goal of scie nce- -or any goal ofscience at a ll- has not been universally adopted. In the 1911 edition of



Explanat ion versus Description 5

The Grammar o f Science Karl Pears on spoke for a large group of physi calscientists when he said, Nobody believes that science explains anything;we all look upon it as a shorthand description, as an economy of thought'·( 1957, p. xi). Around the turn of the century, many physicists and c hemists

held this view about the nonexplanatory character of science because theywere convinced that the existence of such micro-entities as electrons,atoms, and molecules could not be scie ntifically esta blished; indeed, manybelieved that knowledge of the microstructure of matter is beyond thescope of science. Some argued that this is not really a limitation on th epower of science, for the proper bu siness of science is to enable us tomake reliable predictions--{)r, more accurately, to make inferences fromobserved facts to unobs erve d facts - but not to try to explain anything. In

its most extreme form hich is, incidentally , the version Pearson espoused - - this thesis concerning the nature of science holds that the solepurpose of science is to enable us tp predict our future experiences.

Several distinct issues are involved here . The first issue has to do withphenomenalism . This is the doctrine that the only reality consists of humansensations-that what we nom1ally t ake to be ordinary physical objectsare actually nothing more than complex combinations of sensations. Oneclassic expression of this thesis can be found in Ernst Mach's The Analrsis

o f Sensations ( 1914), and it W IS embraced by number of early a d h e r ~ n t

of logical positivism. According to this view , the aim of science is todiscover regular patterns among our e n s ~ t i o n sth<1t will enable us to predictfuture sensations. Since, in my opinion, phenomenalism has been effectively discredited on a number of grounds, do not intend to di scuss itfurther in this book. shall adopt the standpoint of physicalism whichholds that perception (of a fairly direct sort) provides us with reliablethough not incorrigible nowledge of ordinary middle-sized physical objects and events. We have, believe, so und reasons for taking such entitiesas flowers, rocks, and sneezes to be real.

The controversy regarding instrume11talism and theoretical realism isanother issue that arises in this context. Instmmentalism is the doctrinethat scientific theories that seem to make reference to unobsetvable entitiesare not to be considered litera lly true; instead, they should be regarded asuseful instruments for dealing with the obsetvabl .e events, objects. and

properties that we find in our environment. For examp le, some phi losorhersand scientists have claimed that although such perceptible entities as th ermometers, pressure gauges, and containers of gas are real enough, atomsand molecules , which are too small to be observed, are unre al. In responseto the successes of the molecular-kinetic theory of gases in the latter pat1of the nineteenth century, instmmenta lists like Pearson and Mach couldreply that atoms and molecules are useful fictiPns that e .nable us to make



6 I Scicnlifi c Expl:lnation

excellent predictions about the behavior of gases . At that time. however,many still insisted that there was no compelling evidence that such entitiesactually exist. As 1 shall explain in chapter 8, it seems to me that nineteenthcentury physical scientists were justified in taking an agnostic view con

cerning the nature of microphysical entities, but that decisive evidence forthe existence of atoms and molecules emerged ea rly in the twentieth century . For the moment, however, I merely want to ca ll attention to thethesis of instrumentalism .

If one maintains that atoms and molecules are rea l. entities, then it maybe plausible to claim that they enable us to explain the behavior of gases.Pressure, for examp le , is exp lained on the basis of co llisio ns of sma llmaterial patticles (which obey Newt on's laws of motion) with the walls

of the container. If however, atoms and molecul es a re mere fictions, it .does not see m reasonable to suppose that we can exp lain the behavior ofa gas hy say ing that it ac ts as i f it were composed of sma ll particles. Sinceinstrumentalis ts do not have any alternative explanation of suc h phenomena, they have often taken re fuge in the view thilt providing explanationsis no part of the scien tific enterprise anyhow.

The main idea of Mach, Pearson, and others concerning shmthanddescriptions and eco nomy of thought may be illustrated by JohannJakob Balmer's fom1Ula for a parti cular series of line s in the spectrum ofhydrogen. In 1885, Balmer published a fonnul a for the wavelengths ofthe spectral lines, which can be written,

~ = R(l /2 2 - l/n 2 ,

where n takes successive integral values beginning with 3, and R is nowknown as the Rydberg constant. When he worked out this relationship,

Balmer was aware of the wavelengths of four lines that had been measuredprecisely:

H = 6562.10 A (n = 3, redH = 4860 .74 A n = 4, green)Hy 4340 . 1 A 11 5 , blue)H = 410 1.2 A 11 = 6. vio let)

Additional lines in the Balmer series have subsequently been disc0vered,and the measured values of their wavelengths agree well with the valuesthat come from Bal mer's formula . Obv iously, it is more economical towrite the simp le formula than to keep a list of the numerical values forthe wavelengths of all of the lines. Balmer also speculated•that hi s formulacould be genera lized in the following way:

IIA = R(l m1 - l/n2 .



Exrlanalion versus Descriplion I 7

where m has b een substituted for 2 in the previous formula. Additionalseries that fit this generalized formula hav e been found:

m 1 Lyman seriesm = Balmer seriesm 3 Paschen seriesm 4 Brackett seriesm 5 Pfund series

This formula yie lds a n even more economical description of the lines inthe hydrogen spectrum, but manifestly neit her fonnula explains anythingabout the spectral lines . According to the instrumentalist, it seems, theultimate aim of sc ience is to find formulas of the foregoing sorts that enable

us to describe concisely certain known facts and to predict accurately othersthat may be discovered in the future.

The instrumentali st view has also hcen highl y innu cnt ial in psychology ,where behaviorists have maintained ihat their science is concerned oulywith the stimuli that impinge upon their subjects (humans or animals ofother species) and the physical responses that are elicited under specifiablecircumstances. These stimuli and responses can be objectively observedand measured . Internal p syc holog ical states-for examp le , feelings of anx

iety , hunger, or love a re highly subj ective: they are not amenable toobjective measurement. They ca n, however, be regarded as fictions-sometimes called intervening var iables --whic h do not literall y describeany psychological reality, but which do have instrumental value in describing stimulus-response relationships and in predicting observable behavior . This kind of instrumental ism has remained influential in psychologywell into the twentieth century. A classic account of the situation in psychology at mid-century can be found in (MacCorquodale and Meehl, 1948) .

ln his preface to the third edition of The rammar of Science, just priorto the previously quoted statement about the nonexplanatory character ofscience, Pearson remarks, Reading the book aga in after many years [thefirst edition was published in 1892, the second in 1900], it is surprisingto find how the heterodoxy of the 'eighties had become the commonplaceand accepted doctrine of to -day . For some years now , it seems to me.the commonplace and accepted doctrine 0f 1911 has been heterodox .and theoretical realism has become the orthodox view. Howev er, theoretical realism has quite recently come und er strong attack by Bas vanFraassen 1980) and Hilary Putnam ( 1982, J 982a), among others. Onewonders to what extent such changes in philosophic doctrine are merematters of changing fashion, rather than so lid results based upon strongphilosophical arguments. I sha ll return to this question in chapter 8.

Although the view that exp lan ation is outside of the domain of scienceoften goes hand in hand with the denial or theoreti ca l realism, the y do llllt



I8 I Scientific Explanation

always go together. For exam ple, a le tter to th e editor of the Scientifici\merican offered the following comment on an art icle that had claimedgreat explanatory merit for the quark hypothesis:

The quark hypothe s is describes the behavior of the subatomic particles

in a successful and concise manner. No explan ation of these particles behavior is , of co ur se, ever possible thr ough either pure or experimentalsc ience. Why? is a question left to philosophy and is not a properpart of a ny physical theo ry. (McCalla, I 976, p. 8

The author of thi s letter does not appear to be denying the reality eitherof quark s or of other microphysical entities. It would b e incorrect to givethe impr ession that a denial of realism is the only ground on which people

maintain that exp lanation is outside of the domain of science. In somecases , I suspect, such views rest upon the anthropomorphic notion that'explanations' must appeal to human or superhuman purposes. I do notknow what rea so ns mo tivated the author of the foregoing letter.

Conv erse ly , it would also be a mistake to suppose that everyone whorejects theore tical realism denies that sc ien ce ha s exp lanatory power. Aswe sha ll see in greater detail in chapter 4, van Fraassen, whose agnosticattitude regarding microphysical entities makes him an antireali s t, offers

a powerful account of sc ientific explanation ( 1980 . cha p. 5) .The que stion w he ther sc ience ca n provide explanations as well as de

sc riptiv e knowledge arises poignantly in contemporary quantum mechanics . In a popular article enti tled The Quantum Theory and Reality ,Bem ard d'Espagnat (1979) brings out this issue with admirab le clarity. Atth e outset. he focuses explicit attention upon the problem of exp lanation:

•Any successfu l theory in the physical sciences is ex pected to make

accurate predictions . .. .

From thi s point of view quantum mechanicsmust be judged highly successful. As the fundamental modern theoryof atoms, of molecules , of elementary particle s, of e lectr omagneticradiati o n. a nd of the solid state it supplies methods for calculating theres ult s or experiments in all these realms.

Apart from expe rim e nta l confirmation , however , something more i sgenera ll y demanded of a theory. I t is expected not only to detem1inethe re sults of an experiment but also to provide some under standing of

the physical even ts tha t are presumed to und er lie the observed res ults .In other words, th e th eory s hould not only give the position of a pointeron a d ial but also exp lain why the po inter t akes up that po sition . Wheno ne seeks infonnation of this k ind in the quantum theo ry. cer1ain concept ual difficulties arise. (P. 158)

Th e hulk or the artic le is occu pied wit h a di scuss ion or ~ n m eperplexingexperiments that nre c lose ly related to the f mo u ~prohlem ra ised by Ein -



Oth e r T ypes o f Ex p lanation I 9

stein, Podol sky, and Rosen (1935). Given the recalc it rant character of thatproblem, the las t se nten ce in the d 'Espag nat qu o tation ma y just be th eunderstatement of the centur y. After pre se nting the se problems in some

detail , d'Espagnat then considers the possibility o f s imply denying that

science provides explanations, and rejec ts it. Having n o ted that the resultsof the experiments in question agree with the predictions o f quantum the ory ,he continues:

One conceivable respon se to the ex periment s is that their outcome

is incon seq uential. the re sults are merely what was expected . They

show that the theory is in agreement with experiment and so provid eno new infom1ation. Such a reaction would be highly superficial. (P .181)

In this judgment, I believe , d 'Espagnat is altogether correct. I shall saymore about these issues in chapter 9 .

It is now fashionable to say that science aims not merely at des cribin gthe world; it also provid es understanding, co mpr ehension , and enlight-enment . Science presumably accomplishes such high -so unding goals bysupplying scien tifi c explanations . The current attitude leaves us with a deep

and perplexing question, n amely , if explanation does involve somethingover and above mere description. just what sort of thing is it? The use of

s uch h onor ific near -sy nonyms as ' 'und e rs tanding , · ' ' 'comprehension ,

and enlightenment mak es it sou nd important and desirable , but doe snot help at all in the philo so phi cal ana lysi s of exp lanation - s c i e ntifi c orother. What, over and above descriptive knowledge of the world, is required in order to a c hieve under sta nd ing? This seems to me to pose ase rious philo sop hi ca l problem, especially for those who hold what must

be considered the most innuential contemporary theory concerning th enature of scientific explanation (see Salmo n , I978).

The main purpose of the present book is to examine the natur e ofscientific explanation . lt will become clear, as the discussion unf o ld s, thatthe issues to be considered ap ply not on ly to modem physics but r at he r tothe whole ran ge of sc ientifi c disciplines from arc haeology to zoo logy,

includin g a great many that occupy a lph abe tica lly intermediate positions.The approa c h wi ll be both i c o n o c l ~ t i cand cons tru ctiv e . shall level what

seem to me to be grave cr itic isms against what current ly qualifies as ' thereceived view,· but shall also tty to elaborate a more sat isfactory a lte rnative account.

OTH R TYP ES OF EXPLANAT ION

It i s adv isab le, I believe, to begin with some brief remarks circumscribingthe co nc ep t tlwt is the objec t o f our atte nti on. The tenn exp lanation



I 0 Scientific Exp lanation

has a number of uses, many of which are beyond the scope of the presentdiscussion. We are often asked, for example, to explain the meaning of aword; in such cases, somet hing akin to a dictionary definition may constitute an adequate response. Someone might ask for an explanation of the

meaning of a story oi· a metaphor; here a translation of figurative or metaphorical language into litera l terms seems called for A friend might askus please to explain how to find the location of a party;.in this instance,a set of detailed instructions or a map might do quite well. In none of

these cases is a scientific explanation of a natural phenomenon beingr e q u e ~ t e d t is cruc ially important to distinguish scientific exp lanationsfrom such other types. Michae l Scriven once comp lained that one of CarlG . Hempel 's models could not accommodate the case in which one 'ex

plains' by gestures to a Yugos lav garage m e h ~ n iwhat is ~ ~ r o ~ g:vit ll acar. I am in complete agreement w1th Hempel s response: Th1s IS likeobjecting to a metamathematical definition of proof on .the groun d th at itdoes not fit the use of the word 'proof' in 'the proof ot the puddmg IS mthe eating,' nor in '86 pronr Scotch' 1965 , p. 413). i\s Hempel remarks,a model of scientij;c explanat ion that did accommddate such cases wouldipso facto be defective.

It is worth noting that none of the foregoing requests for exp lanationswould normally be posed by asking a why-question. In some of the cases,we are asking what some thing means, or wh t is wrong with the car. Inother cases. we are asking how a mathematical proof goes, or how to getto the part y A request for a scientific explanation, in contrast, can alwaysbe reasonably posed by means of a why-question. If the request is notoriginally formulated in such terms, it can, I believe, be recast as a whyquestion without distortion of meaning (see Brombergcr, 1966). Indeed,

van Fraassen 's recent accou nt of scientific explanations characterizes themessentially as answers to why-quest ions (1980 , chap. 5).It is crucial to recognize. however, that not all- -o r even most - w hy-

question s are requests for scientific explanations. Why, we might ask. didone employee receive a larger raise than another? Because she had beenpaid less than a male colleague for doing the same kind of job. Tn thiscase, a moral nr le gal justification is the appropriate response. Why,someone might ask. did you go to the drugstore today? The answer, To

get some aspirin for my headache, constitutes a practical justificationWhen Job asked God why he had been s ingled out for such extraordinarymisfortune and suffering, he seems to have been seeking religious con-

solation 1t would. of course, spawn endless philosophical mischief toco nfu se justilication or consolation with scientific explanation. Moreover,it would bl a sadistic joke to offer a scientific explanation when consolationis snught. ro tell a bereaved wtdow, who asks why her husband was taken



Lap lacian Explanation I I

from her, that (as she is fully aware) he was decapitated in an automobileaccident and that decapitation is invariably fatal. would simply be wantnncruelty.

LAPLACIAN EXPLANATION

Although one particular philosophic a l account of scient ific explanationhas enjoyed considerable recent influenc e, there are others that have bothhistorical and co ntemporary significance. As a point of departure for discussion of various general conceptions of scientific explanation, I shouldlike to consider a historical example. Father Eusebio Francisco Kino wasa Jesuit missionary who, in the latter part of the seventeenth century.founded a series of missions in what is now northern Mexico and southernArizona. Kino was a highly educated man who was well trained in astronomy, cartography, and mathematics . While in Spain, just prior to hisjourney to North America, he observed the great comet of 1680 . Writingto a colleague, he began by giving a scientirically accurate description ofits appearance and motion; he then went on to comment upon its port e ntfor humanity:

It appears that this comet, which is so large that I do not know whetheror not the world has ever seen one like it or so vast, promises, signifies.and threatens many fatalities i t~ influence will not be favorable .And therefore it indicates many calamaties for all Europe . . . andsignifies many droughts , hunger, tempests. some earthquakes. greatdisorders for the human body. diswrds, wars, many epidemics, fevers,pests, and the deaths of a great many people. especially of some very

prominent persons. May God our Lord look upon us with eyes filledwith pity .And because this comet is so large it sig nifi es that its fatalities will

be more universal and involve more peoples, persons, and countries.And since it is lasting so long a time it indicates that its evilinfluence will afflict mortals for many years. (Bolton. 1960, pp. 62- 63)

In those days, even the well-educated saw comets as divine omens of

disaster. Kino was no ignorant, co un try parish priest.Just a few years later, in 1687, lsaa c Newton published the PrincipiaIn this work, comets were explained as planetlike objects that move aroundthe sun in highly eccentric orbits. The astronomer Edmund Halley - forwhom the great comet of 1682 was named, and who was instrumental insecuring the publication of the Principia composecl an ' 'Ode to Newton.'·which was prefixed to that work. In it he wrote:



12 I Scientific Exp lan atio n

Now we knowThe sharp ly vee ring ways of comets, onceA sou rce of dread. nor longer do we qu a ilBeneat h appenranccs of bearded stars.

(Newto n , 1947 , p. xiv)

The return of Hal ley's comet in 1759 marked a major triumph of classica l

me cha nics.Writing at the be g inning of the nin e teenth century, P. S. Laplace, the

famous advocate of mechan istic determinism, drew an e loquent co ntra stbetwe en th e scie ntifi c unders tandin g of comets provided by Newlonianphysics and th e superstitious lac k of und ersta ndin g that led people, in pre

Newtonian tim es, to regard them as mysterious s igns se nt by inscrutablesup erna tura l power s.

But as these phenomena occurri ng and di sa pp earing at long int erva ls ,seemed to oppose the order of nature , it was s upp osed that H eave n ,irritated by the crimes o f ea rth , had created them to an nounce it s vengeance . Thus, the long tail of the comet of 1456 spread terror throughoutEurope. . . . This star aft e r four revolution s has excited among us a

very different interest. The knowledge of the Jaws of the system of theworld ~ q u i r e din the int erval had dissipated the fears begotten by theignorance of the true relat io nship of man to the uni ver se; and Halley .,havin g recognized the identit y of this comet with those of 1531, 1607,and 1682, annou nced its nex t re turn for the e nd of the yea r 1758 or thebeginning of the year 1759. The learned world awai te d with impatiencethis return which was to confirm one of th e g rea tes t discov eries thathave been made in the sciences. (Laplace , 195 J , p . 5)

Although the e limination of super stitiou s fears is, undoubtedly , one of

the major benefit s of scie ntific ex planations a n d I think it is importantto keep th at fact in mind at time s, s uch as the pr ese nt , when irr at ionalismseems rampant 1- it can hardly be regar ded as their main purpose. Forsuppose that cornets had been considered good omens much to be hopedfor, and so ur ces of great joy w he n they ap pea red - but equally hap hazar das far as anyone could tell . As objec ts of rejoicing rather than of fear , they

would not have been any less mys teriou s .There is some temptation, J believe, to view sc ie ntifi c explanation so le ly

in psych o logical t e rms . We beg in to wonder abou t some phenomenon -

• /\s jus t one indi cation . co n, ider the remarkable strengtl1 o f the move ment to mandatethe teachin g . in publi c sc hools. of the bib lica l acco unt of the or ig in of the wor ld. and of lifewithin it. utu.lc r the title · cre:1t ion-sr iencc. Sre (Martin Ganfncr, I081 fnr man y additional~ x m p l c . and (S ingn :mrl Acna,si. 1 R I l for an inter es tin g disr11ss ion .



L.aplacian Exp lanat ion I 13

be it familiar or unf amiliar, an ohject of jo y or an object of fear - and we

are intell ect ua lly (and perhaps emotionally) ill at ease until we can summona set of facts that we take to explain it. Whether we have successfullyexp la ined th e phenomenon. on this v iew. depends entire ly upon whetherw ~have overcome our psychological uneasin es<; and can feel comfortabl ewith the phenomenon that origi na ll y se t u s to wonderi ng. We need notobject to thi s conception merely o n the ground th at people often in vokefalse beliefs and feel comf01tab le with th e 'exp lanatio n ' thus providedas w hen the behavior of a me mb e r of a racial minority is 'explained' interm s of a racial stereotype. We can . quite cons isten tl y wit h this approach.insi s t that adequate ex pl ana tions mu st rest upon true ex planat ory bases .Nor need we object on the gro und that supemat ural 'ex pl anations ' are often

psychologically appealing . Again, we ca n in sis t that the explanation begrounded in scientific fact. Eve n with th ose restrictions, however, the view

that scientific explanation cons ists in release from psyc hological un easinessis unacceptable for two r easo ns . Fir s t. we must surely require that th ere

be some sort of objective relationship between th e explanatory facts andthe fact -to -be -explained . Eve n if a person were perfectly content with an'exp lanation ' of the occurren ce of storms in term s of falling barometricreadings, we should still say that th e behavior of the barometer fail s ob

jec tive ly to exp lain such facts . We must , in stead , appeal to meteorologic a lcondi tions. Second, not only is ther e the danger that people will feelsa tisfied with scientifically def ective exp lana tion s; the re is also the riskthat they will be un satisfied with legi ti mate scie ntific explanations . Ayearning for anthropomorphic ex plan atio ns of all kinds of natur al phenomena - for example, the demand th at every explanation in vo lve consc ious purpo ses - some tim es lead s people to conc lude that physi cs doesn't

re lly explain anything at all (reca ll the letter about the quark theory that

was previousl y quoted in part). Some people have rejected exp lana tionsfurnish ed by general relativity on the ground that they cannot visua lize acurved four-dimensional space - tim e. The psych o logi ca l interpretation ofscientific explanation is patently inad equate.

Another conception, equa ll y inadequ ate, may be suggested by Lap lace's

remarks . It is sometimes c laimed that sc ientific exp lana tion consis ts inreducing the unf am ili ar to the familiar as when Newton exp lained comet sby showi ng that they are objects that behave in esse nti a lly the same way

as planets, whose motion s were by then quit e well known. Olbers ' paradoxconstitutes a clear refutation of that th esis. In 1826. the astronomer HeinrichOlbers showed that, on Newtonian pr in c iples , the entir e ni g ht sky shouldbe brilliantly luminous- eve ry region shining at least as brightly as themidday su n .2 Yet t he fact is that for the mo s t part , except for the moon

2 Edrnund I I a lley anti cipa ted O lhcr' when he wrpte ( 1720 . If the numb er nf th e Fixt

I



14 Sci e ntific Exp lanation

and a few points of light coming from visible stars or planets, the nightsky is dark. The explanation of this familiar fact. it turns out, takes usinto some of the more esoteric reaches of modern cosmology, involvingsuch concepts as the mean free path of photons and non-Euclidean geometry. Jn thi s case, 'the familiar is explained in terms of highly unfamiliarnotions. Other examples, which exhibit a simi lar character, are easilyfound. It is a familiar fact that a meta l table will support a coffee cup, butthe explanation of this fact involves the details of atomic structure and theappeal to conceptions as unfamiliar to everyday expetience as the Pauliexclusion principle. We need not linger longer over this conception , eventhough it has had prominent adherents. 3

Still another characterization of scientific explanation is suggested byLaplace's discu ss ion of comets, name ly, that exp lanation consists in showing that what appears to be haphazard does actually exhibit some regularity.Although this appeal to regularities in nature is a step in the right direction,it certainly cannot be the whole story. The basic reason is that althoughsome r e u larities have explanatory power others simply cry out to beexplained. Newton s theories explai ned comets by bringing them withi nthe scope of the laws of universa l gravitation and motion. In the co ntext

of classical physics, that sort of exp lanation is extremely satisfying, asLaplace made clear. At the same time, Newton's theories a lso explai nedthe tide s. The regular ebb and flow of the tides had been known to marinersfor centuries prior to Newton; moreover, the correlation between the behavior of the tides and the position and phase of the moon was well known .These are the sorts of regularities that arouse, rather than satisfy, ourintellectual curiosity. Without Newtonian physics, the tides were not reallyunderstood at all .

Consider another example . Suppose it has been noticed that on dayswhen clothes hung out on the line tend to dry more slowly than usual,airplanes at nearby airports require greater than averag; distances to getoff of the ground. Mr. Smith, whose business involves the use of a smallairplane, complains to his wife one evening about the difficulty he hadgetting airborne that day. Even if both of them are fully aware of thepreviously mentioned regularity, it could hardly be considered an expla-

Stars were more than finite. the whole superfi c ies of their apparent Sphere [i .e .. the sky]would be luminous . Quot ed in (Misner et al., 1973 . p. 756).

' For eKample , Holton and Brush (1973, p. 185) remark , Perhap s it is not too frivolousto ho ld that 'to explain ' mea ns to reduc e to the familiar , to es tablish a re lat ionship betweenvdutt is to be cxp laineci and the (correctly or incorrec tly) unquestioned preconc ep tions. ··

r i d i ~ l l l l l l t 9 2 R p. 37 eK prcsscs si milar view, J be li eve that e: \amination will show thatthe esse nce of an ex plan at ion co ns ists in reducing a situati on to eleme nts with whi ch we ares • familiar that we :1cccpt them a s a matter of course. sn that our curiosity rests .··



Three Basic Concepti o ns 15

nation of his difficulty for her to mention that the wash had taken anunusually long time to dry that day .

I am dwelling upon Laplace s conception of scientific explanation notmerely to exhibit and refute what I take to be patently mistaken notions ,but primarily because he provided an acco unt that was especially appropriate w ithin the context of classical physics. Without making any claims

about the historical influence of Laplace, l suspect that the conceptions heexpressed have had a large nd not altogether beneficial--effect uponcontemporary thought about scientific explanation. In the remainder of thischapter, I should like to support and amplify this general assessment . Inso doing, I sha ll extract three distinct general conceptions of scientificexp lanation, all of which have historical as well as contemporary significance .

THR BASIC CONCEPTIONS

Laplace attributed our. ability to explain comets to our knowledge of thelaws of nature. Twentieth -ce ntury philosophers have echoed that view bymaintaining that, with the aid o suitable initial conditions an event is

explained by subsumingit

under one or more laws o nature. If these lawsare regarded as deterministic, th .s fonnulation becomes hardly more thana translation into more up-to -date and less colorful tem1inology of Lap lace s

famous statement:

Given for one instant an intelligence which could comprehend all of the.forces by which nature is animated and the respective situation of thebeings who compose it- an intelligence sufficiently vast to submit allthese data to analysis- i t would embrace in the same formula the movements of the greatest bodies of the universe and those of the lightestatom; for it, nothing would be uncertain and the future, as the past,would be present to its eyes. (1951. p . 4)

Such an intelligence would exemplify the highest degree of scientificknowledge ; it would, on Laplace s view, be able to provide a completescientific explanation of any occurrence whatsoever.

There are, it seems to me, at least three distinct ways in which suchLaplacian explanations can be construed. In order to relate them to themodern context, we will need to introdu ce a bit of technical tem1inology .It is customary, nowadays, to refer to the event- to -be-explained as theexp lanandum -eve nt and to the s tat emen t that such an event has occuned

' Those reader s who ar e unarquaint ed with this exampl e ca n find an explanatinn nf theregularity in clwpter 9, pp . 26R- fio.



6 I Scientific Explanation

as the exp/anandwn-statement. Those facts-both particular and genera lthat are invoked to provide the exp lanation are known as the exp lanans.

f we want to refer specifica lly to statements that express such facts, wemay speak of the explanans-statements. The exp lanans and the explanandum taken together constitute the explanation. Let us now lo ok at thethree conceptions .

1) Epistemic conception. Suppose that we attempt to exp lai n some

occurrence, such as the appearance of a particular comet at a particularplace a nd time. By c itin g certai n law s, together with suitab le initial conditions, we can deduce the occurrence of the event -to-be-explained. Byemploying observational data collected when his comet ar.reared in 1682,

Halley predicted its return in 1 7 5 9 These data, along with the law semployed in the deduction, su bsequently provid ed an explanation of thatappearance of th e comet. This explanation could be described as an argument to the effect that th e event-to-be-exp lain ed was to be expected byvirtue o the explanatory facts. The key to this sort of explanation is nomicexpectahility. An eve nt that is quite unexpected in the ahsence of knowledgeof the explanatory facts is rendered expectable on the basis of lawfulconnections wi th other facts . Nomic expectability as thus characterized isclearly an epistemological concept. On this view, we can say that there isa relation of logical ne ces sity between the laws and the initial conditionson the one hand, and the event-to-be-explained on the other-though itwould be more accurate to say that the relation of logica l necessity hold sbetween the explanans-stateme nt s and the ex pl anandum -state ment.

(2) Modal co nception. Under the same circumsta nces we can say, alternatively , that because of the lawfu l relations between the antecedent

conditions and the event-to-he-explained th ere is a re lation of nomologicalnecessity between them. In La place's Essay the discussion of deteTminismis introduced by the fo llowi ng remarks:

All events, even those which on account of their insignificance do notseem to follow the great laws of nature, are a result of it just as necessruilyas the revolutions of the su n . In ignorance of the ties which unite suchevents to the entire system of the universe. they have been made to

depend upon final causes or upon hazard but these imaginary causeshave gradually receded with the widening bounds of knowledge anddi sappear entirely before so und philosophy, which sees in them onlythe expression of our ignorance of the true c a u ~ e s( 195 t, p . 3)

' Actually Hallcy did not make a very precise prediction , for he did not take account ofthe pcrturhations in the orb it due to Jupiter and Saturn . This was done by Clairaut; see(L.nplacc. l ' l ) l . p. 6).



Three Basic Conceptions I 17

Nomological necessity, it might be said, derives from the laws of naturein much the same way as logical necessity rests upon the laws of logic.In the absence o knowl edge o the explanatory facts. the explanandllmevent (the appearance of the comet) was something that might not have

occurred or all w know; given the exp lanatory facts it had to occur. Theexplanation exhibits the nomolog ical necessity of th e fact-to-be-explained,given the explanatory facts. Viewing the matter in this way, one need notmaintain that a n exp lan ation is an argumen t showing that the explanandumevent h ad to occ ur, given the initial conditions. A lthough a deductiveargument can be cons tru cted (as in the epistemic account) within which arelation of logical entailment obtains, an exp lan at ion n eed not be regardedas such an argument. or as any ki nd of argument at all. In comparing the

epistemic and modal conceptions, it is important to be clear on the rolesof the two kinds of necessity. In the epistemic conceptio n, the relation of

logical nece ssi ty obtains between th e e ntir e explanans and the explanandumby virtue of the laws of deductive logic. In the modal conception, therelation of physical necessity holds between particular antecedent condition s and the explanandum-event by virtue of the genera l law s, which weare taking to be part of the explanans.r.

(3) Ontic conception. There is st ill another way of looking at Laplacianexplanations. lf the univ erse is, in fact, deterministic , then nature is governed by str ict laws that cons titut e natural regularities. Law-statementsdescribe th ese regularities. Such regularities endow the world with patternsthat can be discovered by scientific inv es tigation. and that can be exp loitedfor purposes of scient ific exp lanation. To explain an eve nt t o relate theevent-to-be-explained to some antecedent conditions by means of lawsis to fit the exp lan andum-event into a discernib le pattern. This view seems

to be present in Lap ace's thought, for he remarks that comets seemed

6 The contrast being suggested is we ll illu strnted by a controversy beTween Hempel andScriven concern ing the role of laws in scientific ex planation . As we have seen, Hernp e l(1965) insists that ge nera l laws be prese nt in the explanans. Scriven (though he is n0t aproponent of the mooal conception) argues that a set of particular antecedent conditions mayconstitute an adequate explanation of a particular even t; consequently, the explanans nee<.not inc lude reference to any general laws . A law that provid es a connection between theexplanans and the explanandum cons titute s a "role -justifying ground" for the explanation

by showing, roughly speaking, that the explanans is explanatorily relevant to the explanandurn. For Hempcl, the laws of logic-w hich provide the relation of relevance of theexplanans to the explanandum-are not part of tlH exp lanation, but can be called upon t ojustify the claim that a given explanans has explanatory force with respect to some explanandum. Scriven invokes similar considerations to argue that general laws of nature shou ldremain outside of scientific explanations to he called upon , if necessary, to support the claimThat a given explanation is adequate. See (Scriven . I l59) for details: Jlempel"s reply is givenin his (1965. pp. 359-364).



l \ Scientific Exp lana tio n

to oppose the order of nature before we knew bow to exp lain th em, b ~ ~

th at subsequent ' 'know ledge of the laws of the system of the world.provided und erstanding of th em ( 195 1, p . 5). Mor eove r , as noted pr e vious ly, he speaks of " the ties which unite such eve nts to the enttr e ~ y ~ t e

of the universe'' (1951, p . 3). Because of the universal (non s tattstical)

character of the laws involved in Laplacian explanations, we can also saythat given certain portions of the pattern of events, and given ~ h el a ~ f u l

relations to wh ich the constituents of the patt erns conform, other portiOn sof the pattern of eve nt s must ha ve certain c h a r a c t ~ ri s t ic s . o ~ k i ng at ~ x -

plan ation in this way, we might say that to exp lam .an e v e~ t S t ~ exlubztit as occ up ying its (nomologica lly necessary) pl ace 111 th e d s cer mbl e pat -

ter ns nf the wor ld .

The se three general conceptions of scientific e x pl a na ti o ~a ~ see.m to .goback at leas t to Aristotle. We have already r e marked on hts J d e n t l ~ c a t t ~ n

of certain sorts of sy llogi sms as exp lanations; thi s conforms to th e ep tstemtcconception that regards explanations as deductive argument s. : . He seemsto be expressi ng the modal conception w h e ~he r e m ~ r k stha.t the properobject of unqualified scientific knowl edge s so methmg w h t c ~cannot beother than it is " (Posterior Analytics, 1. 2. 7lbl4 - l6). And m the samecontext, discu ss ing the nature of the syllogism that yields "scientif ic n ~ w l -

edge, he says. The premises must be the causes of the c o n c ~ u s J ? nbetter known than it, and prior to it; its causes, s ince we po ssess sctenttficknowledge of a thing only when we know its cause; p r i o ~in order to bec a u s e s ~antecedently known, this antecedent knowledge betng not ?.ur .n:ereunderstanding of the meaning, but knowledge of the fact a s well (1b1d . ,71 b29 -33). These remarks suggest an antic conception . . .

In the tw en tiet h ce ntury , we still find th e same three nottons figunngprominently in philosophical di sc uss ions of ~ ci c n t i lc e ~ p l a r . w t i o ~ .Theepistemic concep tion represents the currently recetved. VIew; whtch hasbeen advocated by such influential philosopher s as BraJthwatte.' Hempel,Nag el, and Popper. t was succinctly formulated by Hempel m the fol -

lo wing way:

' In the Posterior Anall lics (1928 , t 2 . 71bl8-24) , Ar istotle w rites : By demonstration1 mean a sy llog ism prod-uctive of sc ientific knowledge. a sy llogism , that is , the g ~ a s p ofwhi ch is eo ipso suc h knowledg e. Assum ing that my thesis as to the nature o ~sctent.tficknowledge is co rrect, the premises of demonstrated knowledg e must be true, pnmary. Immediate , better known than and prior to the co nclusion. w ~ i c his f u r t ~ e rr e a t e ~to t h e ~a ~

effect to cause . Unless these co nditi ons are sa tis fi ed, the bas te truths wttl not be a p p r o p ~ a t e

to the conclusio n. Syllogi sm there may indeed be without lhese co ndition s , but such sy llogtsm ,not being produ ctive of sc ientific know ledge, w ill not be demonstration." Richard Jeffrey(1969) offe rs an illumin atin compar ison between this Aristotelia n view and Hempe l's D-N

account.



1An] exp la natory account may be regarded as an argument to the effectthat the event to be explained . . . was to be expected by reason o fcerta in explanatory facts . These may be divid ed into two groups: (i)particular facts and (ii) unif onnit ies ex pr esse d by genera l law s. ( 1962a,

p 10)

The modal co ncep ti on ha s bee n clearly a ffirmed by D . H . Melior: Thethesis is that we call for explanation only of what, although we know itis so, might have been oth erwise for all e lse of some suitabl e sort we

know'' (1976, p . 234). In what does a n exp lana tion co nsis t?

We want to know why what mi ght n ot have happ e ned n o nethel ess did.Causa l explanation clo ses the gap by deducin g what happen e d fromknown earlier events and detetm ini s tic laws. So in this r es pect it satisfies

the demand for exp lan ation : w hat follo ws fro m what is true must alsobe true. Given the causal explanans , thin gs co uld not have happ enedotherwise than the exp lan andum says . .( 1976, p . 235. ita lics added)

G. H . von Wright 1 971) giv es concise expression to thi s same conception :' 'What make s a deductive-nomological explanation 'ex plain.' is, one mightsay , that it tells us why E had to be (occur), why E was necessary oncethe basis [body of expla nat ory facts is there and the la ws are accepted "

(p . 13, italics in original). This same vi ew can be found explicitly inC . S . Peirce (1932, 2 :776).

T he antic conception is th e one for which I shall be arguing. In Salmon(1977a, p. 162), I offered the following characterization: T o give scientific explanations is to show how events fit into the causal structureof the world.' ' Hempel summarizes the import of his major monographicessay, Aspects of Scientific Ex plan a tion· · ( 196 5a, p . 488), in rathersimilar terms: The central theme of thi s essay ha s been , briefly, that a llscie ntific explanation involves, explicitly or by implication, a subsumpti onof its subject matter under general regularitie s; that it seeks to provide asystematic under s tanding of empirical phenomena by showi ng that th ey fitinto a nomic nexus . " I find thi s s tat ement by Hempel in ai most complete

accord with the viewpoint I shall be advocating; m y s ugge stion for modifica tion would be to substitute the words ~ o wthey fit into a causalnexus for that they fit into a nomic nexus . It seems to me that Hernpel

began the Aspects article with stateme nts clearly ind icating that heembraced the epistemic conception, but he end ed with a summary thatseems closer to the ontic conception. Because these three co nception s hadno t been explicitly formulated and di stingui shed at the tim e of his writing,he was , I think . unaware of any con flic t. As we s ha ll see in subsequent



20 I S c ie ntifi c Ex pl a na tio n

chapters. there a re profound differences, especia lly in the context of statistical explanatio n .

Those phi losophers who have adopted the ontic co nce ption of sc ie ntificexp lanation have ge nerally regarded the patt ern in to w hi ch events are tobe fit as a causal pattem . Th is feature of the view is brought out ex plicitlyin the quotation from Aristo tle. It certain ly was prese nt in Lapla ce 's mindwhen he wrote, Present eve nts are connected w ith prece din g ones by atie ba sed upon the evide nt princip le that a th ing ca nn o t occur wi thout acause wh ic h produces it (195 1, p . 3), and W e oug ht then t o regar d thepresent state of the univ erse as the effect of its anter ior s tat e and as thecause of the one w hi c h is to follo w ( 1951 , p. 4). I t was also exp licit inmy form ul a tion quoted previous ly. Hemp e l , howeve r , does not share thisnoti o n ; for him the pattem is lawful (nomic), but the Jaws involved neednot be ca usal law s (1965, pp. 352 - 354). In view of we ll -known Humeanpr o bl ems associated with causa lit y, it might see m desirable to try to avoidref erence to ca usa l law s in dealing with scie ntifi c exp lanation . Nevert heless, _L hal l try to show tha t we need not purg e the ca usa l noti o ns; indeed ,I sha ll arg ue th a t the y are required for an adequate theory of sc iei1tiHcexp lana tion . In o rde r to implem e nt the causa l vers ion of the ontic conception, ho weve r, it wi ll he necessary to examine the nature of causalre lations with considerable care, and to show ho w they can be e mployedunobj ec tio nab ly in a th eo ry of scien tific exp lan ation . This probl e m wi ll bepostponed unt il chapters 5- 7.

The foregoing three ways of thinking about sc ien tific explanation mayseem more or less equiva lent ith somewhat di s tinct emphases perhaps but hardly more than different verbal formulati o ns. This is tru e as long aswe are talking abou t the kind of ex planation that invo lve s appeal to universal laws on ly. A striki ng divergence will appear, however , w hen weconsider explanations that invoke s tati stica l (or probabilistic) law s. In thedeterministic framework of Laplace's thought, all of the fundamental lawsof nature are taken to be s trictly univer sa l; any appeal to probabilities ismerely Ia reflection of human ign orance. In twentieth-century science, thesituat ion is qu ite different. There is a strong presumption in contemporaryphysic s that some of the basic la ws of nature ma y be iri·educibly stat isticalthat probability re la tions may const itut e a fundamental feature of the phys

ical world . There is , to be sure, some di sagreemen t as to whether determini sm is true or false-whether modem physics requires an indetenninistic interpreta tion . I do not want to prejud ge this issue. In the attempt toelaborate a phi losophical theory of scientific explanation, it seems to me ,we must try to construc t one th at will be viable in ei th er case. Therefore,we must leave open the possibility th at some sc ient ific exp lanations wi llbe unavoidably statistical. This means th at we must pay carefu l atte nti onto the natur e o f stati stical explanat ion .



An Out li ne of Strategy I 21

N O U T L I N E f ST R AT EGY

Much of the co ntemporary li terature on scientific explanation arise sdirectly o r indir ec tl y in response to the classic 1948 Hempei-Oppenhei mpaper, Studies in the Logic of E xplanation . " 8 In it the au th ors attemptto provide a precise explication of what has come to be known as thed e d ~ c t iv e n o ~ 1 0 o g i c a or D-N model of scie ntific exp lanation. They didn o ~mvent th1s mode of scient ific explanat ion , nor were the y the firstphlio so ph ers to a ttemp t to charac ter ize it . As mentioned previou sly, itsroot s go back at least to Ar istotle, and it is strong ly s ugg es ted in su c hworks as Arnauld's The r t o f Thinking ( Po rt -Royal Logi c ) and Laplace 's Philos ophi ca l Essay on Prohahilities In no ne of the anticipationsby these or other a uth ors, however, do we find the precision and detail of

the Hempei -Oppe nheim account. One might almost say that 1948 marksthe division between th e prehistory and the hi story of the philo so phi ca lstudy of scientific explanation. When o the r s uch influ ential phil oso ph ers

as R . B. Braithwait e ( 1953), Emest Nagel (1961), and Karl R . Popper(1935, 1959) espoused a similar accou nt of deduct ive explanation, it Hchievedvirtual ly the status of a ' rec e ived view . 9

A ~ c o r d i n gto th e ' receiv ed view,' par ti c ular fa cts are explained by subsummg them under genera l laws, while ge neral regu laritie s are explainedby subsumption under still broader laws. If a particular fact is succes sfull ysub sumed under a lawful genera lization, it is, on thi s view , complet e lyexplained . On e ca n legitim ate ly ask for a n exp lan ation of the general Jawthat figure s in the explanation, but an exp lanation of the general Jaw wouldbe a differ e nt and additional exp lanati on, not an essential part of the original~ x p l a n a t i o nof the particula r fact. For exam ple , to explain why this particular penny co ndu c ts electricity, it suffices to point out that it is composedof copper and that all copper o bjec ts co nduct electricity. If we are askedto explain why copper conducts e lectricity , we may give a further distinctexplanation in term s of the fact that copper is a met al with conductione lectro ns that are not tightly bound to individua l a toms an d are free tomove wh e n an electric potential is app lied .

Most proponents of this subs umpti on theory maintain that some eventscan he explained statist ica lly by subsurnrt i0n und e r statis tical l aws i n muchthe same way th at o th er events - suc h as the fact that the penny just inserted

' See (Rescher. 1970 for an exten sive bibliography on sc ientific explanation up to thedate of its pub Iicat ion.

9 Althoug h _Popper's Lo8ik der Fors chrmg ( 1935 contains an important anticipation of theD-N mode l. It does not provide as precise an nnalysis as was embodied in He mpel and? ppenh enn. 1_948 . Moreover, Popper's v iews on scient ific explanation were not wicle lytnfluen l al until the E ng lish translation (Popper. 1959 of his 1935 book appeared . Jt is forthese reasons that l chose t948, rather than 1935, ~ the critic al point of divi s ion betwe enthe history and the prehistory of t l1c subjec t.



22 I Scientific Explanation

behind the fuse conducts electricity-are explained by appeal to universallaws. Thus we can explain the fact that a particular window was brokenby pointipg out that it was struck by a flying baseball, eve n though notall, but only most, windows so struck will shatter. .

Although l disagreed from the beginning with the proponents of thestandard suhsumption view about the nature of the relation of suhsumptionof particular facts under universal or statistical generalizations, J did forsome time accept the notion that suitable subsumption under generalizationsis sufficient to explain particular facts. In Statistical Explanation nd Statistical Relevance (Salmon et al., 1971), 1 tried to give a detailed accountof what seemed to me the appropriate way to subsume facts under generallaws for purposes of explanation. This effort led to the elaboration of thestatistical-relevance (S-R) model of scientific explanation. As the namesuggests, statistical relevance relations play a key role in this model ofscientific explanation.

Subsequent reflection has convinced me that subsump tion of the foregoing sort is on ly part--not all-Qf what is involved in the explanation ofparticular facts. lt now seems to me that explanation is a two -tiered affair.At the most basic leve l, it is necessary, for purposes of explanation, tosubsume the event-to-be-exp lained under an appropriate set of statistica lrelevance relations, much as was required under the S-R model. At thesecond level, it seems to me, the statistica l relevance re lations that areinvoked at the first level must be explained in terms of causal relations.The explanation, on this view, is incomplete until the causal componentsof the second level have been provided. This constitutes a sharp divergencefrom the approach of Hempel, who explicitly rejects the demand for causalJaws 1965. pp . 352-354).

Itwould be advisable, I believe, to adopt an approach similar to onesuggested by Wolfgang Stegmi.iller I 973, p. 345), who characterized the

kind of subsumpt ion under statistical relevance relations provided by theS-R model as statistical analysis rather than statistical explanation.The latter term is reserved for the entity that c o m p r i s e ~both the statist icalrelevance level and the causal level as well . As (Hump hreys, 1981 , 1983)and (Rogers, 1981 persuasive ly argue, statistica l analyses have importantuses, but they fall short of providing genuine scientific understanding. To

emphasize this point, I shall use the term S-R basis to refer to the statisticalcomponent of an explanat ion . 10

The remainder of the present book is divided into two main parts,

' In relinquishing the thes is that the S-R model provides an adequate characterization ofsc ientifi c ex planation , I accept as valid most of the criticisms lcve lcd against it by Achinstein

1 9 8 ~ ) .h ~ e criticisms do not, however, undermin e thl' utilit y nf the S-R bas is as a foundatiorl fnr scien tific ex pl anat ions .



An Outline of Strategy I 23

corresponding to the two tiers just mentioned. Chapters 2-3 which constitute the first part, deal essentially with the S-R basis; it is regarded asan indispensable portion of the present account of scientific explanation ,even though it is no longer taken as a complete model. These two chapterscontain a numb er of important revisions of the S-R model itself, vis-a-visprevious presentations, as well as a good deal of supplementary material.Chapter 3 deals with the dif'ficult concep t of physical randomncss thatis, objective homog eneity of reference classes - which is required to implement an ontic treatment of statistical explanation adequate to the possiblyindetennini stic context of contemporary science. The result, I hope, is asubstantially improved version of the S-R model-Qne that can provide asatisfactory basis for the second level. These two chapters contain. roughly,all that can be said, as far as I am aware, regarding scientific explanationof particular facts without invok ing causal considerations.

Chapter 4 is a transitional discuss .ion of the three general conceptions(introduced in this chapter) in the light of statistica l considerations . Itconstitutes a serio us attempt to provide a much-needed clarification of theexplicandum. The conclusion drawn from the discussion is that the onticconception is the only acceptable one.

Chap ters 5- 8, which make up the second main part, deal explicitly withcausality in scientific explanation. In order to achieve the goal of exp licatin gthe role of causality in scientific explanation, it is necessary to develop atheory of causality that, though it borrows heavily from other authors,incorporates various novel elements. Among its conspicuous (though notnecessarily novel) features are 1) it is a probabilistic or statistical concept,(2) it places great emphasis upon the distinction between causal processesand causal interactions and (3) it takes processes to be more fundamentalthan events. The reader will have to judge for herself or himself concerningthe adequacy of the account of causality offered in these chapters. and itsfe1tility in providing an improved theory of scientific explanation. lt is myhope that the result is a theory of scientific explanation that constitutes asignificant advance beyond its predecessors.

Chapter 9, the concluding chapter, deals with some general features ofmy approach to scientific exp lanation. including some of its shortcomingsand limitation s. It points the way, I believe, to the kind of research that

should extend and deepen our philo sop hical unders tanding of scientificexplanation .

scientific explanation and the causal structure of the world - wesley c. salmon(cut)

Documents