explaining explanation - philarchive

d d d

Explaining Explanation

David-Hillel Ruben offers a discussion of some of the main historicalattempts to explain the concept of explanation examining the worksof Plato Aristotle John Stuart Mill and Carl Hempel Building onand developing the insights of these historical figures he introducesan elaboration and defense of his own solution

In this volume Ruben relates the concept of explanation to bothepistemological and metaphysical issues Not content to confine theconcept to the realm of philosophy of science he examines it within a farmore broadly conceived theory of knowledge He concludes with his ownoriginal and challenging explanation of explanation

Explaining Explanation will be read with interest by students of generalphilosophy as well as those specializing in the philosophy of science andscholars with a more advanced level of interest

Private OwnershipReligious Belief and the WillRationalityThe Rational Foundations of EthicsMoral KnowledgeMind-Body Identity TheoriesPractical ReasoningPersonal IdentityThe InfiniteThought and LanguageHuman ConsciousnessExplaining ExplanationThe Nature of ArtThe Implications of DeterminismWeakness of the WillKnowledge of the External WorldIf P Then Q Conditionals and

the foundations of reasoningPolitical FreedomScepticismKnowledge and BeliefThe Existence of the WorldNaming and Reference From

word to object

Also available in paperback

James OGrunebaumLouis PPojmanHarold JBrownTLSSpriggeAlan GoldmanCynthia MacdonaldRobert AudiHarold WNoonanAWMooreJulius MoravcsikAlastair HannayDavid-Hillel RubenALCotheyRoy WeatherfordJustin GoslingBruce Aune

David HSanfordGeorge GBrenkertChristopher HookwayFrederick FSchmittReinhardt Grossman

RJNelson

The Problems of PhilosophyTheir Past and Present

General Editor Ted HonderichGrote Professor of the Philosophy ofMind and LogicUniversity College London

Each book in this series is written to bring into view and to deal witha great or significant problem of philosophy The books are intendedto be accessible to undergraduates in philosophy and to other readersand to advance the subject making a contribution to it

The first part of each book presents the history of the problem inquestion in some cases its recent past The second part of a contemporaryand analytic kind defends and elaborates the authorrsquos preferred solution


David-Hillel RubenSenior Lecturer in PhilosophyThe London School of Economics and Political Science

London and New York

First published 1990 by Routledge

This edition published in the Taylor amp Francis e-Library 2004

First published in paperback in 1992by Routledge

11 New Fetter Lane London EC4P 4EE

Simultaneously published in the USA and Canadaby Routledge

a division of Routledge Chapman and Hall Inc29 West 35th Street New York NY 10001

copy 1990 1992 David-Hillel Ruben

All rights reserved No part of this book may be reprintedor reproduced or utilized in any form or by any electronic

mechanical or other means now known or hereafterinvented including photocopying and recording or in anyinformation storage or retrieval system without permission

in writing from the publishers

British Library Cataloguing in Publication Data

Ruben David-HillelExplaining explanationmdash(The problems of philosophy)

1 ExplanationI Title II Series

160

Library of Congress Cataloging in Publication Data

also available

ISBN 0-203-16930-1 Master e-book ISBN

ISBN 0-203-26475-4 (Adobe eReader Format)ISBN 0-415-08765-1 (Print Edition)

For my parentsBlair S Ruben

Sylvia Ginsberg Ruben

Hear my son the instruction ofthy father

And forsake not the teaching of thy mother

vii

Contents

Preface and Acknowledgements ix

I Getting our Bearings 1Some explanations 3Process and product 6The methodology of explaining explanation 9Restricting the scope of the analysis 15Scientific and ordinary explanation 16Partial and full explanation 19Bad explanations and no explanations 21Some terminology 23Theories of explanation 25Dispensing with contrastives 39

II Plato on Explanation 45The Phaedo 47Platonic explanantia and explananda 51Problems for the physical explainers 53Some terminology 56Platorsquos Principles 58Platorsquos (PP2) 64Platorsquos (PP1) 66The Theaetetus 72Summary 75

III Aristotle on Explanation 77The doctrine of the four causes 77Does Aristotle have a general account of explanation 83Incidental and per se causes 87Necessitation and laws in explanation 93Aristotle on scientific explanation 95Aristotlersquos demonstrations 101Summary 108

viii


IV Mill and Hempel on Explanation 110Millrsquos account laws of coexistence and succession 115Millrsquos account the symmetry thesis 123Mill on ultimate explanations 125Mill on deduction and explanation 129Hempelrsquos account of scientific explanation 138Hempelrsquos methodology 141Hempel on the symmetry thesis 145Hempel on inductive-statistical explanation 149Hempel on epistemic ambiguity 152Summary 154

V The Ontology of Explanation 155Explanation and epistemology 155Extensionality and the slingshot 156The relata of the explanation relation 160Explaining facts 168The non-extensionality of facts 171Facts worldly or wordy 172The co-typical predicate extensionality of facts 173The name transparency of facts 177

VI Arguments Laws and Explanation 181The standard counterexamples irrelevance 183The standard counterexamples symmetry 191A proposed cure and its problems the causal condition 192Generalizations get their revenge 205

VII A Realist Theory of Explanation 209Are all singular explanations causal explanations 211What would make an explanation non-causal 217Identity and explanation 218Are there other non-causal singular explanations 222Disposition explanations 225Again determinative high and low dependencyexplanations 230

Notes 234

Bibliography 256

Name Index 262

Subject Index 264

ix

Preface and Acknowledgements

This book is written in the conviction that the concept of explanationshould not be exclusively hijacked by the philosophy of the naturalsciences As I repeat often in the following like knowledgeexplanation is an epistemic concept and therefore has a philosophicallocation within the theory of knowledge widely conceived Thephilosophy of science has great relevance for a theory of explanationjust as it does for discussions of knowledge But it is not the soleproprietor of either concept

It is a pleasure to acknowledge the many debts I have incurred in thewriting of this book A Nuffield Foundation Fellowship for the period ofJanuary-April 1988 and a grant from the Suntory-Toyota InternationalCentre for Economics and Related Disciplines which funded a period ofleave from January to April 1989 were both invaluable in providing mewith time to write the book I am extremely grateful for their help andwish to thank them publicly for it In addition to funding leave both alsoprovided me with a small sum of money for the purchase of books whichI found immensely helpful in ensuring that I had all that I needed to workand write efficiently

My intellectual debts are many Peter Milne read ancestors of chaptersII and V and generously helped me with some of the more technical partsof chapter II Jonathan Barnes read and commented on an ancestor ofchapter III Graham Macdonald and Mark Sainsbury commented on andmade many helpful suggestions for the improvement of early versions ofchapters I and V Peter Lipton provided me with many fruitful discussionsof explanation generally and also commented in detail on chapters I IVV and VI Gary Clarke and Paul Noordhof read over the whole manuscriptin an almost final form both made many useful suggestions throughoutthe manuscript and saved me from numerous errors It would perhaps

x


not be inappropriate in a paragraph on intellectual debts to mention mydeep respect for the literature I discuss (even when I argue with it) andthe extent to which I have learned and profited from it This is obvious inthe case of the historical figures but obvious or not it is similarly thecase with the contemporary literature on explanation which I cite (andsome which I do not have space or time to cite) Whatever I have beenable to discern has only been by standing on their shoulders I have learneda great deal from everything I have read but perhaps the greatest singleinfluence on my thinking has been the work of Peter Achinstein

It is so self-evident that only the writer himself can be responsible forany remaining mistakes and errors that writers often attempt to discoverincreasingly novel or amusing ways in which to say this I shall not try Iknow that the philosophical influence of all these people made the bookmuch better than it would otherwise have been and it cannot be the faultof any of them that they were unable to detect all of the errors I made orunable to ensure that I was capable of making good every error they pointedout to me

In each of my previously published books and articles I have thankedMark Sainsbury for philosophical conversation whichmdashall too oftenmdashhas been one-sided with him as teacher and me as pupil I like mostphilosophers cannot work without constant philosophical discussion andI have him principally to thank for bringing it about that I live in aphilosophically acceptable environment

The strategy of the book is almost but not quite straightforward Inthe historical portion of the book chapters II III and IV I discuss thetheories of explanation of Plato Aristotle John Stuart Mill and CarlHempel Although there is little explicit philosophical work on explanationbetween Aristotle and Millmdasha gap of over two thousand yearsmdashthere ismuch implicit in the writings of Bacon Berkeley and many otherphilosophers that is relevant to explanation but which considerations ofspace have forced me to neglect I discuss and state my view on someissues as I move through these historical chapters but in the main I reservechapters V VI and VII for the elaboration of my own views on explanation

I have not yet mentioned the purpose of chapter I The placement ofthis chapter has given me some pause As I began my discussions of thehistorical figures I found myself in constant need of a technical vocabularywith which to make the issues they treat clear and precise I thereforedecided to devote an opening non-historical chapter to questions ofterminology and to classification of kinds of theories of explanation Thedanger in this strategy is that the reader will not really see the point of

xi


chapter I until much later in the book I might suggest for readers whobegin to tire of chapter I that they proceed to chapter II and return tochapter I only when they find a need for a discussion of the issues it dealswith I decided not to relocate chapter I to a later position in the book butto leave it in place allowing readers to decide when the reading of thechapter would be appropriate

David-Hillel RubenLondon 1990

1

CHAPTER I

Getting our Bearings

The series in which this book is appearing is called lsquoThe Problems ofPhilosophy Their Past and Presentrsquo this volume since it is about theconcept of explanation discusses some of the philosophical problemsabout explanation as they arise in the writings of past philosophers

It is necessary to introduce certain distinctions and settle a fewsubstantive matters before beginning the discussion of explanation inthe succeeding chapters One possible consequence of this approach isthat readers will not always see the motive for the distinction or decisionI can only ask them to be patient for the discussion in the followingchapters returns to these issues time and time again I engage in a separateintroductory treatment of these common and recurring themes rather thanweave them into the body of the ensuing text But perhaps a lsquomaprsquo ofwhat this chapter contains will help

First it is essential to identify more precisely the concept I shall bediscussing Which concept does the term lsquoexplanationrsquo designate Theliterature is somewhat remiss in this respect Usually the authorpresupposes that the audience will have no difficulty in identifying whichconcept it is about which the author wishes to raise certain problemsThis may be an acceptable presupposition in discussions of concepts likecausation and knowledge It does not seem to me to be an acceptablepresupposition in the case of explanation (or for that matter in the caseof the concept of a person) Hence it is not a presupposition that I shallmake One of my main motives in the sections entitled lsquoSomeexplanationsrsquo lsquoProcess and productrsquo lsquoRestricting the scope of theanalysisrsquo lsquoScientific and ordinary explanationrsquo lsquoPartial and fullexplanationrsquo and lsquoBad explanations and no explanationsrsquo is to specify

2


as precisely as I can which concept it is that I shall be discussing bydistinguishing it from others with which it might easily be confused

I also use this chapter to introduce some terminology and draw variousdistinctions that I need for my later discussion One needs a perspicuousterminology in which to raise the central questions properly Thephilosophical implications (for surely there are such) of choice ofterminology are not always apparent to the writer it is therefore especiallyincumbent on the writer to be as clear about this as possible so that othersmay be able to see those implicit and unnoticed ramifications which mayescape notice Introduction of terminology and drawing of pertinentdistinctions occur in the sections mentioned above but also in the sectionsentitled lsquoSome terminologyrsquo lsquoTheories of explanationrsquo and lsquoDispensingwith contrastivesrsquo In the last section lsquoDispensing with contrastivesrsquo Idiscuss a certain view about what it is that one explains in an explanationI discuss explanation in a lsquotraditionalrsquo terminology which the contrastiveview seeks to overturn hence my motive for taking on the contrastiveview in this introductory chapter

The section on theories of explanation is the longest in the chapter Itoffers a typology by which to identify and describe specific theories ofexplanation In order to help the reader see what is going on in that sectionI introduce its own lsquomaprsquo at the beginning of the section But I wouldstress that the motive for drawing the distinctions in the way I do can onlyemerge in the subsequent chapters in which the distinctions are appliedto specific theories

Many writers on explanation fail to make the lsquoground rulesrsquo of thediscussion of explanation at all clear One is presented in the literatureon explanation with many extremely plausible but competing accountsof explanation In virtue of what features is one account better thananother What acceptance tests should an account of explanation beprepared to meet I address this question in the section entitled lsquoThemethodology of explaining explanationrsquo

Throughout the book I make use of a contrast between epistemologyand metaphysics and the various concepts whose analyses belong to oneor the other of these two branches of philosophy For example a themethat recurs throughout the book is that explanation is an epistemologicalconcept but one which requires a metaphysical lsquobackingrsquo

I am content for this contrast to be understood in a rough and readyway Metaphysics is the study of what there is and what it is likequite apart from questions about our knowledge of these mattersTypical metaphysical questions include are there universals what is

3


an event does every event have a cause is the concept of causationa deterministic concept Epistemology is the study of knowledgebelief reasons and evidence Typical epistemological questionsinclude must all beliefs be justified by other beliefs is all knowledgecertain which if any non-deductive arguments with true premissesprovide reasons for belief in their conclusions I am quite prepared toadmit that there are some concepts which do not fit easily into onecategory rather than the other (perhaps the concepts of truth and offact are examples) but this does not I think detract from the usefulnessof the distinction

I do occasionally refer to the views of Carl Hempel throughout thischapter I discuss Hempel fully in chapter IV However since his writingson explanation have proved to be so central to contemporary discussionsreference to him here is intended to be merely a useful illustration ofwhatever specific question is at hand

Some explanations

Giving explanations is a common activity engaged in by layman andscientific specialist alike Most books about explanations begin bygiving examples of scientific explanation The following arerepresentative cases of the sort of explanations that scientists offer

(a) Two kilograms of copper at 60 degrees C are placed in three kilogramsof water at 20 degrees C After a while water and copper reach thesame equilibrium temperature 225 degrees C and then cool downtogether to the temperature of the surrounding atmosphere Why isthe equilibrium temperature 225 degrees C Since the specific heatsof water and copper are 1 and 01 respectively and since theconservation of energy requires that the total amount of heat be neitherincreased or diminished the heat loss of copper namely 01x 2x(60-T) must be the same as the heat gain of water namely 1x3x(T-20)where T is the final equilibrium temperature And this yields 225degrees C as the value of T

(b) Two nerve impulses I1 and I2 in close physical proximity in a neuronarrive within 03 milliseconds of each other at the synapse of thatneuron Neither has a local potential quite strong enough to fire acertain adjacent dendrite Nevertheless the dendrite in question firedWhy Because the local potentials of I1 and I2 have summated to adegree high enough to evoke a spike potential in the adjacent dendritea phenomenon that will occur in the described circumstances provided

4


that the arrival time of the distinct nerve impulses does not exceed05 milliseconds

(c) It is observed that certain human beings suffering from extreme fatigueand lengthy food deprivation show little or no desire to eat whenpresented with food The explanation for this is that extreme fatigueinhibits the rhythmic contractions in the duodenum that initiate bloodchemistry changes which in turn trip off the central mechanismsleading to eating behaviour1

The three examples of scientific explanation cited above are pickedalmost at random from many equally good ones with which the readerwould be presented in any adequate book on the concept of scientificexplanation

The third example is an example of the explanation of a generalization(well almost a generalization the point is that it does not concern a specificor particular case) lsquocertain human beings suffering from extreme fatigueand lengthy food deprivation show little or no desire to eat when presentedwith foodrsquo The first and second examples are examples of the explanationof particular cases two specific nerve impulses which fire a dendrite anda specific sample of copper weighing two kilograms placed in a containerwith three kilograms of water Carl Hempel cites a particular caseexplanation in the opening pages of his Aspects of Scientific Explanation

John Dewey describes a phenomenon he observed one day whilewashing dishes Having removed some glass tumblers from the hotsuds and placed them upside down on a plate he noticed that soapbubbles emerged from under the tumblersrsquo rims grew for a while cameto a standstill and finally receded into the tumblers Why did thishappen Dewey outlines an explanationhellip2

I have relatively little to say in this book about the explanation of lawsand generalizations I concentrate on what I call lsquosingular explanationrsquoSome writers for instance Michael Friedman have claimed thatexplanation in science is almost always explanation of laws

hellipwhat is explained is a general regularity or pattern ofbehaviourmdasha law if you likehellip Although most of thephilosophical literature deals with the explanation of particularevents the type of explanation illustrated by the account aboveseems much more typical of the physical sciences Explanations

5


of particular events are comparatively raremdashfound only perhapsin geology and astronomy3

I think Friedmanrsquos claim is exaggerated Two of the cases which Icited above which are taken from scientific journals are examplesof the explanation of particular events (and neither is from geologyof astronomy) It is true that science oftenmdashperhaps alwaysmdashhasexplanatory interest in particular cases only in so far as they areexamples of a general sort It would not really have mattered if theabove explanations had been of two similar impulses firing a similardendrite or of a similar sample of copper placed in a similar amountof water As Raimo Tuomela says lsquoSingular facts events etc arenot per se of any interest to at least pure science All interest in themis ultimately interest in their being instantiations of some universalrather than anotherhellip for indeed there are no bare particularsrsquo4

This may have something to do with the nature of explanation itselfWhenever a particular case is explained perhaps the same explanationcould be given for any relevantly similar example and so the explanatoryinterest is never in the particular case as such but only in it in so far as itis a particular case of a general sort But this if true is not the same thingas having little or no interest in particular cases In any event if my neglectof the explanation of laws is a weakness of the book at least I can claimthat what I have to say is consistent with the truth about the explanationof laws or generalizations whatever it may be

A theory of explanation does not only address itself to cases of explanationin science It must address itself to other cases as well in which non-specialistsexplain things to one another I am not thinking of explanations of humanaction about which I will have very little to say in this book Rather I have inmind the perfectly acceptable ordinary explanations we are able to give oneanother of natural occurrences the onset of warm weather explains the meltingof the snow overexposure to the sun explains my painful burn my match litbecause I struck it The person who explains the melting of the snow by theonset of warm weather may not be able to explain how or why higher airtemperature causes the snow to melt for this latter they may need amicrotheory which only scientific specialists possess But inability to explainhow or why an air temperature increase leads to the melting of the snow doesnot imply inability to explain the snowrsquos melting on the basis of an increasein the air temperature Nearly everyone whether or not they have a degree ina natural science knows that the snow melts because spring has come

6


The analysis of explanation then belongs to general epistemology inthe same way as the analysis of knowledge does and not just to thephilosophy of science narrowly conceived Scientific explanation likescientific knowledge has a special importance and pride of place in ageneral theory of knowledge But just as there is more to knowledge thanscientific knowledge so too there is more to explanation than scientificexplanation The knowledge that I now have that I am sitting at my deskand writing is not scientific knowledge The explanation that I can give ofthe snowrsquos melting in terms of the warmer weather is not scientificexplanation Ordinary explanations like ordinary knowledge are notimpervious to error and it may sometimes happen that science overturnswhat we wrongly took to be an example of ordinary knowledge or ofacceptable ordinary explanation But when not so overturned suchordinary explanation or knowledge is not per se scientific explanationor knowledge I do not intend these introductory remarks to beg anyquestions about the nature of the distinction between ordinary and scientificexplanation nor to suggest that there is some hard and fast contrast betweenthem I deal with these issues in the course of the chapter Rather theseremarks are intended only to serve as a reminder about the scope of ourtopic Far too many discussions of explanation assume that what can besaid about scientific explanation exhausts what of interest there is thatcan be said about explanation tout court and this is in my view simplynot so

For the present I shall move rather cavalierly between lsquoexplanationrsquoand lsquoscientific explanationrsquo I ask the readerrsquos temporary indulgence Ideal with this (alleged) distinction later in the chapter

Process and product

lsquoExplanationrsquo itself is susceptible to a well-known process-productambiguity as are many other words ending in lsquo-ionrsquo lsquoIn the process-product shift a word often one ending in ldquo-ionrdquo or ldquo-tionrdquo maysignify an activity or its resultrsquo5 A simple example is this lsquoI saw thedestruction at Rotterdamrsquo The sentence might mean either that I sawthe act of Rotterdam being destroyed or that I saw the results of suchan act

lsquoExplanationrsquo is ambiguous in the same way as lsquodestructionrsquo AsBromberger points out in one sense

7


an explanation may be something about which it makes sense toask How long did it take Was it interrupted at any point Whogave it When Where What were the exact words used Forwhose benefit was it given6

On the other hand an explanation lsquomay be something about whichnone of [the previous] questions make sense but about which it makessense to ask Does anyone know it Who thought of it first Is it verycomplicatedrsquo

The linguistic evidence points to two different senses of lsquoexplanationrsquoThe first suggested by Brombergerrsquos evidence is the process or act sensethe second the product sense Other examples of words which have thisambiguity range from philosophically uninteresting ones like lsquosimulationrsquoand lsquodestructionrsquo (Clark and Welshrsquos example) to ones which raisephilosophical issues similar to those raised by lsquoexplanationrsquo lsquopredictionrsquolsquodeductionrsquo lsquoderivationrsquo lsquopropositionrsquo lsquoargumentrsquo lsquostatementrsquo andlsquoanalysisrsquo (although the last three do not end in lsquo-ionrsquo)

So in speaking of an explanation one might be referring to an act ofexplaining or to the product of such an act How are these two sensesrelated There seem to be just four possibilities

(1) The idea of an explanatory act can only be analysed by using theidea of an explanatory product but not vice versa

(2) The idea of an explanatory product can only be analysed by usingthe idea of an explanatory act but not vice versa

(3) The ideas of explanatory act and explanatory product mutually dependon one another

(4) The ideas of explanatory act and explanatory product are independentof one another

Most of the literature on explanation and certainly the four writerson explanation whom I shall be discussing Plato Aristotle Milland Hempel were interested only in the idea of an explanatoryproduct They believed (and I agree with them) that an explanatoryproduct can be characterized solely in terms of the kind ofinformation it conveys no reference to the act of explaining beingrequired Hence each would have rejected (2) and (3) Their questionwas this what information has to be conveyed in order to haveexplained something

One recent writer Peter Achinstein has advanced (2)7 If (2) weretrue then the idea of an explanatory act would have a far more central

8


position in the analysis of explanation than it has previously been givenAccording to Achinstein an explanatory product is neither just anargument (the Hempelian view) of a certain sort nor just a proposition ofa specific kind nor any other entity which can be characterized solely interms of its syntactic form andor the type of information that it conveysRather according to him an explanatory product is an ordered pair inpart consisting of a proposition but also including an explaining act type(eg the type explaining that such-and-such) For example onAchinsteinrsquos view the explanation of why Nero fiddled might be theordered pair lsquoNero fiddled because he was happy seeing Rome burnrsquothe act type explaining why Nero fiddled

Why does Achinstein think that an explanatory product cannot becharacterized solely in terms of its information content His argumentrests on the uncontroversial fact that the same information content mightbe conveyed by both an act of explaining and an act of another type8 egan act of criticizing For instance in saying that Nero fiddled because hewas happy seeing Rome burn Achinstein claims that I could be eithercriticizing Nero or explaining his action (Achinstein 198388ndash9)Achinstein reasons that since the same information content can beconveyed in two different kinds of acts and since no product of anexplaining act could be identical with for example the product of acriticizing act the explanation product (and the criticism product) mustbe more than just the information conveyed

But why canrsquot the product of an act of criticizing and an act of explainingbe identical Achinstein relies on the following sorts of principles to showthat they cannot be

(5) The product of Srsquos act is an explanation only if S explained(6) The product of Srsquos act is a criticism only if S criticized

These principles will lead to the conclusion that Achinstein wantsIf I am explaining Nerorsquos actions but not criticizing them and youare criticizing them but not explaining them then the explanationproduct of my act cannot be identical with any criticism productand the criticism product of your act cannot be identical with anyexplanation product If explanation and criticism products are to bedistinguished in this case the products ought to be distinguishedeven in the case in which one person is engaging in two or moreacts9 at one and the same time Each of the acts will have its ownlsquointernalrsquo product

9


But what reason is there to think that (5) (6) and other analogousprinciples which claim that a necessary condition for something to be apersonrsquos product10 of a certain kind is that he has actually produced it inan act of that kind are true These principles simply presume what theyare used to prove I can see no good reason to deny that objectivelyspeaking quite apart from whatever intention you (or anyone else) mayhave had in acting the information you impart in criticizing (explaining)Nero may also be an explanation (criticism) of what he did in the productsense One can in criticizing Nero convey information which is also anexplanation (in the product sense) of why he fiddled whether the criticizeror indeed anyone else has ever engaged in an act of explaining what hedid There can be explanations (in the product sense) even if no one hasever explained anything (5) and (6) are false

Explanatory products can be fully characterized in terms of theirinformation content independently of explanatory acts so (2) and (3) arefalse (I do not wish to pronounce on the choice between (1) and (4)) Ofcourse we may tend to call such information lsquoan explanationrsquo (in theproduct sense) as opposed to a criticism or an argument only if it figuresas the product of an explaining act But that gives us no more reason todeny that an explanation product may be the same as a criticism productthan there is to deny that the Morning Star=the Evening Star on the groundsthat we tend to call the heavenly body the latter only when it appears inthe evening and the former only when it appears in the morning

The methodology of explaining explanation

The title of this book is Explaining Explanation The suspicious mightthink that there is something self-defeating in such a title How onemight ask if one were genuinely in need of enlightenment about theconcept of explanation could one undertake to explain whatexplanation is Would it not be rather like trying to pull oneself upby onersquos own bootstraps

Of course there is no real difficulty here In offering a philosophicalanalysis of any concept one must attribute to oneself an (at least partial)implicit understanding of that concept which the analysis is attemptingto make explicit Some sophistication or other of this basic idea of what itis to offer an analysis is necessary if one is to escape the paradox ofanalysis The alleged paradox asserts if one knew what was involved in

10


the concept one would not need the analysis if one did not know whatwas involved in the concept no analysis could be forthcoming The escapeis through some implicitexplicit distinction One can know implicitlybut need the analysis to make the knowledge explicit

Moreover there is a second reason why explaining explanation offersno difficulty What we are explaining is the concept of explanation asemployed not only in science but also in ordinary life But the explainingthat we are undertaking is specifically philosophical explication or analysis(I use these terms interchangeably) of a concept Carl Hempel for examplerepeats in several passages that what he is doing is offering an explicationand that the purpose of an explication or analysis is to lay bare the lsquologicalstructure of the conceptrsquo Hempel often speaks of the analysis of a conceptas lsquoa modelrsquo as he does when he says that there are three lsquomodelsrsquo ofexplanation11 On his view there are really three distinct concepts ofexplanation and the three lsquomodelsrsquo make clear in what ways the threediffer

The literature abounds with competing and incompatible explicationsor analyses of explanation Optimally we should like to be able to chooserationally one from amongst them We need to know then what constraintsthere are on such a choice Should the best analysis lsquofitrsquo the way in whichwe ordinarily use the term lsquoexplanationrsquo Should it rather meet somemore technical requirements of science or philosophy I wish below todraw a contrast between two different ways of answering these questions

I intend the following general remarks on methodology to be as anodyneand uncontroversial as possible for they are not intended as an excursusinto the philosophy of language or philosophical logic I do not think oneneeds to take them as a serious contribution to the understanding of thenature of concepts I will also assume in the discussion the view of analysisor explication adopted explicitly by Hempel when he said that he isengaged in laying bare the logical structure of a concept Although Hempeldoes not say so it would seem to follow from this view that the truths soexposed about a concept have the status of analytic or necessary truths Ido not here distinguish between analyticity and necessity for nothing ofimportance for my discussion hangs on that distinction

Perhaps a brief comparison with the philosophical literature on theanalysis of knowledge will help us to understand the idea of the twodifferent ways of judging competing analyses I gesture to this otherliterature only as a way of drawing the contrast in philosophical methodthat I will then apply to the case of the analysis of explanation Thesedifferent ways of proceeding philosophically whether in discussing

11


knowledge or explanation arise out of different traditions of what it is todo philosophy

First some discussions of knowledge proceed in this way Variouscomplicated situations are described for instance a situation in which aperson has justified true belief but there is no causal connection betweenthe fact the belief is about and the belief itself We are then asked whetherwe would apply lsquoknowsrsquo in such a situation lsquoSo all three conditions forknowledgehellipare fulfilled but we still do not want to say that S knowshelliprsquolsquoSurely we do not want to say that his friendrsquos wild guess endows S withknowledgersquo12

The idea is that our analysis of knowledge should capture all and onlyor (in a weaker and more plausible version) most of those situations inwhich we would prephilosophically be prepared to use the term lsquoknowsrsquoI am of course thinking of the vast literature inspired by Gettierrsquos famousarticle13 For better or worse I call this method lsquothe language usersrsquoapproachrsquo Notice that this language usersrsquo approach might not be weddedto everyonersquos use of the word at all times (it is important to see that thisapproach need not be wedded to the idea of ordinary language) It is opento an exponent of this view to designate some subset of users of the wordas having a special status For instance the philosopher of knowledgemight only be interested in how scientists employ the concept ofknowledge and perhaps only while they are engaged in some specificscientific activity I still think of this as the same view but with the classof users cut down in size and scope

Michael Friedman adopts this language usersrsquo approach in his accountof explanation

hellipmost if not all scientific the orie s tha t we all co ns explanatoryshould come out as such according to our theoryhellip Although it isunreasonable to demand that a philosophical account of explanationshould show that every theory that has ever been thought to beexplanatory really is explanatory it must at least square with most ofthe important central casesrsquo14

Friedman does not say why his requirement is plausible Isnrsquot itlogically possible that all or most of the central and important casesof theories we thought were explanatory fail really to be so PerhapsFriedman has in mind here some version of the paradigm caseargument if so the prospects for his view seem dim15

12


It is more difficult to give a succinct general characterization of thealternative method which I wish to describe I call it lsquothe technicalapproachrsquo In one way or another it dispenses with such reliance on theway in which terms are actually used or employed As far as the analysisof knowledge is concerned a good example of this approach is KarlPopperrsquos lsquoEpistemology Without a Knowing Subjectrsquo lsquohellipscientificknowledge simply is not knowledge in the sense of the ordinary usage ofthe words ldquoI knowrdquorsquo

hellipordinary languagehelliphas no separate terms fo corresponding twosenses of lsquoknowrsquohellipMy quoting The Oxford English Dictionary shouldnot be interpreted as either a concession to language analysis or as anattempt to appease its adherents It is not quoted in an attempt to provethat lsquoordinary usagersquo covers lsquoknowledgersquo in the objective sense of mythird worldrsquo16

Popperrsquos characterization of knowledge must meet some constraintsand his article goes on to specify just what they are But whateverthey are they do not include lsquofitrsquo with the way in which we (or evenjust scientists) employ or use the term lsquoknowsrsquo

My names for these two positions the language usersrsquo approach andthe technical approach are not especially happy ones but they do at leastsuggest the sort of position intended Different philosophical orientationshave tended to favour one or the other of these positions but of coursethese are lsquoideal typesrsquo and the actual practice of many philosophers ismore complicated combining elements of both of these approaches andperhaps others besides

Even on the language usersrsquo approach one might regard usage as vagueambiguous imprecise even inconsistent or incoherent the philosophermay say that there is no single concept that is expressed by all of theordinary uses of some term He then may single out a subset of those usesas a way to disambiguate and to focus on one concept at a time Or theconcept as used may be vague (it is not strictly true that one can speak ofthe concept in such a circumstance) there may be general agreementabout the paradigm cases but dispute about cases in the conceptrsquospenumbra The language usersrsquo approach should permit us to depart fromordinary usage at least to the extent of eliminating vagueness in a conceptrsquosapplication and disambiguating Letrsquos call this lsquotidying up a discoursersquoBut in all versions of this approach actual language-use is where one atleast begins onersquos analysis

13


The other approach I called lsquothe technical approachrsquo A philosopherengaged on some project might eschew interest in the concepts used bythe speakers of a language The philosopher might see as part of his taskthe introduction of some quite novel concept whose criteria are given bystipulative specification Examples of this include some of the technicalconcepts of philosophy sense data the distinction between essence andaccident the ideas of a metalanguage and material implication The greatphilosophical systems eg the Platonic Kantian and Hegelianphilosophies provide examples of this technical concept introductionForms the distinction between reason and understanding the synthesisof the understanding noumena and phenomena transcendental argumentsthe Absolute in and for itself The philosopher might think that such aconcept plays an important role in coming to understand something thatwe simply failed to understand before I call this lsquosimple introductionrsquo

The above examples of the technical approach are cases ofstraightforward concept introduction But there are other cases in whichthe introduced concept is intended to replace or improve upon one alreadyin use by the common man Humersquos lsquoreformedrsquo concept of the selfBerkeleyrsquos idea of a physical object which excludes the commitment tounperceived existence Hobbesrsquos redefinition of desire and aversion interms of internal motions and the idea of truth-in-a-language are examplesof concept replacement Many scientific reductions involve conceptreplacement in the reduced science arguably the pre-reduction conceptof water is not the same as the concept of water after its identificationwith H20 The latter would then be a replacement for the former

Suppose a philosopher practising this technical approach decides thatthere are good reasons for the introduction of a new concept of X toreplace the old one The new replacing and old replaced concepts willsometimes have very similar extensions and their analyses (or lsquomodelsrsquo)may have many features in common But this is hardly essential Replacingconcepts might differ dramatically in intension from the concepts thatthey replace17 Moreover the new and old concepts of X may differ inextension We might even come to believe that nothing correctly calledlsquoXrsquo before when the old concept was in use can be correctly so-callednow that it is the new concept that is in service and vice versa

I have stressed the intensional and extensional discontinuities theremight be between replacing and replaced concepts of something Butsurely there are limits here There must be some difference between (a)replacing the old concept of X with the new concept of X (b) eliminatingthe concept of X and simply introducing the concept of Y as two separate

14


exercises in improving our discourse How could we account for thisdifference if not by introducing some sort of continuity between the oldand new concepts

Indeed it is true that there must be some sort of continuity There isby and large a point in having the concepts we do For example at leastpart of the point and purpose of explanation is that we should come tounderstand why things happen18 That is the function that explanation hasfor us If a replacing concept of explanation is a replacement for thestandard or ordinary concept of explanation it surely must serve at leastthis function The requisite continuity between the old and new conceptof X might be provided by continuity of function

Some philosophers may believe that these lsquofunctionalrsquo facts about aconcept have no place in its logical analysis They will say in the caseof explanation that although it is true that explanations do or shouldlead us to understand that this is so is not a logically or conceptuallynecessary truth about explanation Hempel for example says that lsquosuchexpressions as ldquorealm of understandingrdquo and ldquocomprehensiblerdquo do notbelong to the vocabulary of logic for they refer to psychological orpragmatic aspects of explanationrsquo19 These facts about what explanationsdo for us have on his account no place within the analysis of explanationitself For such philosophers there could be a complete intensional andextensional discontinuity between the old and new concepts ofexplanation with only the sameness of contingent functional factslinking the two concepts as two concepts of explanation

Other philosophers will find room for these functional facts within theanalysis of the concept20 The analysis of explanation will include somemention of understanding For these philosophers there will after all haveto be some at least minimal intensional continuity between replaced andreplacing concepts of explanation

I have great sympathy for the technical approach rather than thelanguage usersrsquo approach in any of its possible refinements However ifthe technical approach is adopted one needs to consider arguments whichattempt to justify the new replacing concept one has introduced Manynew concepts might be introduced which could be said to have the samepoint as the old replaced one How can we justify one candidate over theothers as the replacing concept if language use does not constrain thatchoice How can we show that the replacing concept we select is not justarbitrary ad hoc

Suppose concepts a bhellipn are all put forward by different philosophersas competing new replacement concepts of explanation Each might

15


plausibly be thought of as a concept of explanation For each itsphilosophical champion can produce a set of necessary or analytic truthsThat by itself is wholly uninteresting Which of a b chellipn is the bestreplacement for the old concept of explanation It is only when we cananswer that question that we will know which set of analytic truths hasany real claim to be of interest to us and what it is that we are trying to dowhen we offer an analysis of explanation This is an issue which I willwant to raise when I look at Aristotle Mill and Hempel and which willprovide a thread of continuity that runs throughout the book

Restricting the scope of the analysis

The Hempelian models are not intended as models of all explanationsHempel contrasts the cases of explanation covered by his models ofscientific explanation with others in which we do not explain whysuch-and-such or that such-and-such lsquoexplaining the rules of acontest explaining the meaning of a cuneiform inscription or of acomplex legal clause or of a passage in a symbolist poem explaininghow to bake a Sacher torte or how to repair a radiorsquo (Hempel1965412ndash13) In the cases of explaining the meaning of somethinglsquothe explanandum will be specified by means of a nounphrasehellipwhereas explanations of the kind we have beenconsideringhellipare characterized by means of a sentencersquo (Hempel1965414) Hempel would consider none of these above mentionedsorts of explanation as scientific in his sense and none constitutes areasonable objection to his account of explanation

Similarly to put forward the covering-law models of scientificexplanation is not to deny that there are other contexts in whichwe speak of explanation nor is it to assert that the correspondinguses of the word lsquoexplainrsquo conform to one or another of ourmodels Obviously those models are not intended to reflect thevarious senses of lsquoexplainrsquo that are involvedhellip Hence to deploreas one critic does the lsquohopelessnessrsquo of the deductive-nomological model on the ground that it does not fit the case ofexplaining or understanding the rules of Hanoverian successionis simply to miss the intent of our model

(Hempel 1965412ndash13)

16


Hempel indicates two ways by which to delimit the explanations forwhich he seeks to offer an analysis The first is grammatical lsquohellipexplanations of the kind we have been considering are concernedwith hellip[whatever] is properly characterized by means of a sentencersquo(Hempel 1965414) Elsewhere he speaks rather circularly of theexplanations in which he is interested as being answers tolsquoexplanation-seeking why-questionsrsquo (Hempel 1965412) Fully andcompletely explaining how to ride a bike is not a case of explanationto which Hempel would consider his models of scientific explanationappropriate it fails both the grammatical and the lsquowhy-questionrsquo testsIn chapter III I return to the question of the adequacy of these twoways of characterizing the subset of explanations to which Hempelrestricts his analysis

Since Hempel in the above quotation speaks of lsquothe various senses ofldquoexplainrdquorsquo he seems to commit himself to the thesis that lsquoexplainrsquo inlsquoexplain that prsquo lsquoexplain howrsquo and lsquoexplain the meaninghelliprsquo is ambiguousThat thesis seems to me dubious but we do not need to decide the matterone way or the other in order to delimit the instances of lsquoexplainrsquo inwhich Hempel is interested

Scientific and ordinary explanation

As my opening remarks suggested there are or are thought to besuch things as scientific explanations The contrast is usually withordinary explanations What does this contrast come to Is lsquoscientificexplanationrsquo anything more than a pleonasm for lsquoexplanationrsquo

There are at least two possible senses of lsquoscientific explanationrsquo In thefirst sense it refers to explanations which are actually given in scienceAs we shall see this is not the sense of the expression in which Hempel isprimarily interested In the second sense the meaning of lsquoscientificexplanationrsquo is commendatory or honorific in some way In any eventin this second sense it is an open question whether any of the explanationsactually given in science are scientific explanations at all

There is without doubt a distinction between ordinary explanationsand scientific explanations in the first sense since it is simply a fact thatsome explanations are given in the course of lifersquos ordinary affairs andothers are given by scientists when they do science But Hempel useslsquoscientific explanationrsquo in the second sense Consequently the question I

17


address in this section is whether there is a distinction between ordinaryexplanations (and also scientific explanations in the first sense) on theone hand and scientific explanations in this second sense on the other

In my view the only distinction that can usefully be drawn is thatbetween full and partial explanations and the distinction between scientific(in the second sense) and ordinary explanations is either that distinctionor no distinction at all As I indicated before although I discuss Hempelrsquosviews on explanation fully in a later chapter I use him here as a way ofsharpening the issue (and in this case actually stating my own position)

To begin with Hempel does not think of scientific explanations asexplanations actually given by scientists lsquothese models are not meant todescribe how working scientists actually formulate their explanatoryaccountsrsquo21 The practising scientist may use lsquoexplanationrsquo in as loose orvague a way as the ordinary man on the street What the scientist calls lsquoanexplanationrsquo and indeed his actual explanatory practices too how heactually goes about explaining things may fall woefully short of whatHempel requires of an explanation Actual explanations in science maysuffer from the same deficiencies as do explanations offered by the non-scientist on the Clapham bus

Perhaps then the term lsquoscientific explanationrsquo is meant to conjure upthe fact that there is a goal or ideal of precision and completenessexplicated by Hempelrsquos models which explanations in science can aspireto and can actually meet if so required lsquoThe construction of our modelstherefore involves some measure of abstraction and of logicalschematizationrsquo (Hempel 1965412) lsquowe have foundhellipthat theexplanatory accounts actually formulated in science and in everydaycontextshellipdiverge more or less markedly from the idealized andschematized covering-law modelsrsquo (Hempel 1965424) Hempel compareshis models of explanation with the lsquoidealrsquo (this is his term)metamathematical standards of proof theory (Hempel 1965414) So themodels are lsquoidealsrsquo in some sense Actual explanations in science mayfall short of the ideal by being elliptic incomplete partial or mere sketchesof an explanation Hempel describes these various forms of incompletenessat some length (Hempel 1965412ndash25)

In what sense does Hempel use the terms lsquoidealrsquo and lsquoidealizedrsquo Themodels are surely not ideals for Hempel in the sense that explainers shouldalways strive to do their best to make their explanations complete thereis no doubt that circumstances can justify explainers in explaining onlyincompletely by omitting information known by their audience In normalcircumstances in which no one doubts the prevalent atmospheric

18


conditions a scientist would be a bore if he attempted to explain the fireby adducing both the short circuit and the presence of oxygen It is nottrue that even scientists always ought to give as full an explanation as ispossible

Rather the models Hempel introduces are ideals for him simply in thesense that they are complete they specify a type of complete or fullexplanation In fact Hempel believes that such complete explanationsare rarely given even in science It is possible and it would not matter tohis argument if it were so that no one not even a scientist actually everoffers such a complete and full explanation which includes exceptionlesslaws needing no further qualification and all relevant initial conditionsMoreover it could even be that every actual explanation ever given wasjustifiably incomplete due to the pragmatic constraints on providingexplanations However and this is surely the important point for himincomplete explanations explain only in virtue of there being suchcomplete explanations whether or not anyone ever gives or should giveone One might draw the necessary distinction in Kantian terms Hempelrsquosrequirements provide a constitutive ideal for full explanation they arenot intended as a regulative ideal

A consequence of this interpretation of what Hempel has in mind isthat if these models provide an ideal or goal for explanations in science(lsquoscientific explanationsrsquo in the first sense) there is no reason why theyshould not equally provide an ideal for explanations in ordinary life tooThe ideal sets a standard for explanation tout court Indeed Hempeldiscusses quite explicitly the application of his model to historical eventsto the actions of agents and to functional systems In science a scientistmight give some explanation that because of the constraints of time orthe interests of his audience fails to live up to Hempelian standardsHowever exactly the same is true in ordinary life We normally are happyto explain why the chicken crossed the road by saying that it wanted toget to the other side but if required we could impose all of the Hempelianrequirements to obtain a full explanation of what the chicken did Onseveral occasions Hempel explicitly couples scientific (in the first sense)and everyday explanations together as both being subject to the samelsquoidealized and schematized hellipmodelsrsquo (Hempel 1965424ndash5) As he toldus above explanations in science and everyday contexts lsquodiverge moreor less markedlyrsquo from the ideals set by his models

So a lsquoscientific explanationrsquo (in the second sense) doesnrsquot seem to beeither an explanation actually offered in science or an ideal appropriateonly for explanations offered in science or by scientists The truth is that

19


as far as Hempel is concerned the Hempelian models of scientificexplanation if they provide an ideal for any explanations provide anideal for all explanations (subject only to the restriction described in thepreceding section on the range of cases for which the analysis is offered)They are models of complete explanation in science and in ordinaryaffairs By lsquoscientific explanationrsquo (in the second sense) Hempel meansonly lsquoa complete or full explanationrsquo and nothing more

I have developed my discussion of lsquoscientific explanationrsquo around remarksof Hempelrsquos But I think that the lesson is general lsquoScientific explanationrsquo isan expression that repeatedly occurs in most discussions of explanation Iflsquoscientific explanationrsquo does not mean lsquoexplanation actually offered in sciencersquothe sense of the expression is far from obvious and needs to be made clearMany philosophers of explanation use it merely in the sense of lsquoan ideallycomplete explanationrsquo Much of the potentially mesmerizing mystique oflsquoscientific explanationrsquo will vanish if this is kept in mind

Partial and full explanation

The key then to unlocking the idea of a scientific explanation (in thesecond sense) is the distinction between complete or full andincomplete or partial explanation22 The distinction between partialand full explanations is a distinction between two different sorts ofexplanatory products presumably the activity of explanation-givingcan at least sometimes justify giving partial rather than full ones

It is not possible to draw the distinction between full and partialexplanations in a neutral way equally agreeable to all theories ofexplanation Different theories disagree about what counts as a fullexplanation Some will hold that explanations as given in the ordinaryway are full explanations in their own right others (like Hempel) willargue that full explanations are only those which meet some ideal rarelyif ever achieved in practice A partial explanation is simply a fullexplanation (whatever that is) with some part of it left out On any theoryof explanation we sometimes do not say all that we should say if wewere explaining in full Sometimes we assume that the audience is inpossession of facts which do not stand in need of repetition At othertimes our ignorance does not allow us to fill some of the explanatorygaps that we admit occur In such cases in which we omit information forpragmatic or epistemic reasons we give partial explanations

20


Partiality is sometimes related to falsity Laws may be omitted entirelyfrom a partial explanation Sometimes they are not omitted but rather aregiven an incomplete formulation which ignores certain exceptions If alaw is an exceptionless generalization an incompletely formulated law isa generalization with exceptions and which is therefore not strictly trueSomething not strictly true is just false On other occasions strictly relevantinitial conditions might be too marginally relevant to the explanandumoutcome to include in the explanans and so the explanation in order topresent itself as if it were complete rather than only partial may make aclosure assumption about the environment in which the outcome occurswhich is not strictly true

Of course whether some particular explanation is partial or not maybe contentious Since theorists will disagree on standards for fullexplanation they are bound to disagree about which explanations arepartial All I assert is that every theory of explanation must draw somedistinction between full and partial explanation and that the idea of apartial explanation is parasitic on the idea of a full one

Recall that in the first sense of the term lsquoscientific explanationrsquo refersto the explanations actually given in science Most or all of theseexplanations are like their ordinary counterparts merely partialexplanations for Hempel It is consistent with my interpretation of Hempelthat the way in which explanations actually given in science are partialmay generally differ from the way in which actual ordinary explanationsare partial For example typically ordinary explanations omit all mentionof laws and this may not be so in at least some areas of science Forexample in the first example of a scientific explanation given at thebeginning of this chapter even if it were to count as partial on somegrounds it does mention the law of the conservation of energy In thesecond example although a law is not explicitly mentioned it proffers allthe materials for the formulation of one in the concluding sentence Ireturn to the question of the place of laws in explanation and the idea ofa full explanation in chapter VI

In what follows unless I otherwise indicate I mean to be speakingof full explanation If I want to speak of partial explanation I explicitlyuse the qualifying adjective I sometimes add lsquofullrsquo as a qualification ifthe qualification is especially important and stands in need of emphasis

21


Bad explanations and no explanations

Is the concept of explanation for which we are seeking an explicationthe same as the concept of a good explanation This question is highlycontentious (eg it involves the distinction between semantics andpragmatics) and is inextricably bound up with other questions aboutexplanation I will have something more to say about this in chapterV Whatever the right answer it is important for a philosopher to beclear about how he would answer it

Consider the following remarks by Hilary Putnam

Explanation is an interest-relative notionhellipexplanationhellipexplanationhas to be partly a pragmatic concept To regard the lsquopragmaticsrsquo ofexplanation as no part of the concept is to abdicate the job of figuringout what makes the explanation good More precisely the issue is notwhether we count the pragmatic features as lsquopart of the meaningrsquomdashthat is a silly kind of issue in the case of such notions as lsquoexplanationrsquomdashbut whether our theory does justice to them or relegates them to merelsquopsychologyrsquo23

Letrsquos call Putnam an lsquoexplanatory pragmatistrsquo I take that to meanthat what counts for him as a full explanation of something (and notjust as a good explanation of that thing) is audience-variant theinterests of audiences differ and therefore what counts as a fullexplanation differs as a function of differences in interest Everytheorist of explanation can admit that the idea of a good explanationis audience-variant Putnam is refusing to draw a sharp distinctionbetween explanation and good explanation and therefore argues thatthe idea of full explanation not just that of good explanation isaudience-variant

From my point of view Putnam unjustifiably conflates the analysis ofexplanation with the pragmatics of giving explanations (or the pragmaticsof information giving for following David Lewis24 I think that therequirements for explaining well are included in the requirements forconveying information well) Nor do I see why Putnam thinks this is asilly kind of issue In this I follow Hempel and others in thinking thatthere is a clear distinction between the analysis of explanation and thepragmatics of explanation-giving It will be my view that we can markout what counts as an explanation by the information content of what is

22


said For example on one specific sort of non-pragmatic view a causaltheory of explanation (this is not my view but I use it for the purpose ofillustrating the point) an explanation of an event e is always in terms ofits cause c Perhaps not just any true statement of the form lsquoc is the causeof ersquo would be an explanation But to try to explain e in terms of someevent that is not its cause would be on this view to produce no explanationat all It would be to cite something simply irrelevant from the point ofview of explanation Such a requirement for explanatory relevance wouldnot be audience-variant

What a causal theorist indeed any non-pragmatist about explanationcan concede is that how we select from the full list of explanatory relevantfeatures in order to obtain the ones required in a particular (partial)explanation we may offer is a pragmatic and audience-variant questionA partial explanation is one that omits certain relevant factors a fullexplanation is one that includes all relevant factors In lsquoc causally explainsersquo one might be citing what is in fact only part of the cause (or the causeonly partially described if one prefers) The cause of the matchrsquos lightingwas its being struck But if I say this I assume that my audience knows orassumes that the match was dry and that oxygen was present and that myaudience has no further interest in having the dryness of the match or thepresence of the oxygen mentioned That is a matter of pragmatics A partialexplanation may be good relative to one set of circumstances but badrelative to another in which interests beliefs or whatever differ

There are additional ways in which an explanation can be bad otherthan by being partial in its selection of relevant factors in the wrong wayA full explanation can be bad too if it conveys more information than isrequired (suppose it sends the listener to sleep) A partial explanation canalso be bad for other reasons The cause could be described in a causallyrelevant but too general or too specific a way In a history textbook theoccurrence of a plague can explain a population decline but theexplanation might be bad if it included a detailed microbiologicaldescription of the disease Putnam himself contrasts the goodness of thesimple explanation of why a 1 inch square peg will not pass through a 1inch round hole in terms of geometry compared with the awfulness ofthe far more complex and detailed explanation in terms of a completeenumeration of all the possible trajectories of the elementary particlesmaking up the peg obtained by applying forces and the fact that nocombination of them takes the peg through the round hole25

But the non-pragmatist will insist that all of these remarks are aboutthe goodness of explanations and relate to ill-advised choices concerning

23


selection from or description of relevant features None of theseconcessions shows that there are no audience-invariant constraints on whatcould count as a relevant feature (for the purposes of explanation) andhence on what could count as an explanation

In this book I take it that the topic is the analysis or explication of theconcept of explanation I have nothing to say directly about pragmaticissues That one can produce contrary to Putnamrsquos remarks an accountof explanation that distinguishes between explanations (whether good orbad) and non-explanations on the basis of information content is bestargued for not in the abstract but by producing just such an account It isthis that I hope to do in chapter VII

Some terminology

The expressions lsquoexplanansrsquo (ie that which does the explaining) andlsquoexplanandumrsquo (ie that which is explained)mdashand their pluralslsquoexplanantiarsquo and lsquoexplanandarsquomdashoccur repeatedly in this book Theyalso occur ambiguously and this is intentional on my part

If explanation is a relation one can refer to its relata whatever they maybe as lsquothe explanansrsquo and lsquothe explanandumrsquo What ontological sort ofentities are these explanantia and explananda We shall discuss this issuefully in chapter V Obvious candidates include phenomena events factsand true propositions (or beliefs or statements) Whichever candidate isselected we can call this lsquonon-sentence explanationrsquo If events can explainevents then chunks or bits of reality (like the matchrsquos striking and the matchrsquoslighting) literally explain and are explained Or perhaps it is the fact thatsome event occurred which explains the fact that some other event occurred

On the other hand another possibility is that it is true statements whichexplain true statements rather than events which explain events(Propositions and statements are not sentences) Even if this is sostatements explain and are explained only in virtue of the way the thingsin the world which they are about really are If it is the statement thatthere is a short circuit that explains the statement that there is a fire theexplanation only works in virtue of the real short circuit bringing aboutthe real fire (and although it would not be true strictly speaking that it isthe short circuit that explains the fire)

On one well known theory we will be examining we explain only ifwe can deduce a sentence describing the explained phenomenon from a

24


sentence that describes the explaining reality and a lawlike generalizationIn this way then one might also think of lsquoexplanantiarsquo and lsquoexplanandarsquoas sentences (which should be sharply distinguished from statements orpropositions) eg lsquothe explanans entails the explanandumrsquo We can callthis lsquosentence explanationrsquo26

But if there is sentence explanation it is conceptually dependent onthe primary idea of non-sentence explanation (whether the right choiceof relata for that relation is events or facts or true statements) This is Ihold uncontroversial27 Even a theory that seeks to analyse explanationin terms of the logical form of and logical relations between varioussentences is analysing the idea of explanation in the primary non-sentencesense The theory may attempt to lsquoreducersquo the idea of non-sentenceexplanation to some facts about sentences but it does not reduce non-sentence explanation to the idea of sentence explanation

In this intentional ambiguity of lsquoexplanansrsquo and lsquoexplanandumrsquo Ifollow Hempel himself (except that he conflates lsquosentencersquo andlsquostatementrsquo)

The conclusion E of the argument is a sentence describing theexplanandum-phenomenon I will call E the explanandumsentence or explanandum statement the word lsquoexplanandumrsquoalone will be used to refer either to the explanandum-phenomenon or the explanandum-sentence the context will showwhich is meant28

Context will also determine whether I am using lsquoexplanansrsquo orlsquoexplanandumrsquo in the sentence or non-sentence sense

So I variously employ these expressions to refer to sentencesstatements (or beliefs) the facts and the actual worldly events thestatements are about The ambiguity is harmless it often lets me say lessclumsily what would otherwise involve cumbersome expression In anyevent even if we wished it would not be possible to sort out fully theambiguity beyond what I have said here in advance of the discussion inchapter V concerning the relata of the explanation relation

Salmon introduces all three obvious non-sentence categoriesstatements events and facts

It is customary nowadays to refer to the event-to-be-explainedas the explanandum event and to the statement that such an event

25


has occurred as the explanandum statement Those factsmdashbothparticular and generalmdashthat are invoked to provide theexplanation are known as the explanans If we want to referspecifically to statements that express such facts we may speakof the explanans statements The explanans and explanandumtaken together constitute the explanation29

What the quotation appears to say is that the explanation relation perse relates facts The events such facts are about are the explanansevent(s) and the explanandum event The statements which expresssuch facts are the explanans statement(s) and the explanandumstatement On Salmonrsquos view we explain facts which are aboutevents by means of making various statements One consequence ofthis view is that there must be a significant distinction between factsand statements In chapter V I return to these questions and especiallyto the theme of facts and the role they might play in a theory ofexplanation

Theories of explanation

Let me introduce what shall prove to be some useful distinctionsbetween different types of theories of full explanation although theextent of that usefulness can only be apparent as those distinctionsare applied in subsequent chapters I stress that these are theories offull explanation I shall try and add some remarks about partialexplanation as I go along The distinctions provide allegedly necessaryconditions for explanation not sufficient conditions Thus thesedistinctions do not themselves yield specific theories of explanationbut rather permit us to catalogue specific theories as being of one oranother of the types30

The distinctions make use of concepts such as event causationdeterminism indeterminism certainty probability deductive and non-deductive argument In a book on explanation it will be unnecessary tooffer analyses of these concepts The purpose of introducing them is onlyto show how they relate to explanation I use them hopefully in waysuncontroversial to the matters at hand

I introduce three sets of distinctions by which to categorize theories ofexplanation (A) (B) and (C) The typology which these three sets of

26


distinctions produce permits us to categorize theories of explanation intwo different ways epistemologically and metaphysically The first twosets of distinctions (A) and (B) are epistemological Hempel for examplesays that we explain something when we see that it lsquowas to be expectedand it is in this sense that the explanation enables us to understand whythe phenomenon occurredrsquo31 An expectation is a belief Must our beliefabout the occurrence of the explained phenomenon be certain or might itonly be likely Under (A) I distinguish between theories of explanationwhich offer different answers to these questions

Theories also differ about the form an explanation may take I discussthese distinctions under (B) Must an explanation be an argument Idistinguish between argument theories of explanation (which answer thepreceding question in the affirmative) and non-argument theories (whichanswer it in the negative) Argument and non-argument theories givesomewhat different answers to the epistemological question of the certaintyor probability of onersquos belief about the explanandum phenomenonArgument theories can use the ideas of deductive and non-deductivearguments as a way of giving substance to the ideas of certainty andepistemic probability non-argument theories do not have this manoeuvreavailable to them

The third set of distinctions which I use to classify theories ofexplanation is metaphysical and I discuss this under (C) The relevantmetaphysical distinctions involve among other things the ideas ofcausation determinism indeterminism and nondeterminism That isdifferent theories of explanation presuppose different things about thenature and extent of causation A theme that runs throughout this book isthe way in which an epistemic concept like explanation requires orpresupposes a lsquometaphysical backingrsquo I try to show how those differingmetaphysical commitments partially motivate different epistemic viewsabout explanation

Probability is a highly ambiguous term and although there are manykinds of probability and various further distinctions one can draw withinthe two broad kinds of probability I distinguish I want simply to separateepistemic or inductive probability from physical or objective or descriptiveprobability32 There are many competing accounts of each (eg frequencyand propensity theories are competing accounts of physical probabilitylogical and Bayesian theories are competing accounts of epistemicprobability) Epistemic probability is concerned in some way with supportor degree of rational belief physical probability is meant to be a matter ofobjective fact about the world Obviously the two concepts of probability

27


are related although distinct (Another term that can have both an epistemicand a metaphysical sense is the concept of what is necessary lsquoNecessaryrsquocan either be construed as lsquocertainrsquo or as lsquoobjectively necessaryrsquo)

(A) I begin now drawing the epistemological distinctions betweendifferent theories of explanation There are certainty high epistemicprobability and low epistemic probability models of explanation (Theseare three rival accounts) On a certainty model of explanation an explananscan explain an explanandum only if the explanandum is certain giventhe information contained in the explanans This is what we might callrelative or conditional rather than absolute certainty something may becertain given something else without being certain or indubitable per seThis is one of the ways in which one might interpret von Wrightrsquos remarkwhat makes an explanation explanatory is that lsquoit tells us why [an event]E had to be (occur) why E was necessary once the basis is there and thelaws are acceptedrsquo33 lsquoNecessaryrsquo here might be construed as an epistemicidea it is certain that E would occur given knowledge of the basis andthe laws

On the other hand one might only require of an explanation that theexplanandum be (epistemically) probable given the explanansinformation lsquohellipwe we might try to salvage what we can by demandingthat an explanation that does not necessitate its explanandum must makeit highly probablersquo34 Salmon in this passage is suggesting that anexplanans need only to make its explanandum epistemically probable(we need not discuss just yet whether highly so or not) but need notmake it certain An epistemic probability model says that there can bemore kinds of full explanations than the certainty model allows The formerallows that there can be full explanations which meet the certainty modeland others beside So they are rival accounts

An epistemic probability model comes in a stronger and a weaker formThe strong model is a high epistemic probability model It requires thatin a full explanation the explanandum is at least highly likely given theexplanans information (or the explanans highly supports theexplanandum) Given the information in the explanans we have goodreason to believe that the explanandum is true but perhaps not conclusivereason It is true that the strong model has a certain vagueness about itbut it is not clear whether vagueness here is a strength or a weaknessWhat is highly likely Any cut-off we select will appear arbitrary andunmotivated But we might argue that this captures accurately thevagueness of explanation itself The higher the probability of theexplanandum given the explanans the more clearly we have an

28


explanation We have no clear intuitions the strong modellist might sayabout at precisely what point we cease having even a poor explanationand have instead no explanation at all Moreover this is entirely consistentwith the non-pragmatic view that there is a distinction between poorexplanations and no explanations

Is lsquoexplanationrsquo genuinely ambiguous according to the epistemicprobability model This depends on the way in which the certainty-conferring and probability-conferring models of explanation are set outThe high epistemic probability model need not hold that there is a radicaldifference between the two kinds of explanation since certainty is thelimiting case of high probability On the other hand Hempel often speaksof these as two different types or kinds of explanation That on its own isno more evidence for ambiguity than the fact that there are vertebrate andinvertebrate kinds of animals is evidence for the ambiguity of lsquoanimalrsquoHowever Hempel says that there are different models for explanationand given his views on models and analysis this ought to mean thatlsquoexplanationrsquo for him is ambiguous and stands for no single conceptThus he says lsquowe have to acknowledge that they [explanationsconforming to the I-S model] constitute explanations of a distinct logicalcharacter reflecting as we might say a different sense of the wordldquobecauserdquorsquo (Hempel 1965393)

The weaker version of an epistemic probability model does not evenrequire that in a full explanation the explanans information provide goodalthough not necessarily conclusive reason to believe that theexplanandum is true On this weaker version the explanans informationmay only give some albeit small reason to believe that the explanandumis true in one sense of lsquoexpectationrsquo the explanandum phenomenon wasnot to be expected at all As Peter Railton says the explanation lsquodoes notexplainhellipwhy the decay could be expected to take place And a goodthing toohellipthere is no could be expected to about the decay to explainmdashit is not only a chance event but a very improbable onersquo35

Wesley Salmon has also argued that the explanandum might have alow probability given the explanans The quote from him four paragraphsback continues lsquohellipeven this demand [for high probability] isexcessivehellipwe must accept explanations in which the explanandum eventends up with a low posterior weightrsquo (Salmon et al 197164)

There are two sorts of arguments for a low epistemic probability modelof explanation The first is simply the presentation of cases of explanationwhich appear to support such a theory The most convincing examplesare indeterminisitic ones since they ground the low epistemic probability

29


of the explanandum statement on the objective low conditional probabilityof the corresponding event Both Salmon and van Fraassen use thisexample

hellipa uranium nucleus may have a probability as low as 10ndash38 ofdecaying by spontaneously ejecting an alpha-particle at aparticular moment When decay does occur we explain it in termsof the lsquotunnel effectrsquo which assigns a low probability to thatevent

(Salmon et al 1971152)

The thought here is that in both cases the decayrsquos occurring and thedecayrsquos not occurring precisely the same information is relevant tothe outcome It seems arbitrary to allow that the information hasexplanatory force in the case of one outcome but to deny that theinformation has any explanatory force in the case of the other

This question should be considered by anyone who is inclined to accepta high epistemic probability model and deny a low epistemic probabilitymodel why should exactly the same information which intuitively seemsequally relevant to both events explain one but not the other Of coursethe convinced high epistemic probability modellist can always replybecause the information makes what is to be explained highly probableor likely in one case but not in the other What is wrong with the reply isthat it seems as arbitrary and unmotivated as the original doctrine So theconclusion would seem to be if explanations meeting the high epistemicprobability model are acceptable then we should sometimes be in aposition to explain an explanandum on the basis of an explanans on whichthe explanandum is only improbable or unlikely

A second argument that Salmon uses for the low epistemic probabilitymodel derives from a famous argument due to Kyburg36 Kyburgrsquosargument concerned the class of reasonably accepted statements lsquoanidealized body of scientific knowledgersquo The question he raises is this isthe class of reasonably accepted statements closed under conjunctionClosure of the set under conjunction would amount to this If S is a bodyof reasonably accepted statements then the conjunction of any finitenumber of members of S belongs to S Suppose p and q are members ofthe set of statements which are reasonably accepted by me Closure underconjunction means that for any p and q if p and q are reasonably acceptedthen (pampq) is reasonably accepted

30


Now a statement need not be certain (have a probability of 1) in orderto gain admittance to S Suppose that we decide to admit to the body ofour reasonably held beliefs only those beliefs which are either certain orhighly probable say with a probability of at least 085 We admit p and q(which we assume throughout are statistically independent) each of whichhas a probability of 09 and therefore qualifies for admission Because ofthe basic multiplicative rule of the probability calculus the belief (pampq)will have a probability of 081 below the bottom limit of acceptabilityFor whatever lower limit of acceptability that we set some conjunctionof what is accepted will itself be unacceptable This seems in contradictionto the intuitively plausible closure principle but Kyburg himself in hisarticle counsels abandoning the principle despite its original appeal lsquoItis difficult to give an argument against the conjunction principle partlybecause it is so obvious to me that it is false and partly because it is soobvious to certain other people that it is truersquo (Kyburg 197077)

One implication of the conjunction principle that Kyburg thinks isfalse is that one has a right to believe the conjunction of all the statementsone has a right to believe Even if one has good reasons for believingeach and every statement that one believes one may still have a generalargument for believing that some (but of course one would not knowwhich) of the things one believes are false If such a general argumentwere sound then one would not have the right to believe the conjunctionof all the statements that one has a right to believe

Salmon has used this same basic argument in several of his writings37

but applied it to explanation rather than reasonable acceptance to link thefates of the high and low probability models Suppose S is now taken to bethe body of explained statements (a statement is explained iff there is someexplanans that explains it) Letrsquos pretend we are high epistemic probabilitymodellists and say that a statement gains admittance to this set S only ifthere is some information on which the statement has a probability of atleast 085 Again suppose that p and q each have a probability of 09 If weaccept the following conjunctive closure principle for explanation

If S is the body of statements which have an explanation then theconjunction of any finite number of members of S belongs to S

then we can argue that (pampq) has an explanation which confers on ita probability below the required level Whatever lower probabilitylimit we set for explanation an application of this argument (Salmon197180ndash1) will force us to admit as an explained statement some

31


statement whose probability given the explanans is lower than theintended lowest limit Hence a low epistemic probability model ofexplanation must be true Unlike Kyburg Salmon holds fast to theconjunction principle and accepts the consequence that there is nolowest limit to the epistemic probability an explanans must confer onthe explanandum in order for the former to explain the latter

Salmon as far as I know has never given any argument for holding onto the conjunctive closure principle for explanation which is odd giventhe fact that Kyburg seeks to resolve his own puzzle by denying the parallelconjunctive closure principle for reasonable acceptance Colin Howsonhas suggested rejecting the conjunctive closure principle for explanationin order to hold on to a high epistemic probability model without beingthereby saddled with a low epistemic probability model38 He points outthat lsquothere is no general support for such a closure principlersquo In view ofthe havoc conjunctive closure rules would bring in an example such asthe set of reasonably accepted statements Howson counsels arguing caseby case for their use and not assuming universally as Salmon seems todo that conjunctive closure rules are reasonable However even if we donot accept Salmonrsquos second argument for a low epistemic probabilitymodel of explanation adapted from Kyburg the sorts of examples he andvan Fraassen cite still constitute some evidence in favour of such a theory

There is even a strong and a weak version of a low epistemic probabilitymodel of explanation A strong low epistemic probability model willrequire that the explanans raise the probability of the explanandum fromsome prior probability even though the resulting probability may still below The weak low epistemic probability model allows an explanation tofurther lower the explanandumrsquos probability from some already low priorprobability (There is an analogous variant of the high epistemic probabilitymodel according to which an explanation can lower an explanandumrsquoshigh probability) Salmon for example admits into an explanationexplanatory factors which have such a negative effect on the epistemicprobability of the explanandum39

An example of such a low epistemic probability explanation whicheven further lowers the probability of the explanandum according toSalmon is this consider a mixture of uranium 238 atoms with a verylong half-life and polonium 214 atoms with a very short half-life40 Theepistemic probability of an atomrsquos disintegrating if one knows that it wasuranium is low the epistemic probability if one knows that it was poloniumis high The epistemic probability that some unspecified atom in themixture will disintegrate is somewhere between that for uranium 238 and

32


polonium 214 atoms This gives the epistemic probability of disintegrationfor an unspecified atom in the mixture which we can assume is low

Suppose some atom disintegrates in a short space of time and wewish to explain this If we learn that it was a uranium atom thatdisintegrated then the explanans is it was an atom of uranium 238 Itsbeing an atom of uranium 238 explains the disintegration (Note that theexplanans is not an atom of uranium 238 disintegrated for that entailsthe explanandum) One might object to this example on the grounds thatsimply learning that it was a uranium 238 atom could hardly explain thedisintegration That would almost be like trying to explain somethingsimply by assigning a name to it But it is easy to remedy this defectSuppose there are at least two different nondeterministic causalmechanisms leading to disintegration one in the case of uranium and theother in the case of polonium and what one learns is that one of themechanisms rather than the other leads to the disintegration beingexplained

The difficulty then is this The epistemic probability of its disintegrationgiven that it is uranium 238 is less than the epistemic probability of itsdisintegration given that it was merely an unspecified atom in the mixtureand yet we can explain why the unspecified atom disintegrated by sayingthat it was an atom of uranium 238 So according to Salmon lsquothe transitionfrom the reference class of a mixture of atoms of the two types to areference class of atoms of U 238 may result in a considerable loweringof the weightrsquo41

The epistemic probability models admit that there are different kindsof full explanation (certainty-conferring explanations and epistemicprobability-conferring explanations) They might even admit that therecan be these two different kinds of full explanation for the sameexplanandum Suppose that there is an explanation for something thatmeets the certainty requirement and another (full) explanation for thesame thing which only meets the weaker epistemic probabilityrequirement Must the high or low epistemic probability modellist admitthat the explanation meeting the certainty requirement is necessarily thebetter I do not think that they need to admit this A good explanation isone that meets the interests and assumes what it should assume about thebeliefs of the audience We may prefer the epistemic probability-conferring explanation as being simpler less unwieldy and more intuitivein short a better explanation given what we want

(B) Specific theories of explanation can also be distinguished by theirviews on the form an explanation must take argument theories and non-

33


argument theories of explanation The three examples of explanation atthe beginning of this chapter did not appear to be arguments butappearances might be deceptive Perhaps the examples should be recastin the form of an argument or perhaps they are parts of arguments therest of the argument being implicitly understood

An argument theory of explanation uses the idea of an argument togive substance to the ideas of both certainty and epistemic probabilityObviously on no specific argument theory of explanation does just anyargument count as an explanation Different theories will add differentfurther necessary conditions for explanation

On an argument theory what kind of argument is an explanationDeductivism and probabilism differ about what sorts of argumentsexplanations can be Deductivists are certainty modellists who hold anargument theory probabilists are epistemic probability modellists whohold an argument theory If an explanans can fully explain an explanandumonly when the explanandum is certain given the explanans informationand if all full explanations are arguments full explanations must bedeductive arguments Deductivist theories require that all full explanationsbe inter alia deductively valid arguments This is Karl Popperrsquos view ofcausal explanation42 On a theory of this type only a deductively validargument could count as a full explanation

On a probabilist theory since an explanans can fully explain anexplanandum even if the explanandum is only probable (to a degree lessthan 1) given the explanans information inductive arguments provideanother sort of argument that fits the bill (in addition to deductivearguments) Hempelrsquos theory of explanation since it admits both D-Nand I-S explanations is of this type For the probabilist some explanationsare good non-deductive arguments whose premisses support or makeprobable (to a degree less than 1) their conclusion It is important to recallthat the claim is about full explanation since even a deductivist can allowa non-deductive relation between the explanans and explanandum in apartial explanation The probabilist will accept that if there is sufficientinformation from which an explanandum can be deduced then there willbe an explanation that meets the deductivist requirement But if there isinsufficient information for a deduction then we may have a fullexplanation which shows that the explanandum was epistemicallyprobable given the relevant information in the explanans

Note that since there is no such thing as a valid non-deductive argumentwhose conclusion is improbable on the premisses the only viable sort ofargument variety of the epistemic probability model is one which claims

34


that an explanation may be a non-deductive argument whose premissesrender the conclusion highly probable There is no argument version of alow epistemic probability model of explanation43

Suppose that the assumption that all explanations are arguments wererejected An explanation of something is constituted by a certain kind ofinformation about the thing but such information may not necessarily ormay even never have the form of an argument There are various possiblenon-argument theories explanations are single sentences or lsquoa storyrsquo orthe conjunction of an argument and an addendum sentence or a list orassemblage of statistically relevant factors44 or an answer45 and so onThe only thing such non-argument theories have in common is that theyreject the assumption that all explanations are arguments Wesley Salmonwas one of the first contemporary philosophers to question the view thatexplanations are arguments46 historically the argument view is entirelyabsent in Platorsquos writings and only first makes its appearance withAristotle

On the non-argument view there can still be relative certainty andepistemic probability models of explanation it is just that these ideascannot be cashed out by means of the idea of an argument of a certainsort A non-argument (relative) certainty theorist might say for examplethat an explanation is a sentence which states that the explanandum iscertain (has an epistemic probability of 1) given the explanans a non-argument epistemic probability theorist (a high or low epistemicprobability theorist or a HEP or a LEP theorist) says that an explanationin addition to the above can be a sentence that assigns an epistemicprobability of less than 1 to an explanandum given the explanans Suchnon-argument theorists would take explanations to be sentences whichattribute conditional epistemic certainty or probability Of course thereare many other non-argument theory possibilities I mention these onlyby way of illustration

I have called the general ideas of certainty and high and low probabilityconferring explanations lsquomodelsrsquo of explanation These models have bothargument and non-argument forms I now reserve the terms relativecertainty theory high epistemic theory and low epistemic theory (HEPtheory and LEP theory) for the specifically non-argument forms of thesegeneral doctrines The relative certainty theory is the non-argumentanalogue of deductivism high epistemic probability theory (HEP theory)the non-argument analogue of probabilism low epistemic probabilitytheory (LEP theory) has no argument analogue

35


(C) An explanation is a piece of information and the above distinctionshave been concerned with the form in which such information has to bepresented and its epistemic status I now want to concentrate on whatsuch information must be about We can distinguish theories ofexplanations by the metaphysical presuppositions they make about thereality they seek to explain

It will be helpful in introducing this typology to assume somethingthat I regard as false all explanations of singular events or states of affairsare causal explanations I will discuss this assumption in chapter VIIand broaden the kinds of singular explanations that there can be It willthen be easy to broaden the typology to take account of this having alreadyintroduced it on the narrower assumption But in the interim I will bemaking this (admittedly false) assumption

Is causation (the idea of the whole or full cause) a deterministic conceptSome accounts of causation hold that it is and others deny this I do notwish in this book to become entangled in questions about the correctanalysis of causation (if indeed there is one to be had at all) So mydiscussion allows for the possibility that the idea of causation is notdeterministic and hence that some events may have nondeterministiccauses without assuming that this is so

lsquoIndeterminismrsquo might name either of two doctrines (1) some eventshave no cause (2) some events have a nondeterministic cause (whetheror not there are events with no cause of any kind) A nondeterministiccause if such there be is a cause but of a special sort To keep doctrinesstraight I use lsquoindeterminismrsquo as the name of the doctrine that assertsthat some events have no cause of any kind lsquonondeterminismrsquo as thename of the doctrine that claims that the idea of causation is notdeterministic and that some events have nondeterministic causes

The idea of a deterministic cause is that the cause necessitates theoutcome that only one outcome is possible given the cause (and therelevant laws of nature) Different theories will spell out the idea ofdeterminism and necessitation differently A constant conjunction theorymight explain the idea of a deterministic cause in one way (it might addthat the idea of necessitation if it goes beyond constant conjunction isillicit) a counterfactual analysis will provide an alternative way in whichto understand determinism and necessitation If we remember the earlierquote from von Wright his use of lsquonecessaryrsquo might also be taken in thismetaphysical sense an explanation lsquotells us why [an event] E had to be(occur) why E was necessary once the basis is there and the laws areacceptedrsquo (von Wright 197113)

36


But many philosophers believe that some cases of causation are notdeterministic and hence that causation is not itself a deterministic conceptThe nondeterministic view of causation has been argued by philosophersotherwise as different as David Lewis John Mackie Patrick Suppes andWesley Salmon although these philosophers have tried to capture theidea of nondeterministic causation in different ways47 A cause as Mackieargues is not necessarily strongly sufficient for its effect48 If c causes eMackie says that it does not follow that if e had not been going to occurc would not have occurred A cause on the Suppes view raises theobjective probability or likelihood of the occurrence of its effect fromwhat it would have been had the cause not occurred but it does notnecessarily confer a probability of 1 on its occurrence On Lewisrsquos accountnondeterminism affects the necessity of a cause for its effect not just itssufficiency In Lewisrsquos parlance causation is chancy if the cause had notbeen the effect would have had less of a chance of occurring For all ofthese writers effects depend on their causes without being determinedby them

Proponents of these nondeterministic views of causation are oftenresponding to cases from quantum physics but cite quite ordinary cases aswell as supporting this nondeterministic analysis49 The kind of chance orprobability needed by a nondeterministic view of causation is objectivephysical probability although it is open to such a view to adopt any one ofa number of competing theories about how such chance or probability is tobe understood (relative frequencies single-case propensities and so on)

Given our earlier assumption that all explanation of singular eventsis causal explanation theories of explanation can be classified asdeterminative high or low dependency theories of explanation Adeterminative theory asserts that a necessary condition for a cause cfully to explain its effect e is that c physically determine or necessitatee A high dependency theory asserts that a necessary condition for acause c fully to explain its effect e is either that c necessitate e or that ehighly depend on c (given the cause its effect has a high physical orobjective probability) A low dependency theory claims that in a fullexplanation an effect e might depend only slightly on its cause c (giventhe cause its effect has only a low chance or physical probability ofoccurring)

Suppose that the outcomes that result from flipping a coin wereobjectively nondeterministic If I flip such a coin which is heavily biasedfor heads and obtain a heads and if my flipping of the coin was the causeof its landing heads my flipping caused the landing of heads without

37


necessitating it Its landing heads highly depended on but was notdetermined by the flipping On the other hand if I flip such a coin andobtain the improbable outcome of a tails and if my flipping of the coinwas the cause of its landing tails then its landing tails although causedby the flipping was improbable or unlikely given the flipping

Note that a determinative theorist need not deny that there arenondeterministic causes or deny that there are some events with no causein any sense A determinative theorist can accept indeterminism andnondeterminism He has only to deny that nondeterministic causes explaintheir effects and deny that there are explanations for whatever is uncausedFor him the nondeterministically caused and the uncaused areinexplicable

Plato and Mill are proponents of a determinative view of explanationFor Plato an explanans is the cause of its explanandum and the causationis a form of compulsion or forcing Mill (at least on the standard view ofhim) also holds that all explanation of singular events or conditions is byway of causal explanation and that all causes are deterministic

The determinative theorist will subscribe to some version of thecertainty model of (full) explanation If we explain only by citing causesand if all explanatory causes are deterministic then full explanations arealways relative certainty-conferring A full explanation must give the wholecause and since the effect is the only possible outcome given the wholecause we can be certain that the effect will occur if we know that thecause has occurred Ignorance may mean that we do not know an effectrsquosfull cause or for the sorts of pragmatic reasons discussed above we maythink it worthwhile to cite only some part of the cause (or some part ofthe causersquos relevant description) When this is so there can be partialexplanations that do not meet the certainty-requirement But theexplanation would be only partial50

High and low dependency theories of explanation allow explanationon the basis of nondeterministic causes Statistical information on thisview has explanatory relevance only when that information relates to themechanisms that produce that outcome Such mechanisms can benondeterministic ones whose outcomes are highly likely but not necessaryor may even be unlikely (as in the example above of flipping thenondeterministic coin and getting the improbable tails) I agree with thespirit of Salmonrsquos recent judgement according to which the relevance ofstatistical relations for explanation can only be indirect Statistical relationsare evidence for causation but the causation for which they are evidencemay itself be of a probabilistic or nondeterministic sort51

38


What relation is there between high and low dependency theories ofexplanation and the two epistemic probability models of explanation (ineither an argument or non-argument version) The high dependencytheorist need not deny that there can be a full cause c of an effect e suchthat the conditional probability of e on c is low Such a theorist needsonly to assert that such a cause is unable to explain its effect Howeverneither a high nor a low dependency theorist will adopt a certainty modelof explanation Their commitment to physical chance will be reflected intheir commitment to an epistemic probability model of full explanation Ifsome full causes are nondeterministic and if we can sometimes explainby citing them then on the basis of the laws of nature and the occurrenceof the cause it will sometimes be only highly epistemically probable (onone version) or even epistemically improbable (on the other version) thatthe effect will occur

Although the view that some (unexplanatory) causes arenondeterministic is consistent with a determinative theory of explanation(as I argued above) the sole convincing motive as far as I can see forholding either a low or a high dependency theory of full explanation andhence any version of the high or the low epistemic probability model offull explanation (in either an argument or a non-argument form) is abelief that the world contains in some measure nondeterministic causesand that explanations are sometimes possible in spite of this Since bothof those beliefs seem eminently plausible the motive is a good one

If Hempel had said that all singular explanation was causal explanation(he does not hold this view and we shall want to look at this very carefullyin chapters IV and VII) we could have classified him as a proponent of ahigh dependency theory of explanation The case for a low dependencytheory of explanation has been made both by van Fraassen and Salmon52

The example of the improbable decay of the uranium nucleus cited earlieras evidence for a low epistemic probability theory of explanation canalso be used as evidence for a low dependency theory of explanation

Determinative high and low dependency theories of explanation haveimplications for inexplicability as well as explicability We would expectthat the higher the demands for explanation the more will turn out to beinexplicable So on a determinative theory events with no causes or withonly nondeterministic causes are inexplicable whatever else we mightknow about their occurrence That may seem to make too muchinexplicable

On the other hand one might wonder whether a low dependency theorydoes not set the standards for explicability too low and make too little

39


inexplicable Letrsquos accept that the tunnel effect causes the uraniumnucleusrsquos decay when it does decay even though it has a low conditionalprobability of decay given that effect But it does not follow from the factthat the tunnel effect was the cause of its decay that the tunnel effectexplains its decay

Finally a high dependency theory might seem to divide explicabilityfrom inexplicability in an ad hoc way Salmonrsquos arguments discussedabove bring out one of the oddities of a high dependency theory for thequestion of inexplicability In the case of the tossing of a coin with astrong bias for heads the coinrsquos landing heads is explicable but its landingtails is inexplicable although the same information is relevant in bothcases

The classifications I have introduced may seem overcomplicated butI think they are necessary to make sense of the literature on explanationOn the one hand the classification shows how metaphysical beliefsespecially about determinism and nondeterminism play an essential rolein onersquos theory of explanation On the other explanation is an epistemicnotion and the classifications show how metaphysical commitments haveconsequences for the epistemology of explanation

Dispensing with contrastives

The lsquotraditionalrsquo view to which I subscribe holds that in anexplanation one explains facts such as the fact that p (I shall conductthe discussion in terms of facts but there are analogous statements ofthe view in terms of events in an explanation one explains forexample an event e)53 A number of writers have disputed thetraditional view and argued that what is explained in every explanation(including full ones) is something with a contrastive form thecontrastive fact that p rather than q (or the event e rather than theevent f)54 The contrastive theory claims that a question such as lsquoWhyprsquo is always implicitly of the contrastive form lsquoWhy p rather thanqrsquo and that a question such as lsquoWhy prsquo typically will be ambiguousfor there are likely to be several different contrasts with lsquoprsquo whichare possible

Moreover the contrastive suggestion is that whatever insights arecontained in the contrastive idiom cannot be captured by the traditionaltheories of explanation which ignore that idiom The contrastive view

40


utilizes contrast spaces (Garfinkel) or contrast classes or a set of alternatives(van Fraassen) and these are unaccounted for on the traditional theories55

This contrastive view should not be conflated with a distinct viewargued for by David Lewis that links contrast to the pragmatics ofexplanation-giving He claims that contrastive stress or explicit contrastiveformulation is or can be a good way to indicate what part of a maximallytrue answer to a question is the part one wants to hear on a particularoccasion56 On Lewisrsquos view a maximally true answer to an unambiguousquestion is non-contrastive Lewisrsquos view is not my target

An anecdote about the American bank robber Willy Sutton is intendedto illustrate the contrastive view that I have in mind When asked by apriest why he robbed banks Sutton replied that it was because that waswhere the money was kept The contrastive diagnosis of this anecdote isthat the fact to be explained that Sutton robs banks is merely ellipticalfor some contrastive fact and that there is more than one contrastive factfor which it might be elliptical The priest was no doubt wishing for anexplanation of Suttonrsquos robbing banks rather than not robbing at all Suttonreplied with an explanation of why he robbed banks rather than otherinstitutions

Although I have no objection to the contrastive terminology (and indeedwill occasionally use it) I have two points of dispute with the contrastiveview I have described First I do not believe that all explananda arecontrastive Second even when contrastive terminology is appropriate itseems to me that whatever insight it makes can be set out just as well intraditional terms using non-contrastive facts (or events and event failures)To simplify my discussion I will hereafter speak only in terms of theexplanation of facts the extension of the argument to event explanationis straightforward

The first question is whether all explained facts are contrastive SupposeI want to explain the fact that Carl is a good philosopher I do so bydescribing his excellent philosophical training in the company of greatphilosophical masters Sometimes I might have contrasts in mind whyhe is a good philosopher rather than a good carpenter why he rather thanHans is a good philosopher But sometimes at least there is no obviouscontrast and in the example I have mentioned the information about histraining seems to explain why he is a great philosopher tout court Thereis no obvious lsquorather thanrsquo about it

A contrast theorist has a ready reply to this In such a case I explainthe fact that Carl is a good philosopher rather than his not being a goodphilosopher Sometimes the contrastive fact is this the fact that p rather

41


than ~p This was the contrastive strategy in the Sutton example lsquoThepriest was no doubt wishing for an explanation of Suttonrsquos robbing banksrather than not robbing at allrsquo

However the fact that p rather than ~p is just a tedious pleonasm forthe fact that p I do not claim that if one explains p then one has ipsofacto explained every proposition logically equivalent to p But if it is afact that p it follows by double negation which only those bordering onidiocy could fail to appreciate that it is not a fact that ~p There is nothingmore here to explain a person explains the fact that p rather than ~p iff heexplains the fact that p So some explanations are not contrastive I thinkthat the priest wanted such a non-contrastive explanation from Sutton Intruth what the priest wanted to know was why Sutton robbed no non-pleonastic contrast being required

There is no doubt that some explanations are contrastive What isinvolved in explaining genuine contrastives eg why event e rather thanevent f or explaining the fact that p rather than q One plausible-seemingthought is this to explain the fact that p rather than q is just to explainthe fact that pamp~q This view makes the pleonastic nature of lsquothe factthat p rather than ~prsquo clear for it would be equivalent to lsquothe fact thatpamp~~prsquo

Dennis Temple believes that this is the correct analysis for contrastives57

To explain a contrastive fact is to explain a certain type of conjunctivefact Thus in explaining why he robbed banks rather than other institutionsSutton was explaining why he robbed banks and did not rob otherinstitutions If I explain why I live in London rather than Boston I explainwhy I live in London and do not live in Boston

One argument against Templersquos plausible-seeming view runs as followsFor any arbitrary p and ~q suppose I explain the fact that p and then Iexplain the fact that ~q Let lsquoprsquo be lsquosnow is whitersquo and let lsquo~qrsquo be lsquoit is notthe case that grass is redrsquo If it then followed that I had explained the factthat snow is white and grass is not red we would have a simple argumentagainst Templersquos suggestion Even if I have explained the fact that snowis white and grass is not red I certainly have not explained the fact thatsnow is white rather than grass is red

In order to save Templersquos analysis we cannot allow that explanation isclosed under conjunction That seems independently plausible sinceKyburgrsquos conjunctivitis seemed to teach the same lesson So if I explainthe fact that p and I explain the fact that ~q it does not follow that I haveexplained the fact that p and ~q The argument against Templersquos suggestionis blocked if explanation is not closed under conjunction Let (a) be

42


lsquohellipexplained explained the fact that p rather than qrsquo let (b) be lsquohellip explainedthe fact that pamp~qrsquo let (c) be lsquohellipexplained the fact that p amp explainedthe fact that ~qrsquo Templersquos claim would be that (b) is the analysis of (a)For Templersquos analysis to stand (c) cannot be sufficient for (b) since (c) iscertainly not sufficient for (a)

(c) cannot be necessary for (b) either Untreated latent syphilis is theonly cause of paresis but only a small number of those who have untreatedlatent syphilis develop paresis Suppose Jones but not Smith has untreatedlatent syphilis and Jones gets paresis I can (fully) explain why Jonesrather than Smith developed paresis on the grounds that Jones but notSmith had untreated latent syphilis But in view of the small number ofthose with untreated latent syphilis who develop paresis I might not havefully if at all (depending on onersquos view of explanation) explained whyJones got paresis

On Templersquos view if I explain the fact that Jones rather than Smithdeveloped paresis I have explained the fact that Jones developed paresisand Smith did not develop paresis But in view of the above argumentit does not follow that I have explained the fact that Jones developsparesis If Templersquos analysis is to stand explanation cannot be closedunder simplification either If I explain the fact that pamp~q (p rather thanq) it does not follow that I have explained the fact that p and a fortioriit does not follow that I have explained the fact that p amp explained thefact that ~q

It is not so much that Templersquos analysis is faulty Rather it is that (b)lsquoexplains the fact that pamp~qrsquo does not really illuminate (a) lsquoexplains thefact that p rather than qrsquo We cannot understand the conjunction sign withinthe lsquoexplainsrsquo context in the normal truth-functional way Neither theconjunction nor the simplification rule holds It is not then clear whatlsquoexplains the fact that pamp~qrsquo is supposed to mean and the suspicion isthat we can only understand the purported analysans lsquoexplains the factthat pamp~qrsquo in so far as we understand lsquoexplains the fact that p rather thanqrsquo which is supposed to be the analysandum

A difference (perhaps there are others) between (a) lsquoexplaining thefact that p rather than qrsquo and (c) lsquoexplaining the fact that p andexplaining the fact that ~qrsquo seems to be that (a) requires some sort ofrelevance or connection between lsquoprsquo and lsquo~qrsquo and that (c) does notrequire this I am doubtful that there is a single way in which to capturethe relevance relation between the fact that p and the fact that ~q in allcases of explaining the fact that p rather than q One way which Idescribed in an earlier article is this the fact that p lsquoeclipsesrsquo the fact

43


that q In some cases in which the explanation of the fact that p ratherthan q is more than the explanation of the fact that p and the fact that~q the lsquomorersquo has to do with this eclipsing

For example suppose I want to explain why a certain stone was inLondon rather than Boston in the late evening of 7 January 1986 It isinsufficient to explain only why it was in London in that late eveningbecausemdashfor all we know so farmdashthe stone might have been in bothplaces during the course of that late evening

Often what is needed in order to explain why the fact that p ratherthan q (why the stone was in London rather than in Boston) is anexplanation of how or why the fact that p (its being in London) isphysically inconsistent with the fact that q (its being in Boston at thattime) As I shall say the fact that p eclipses the fact that q wherelsquoeclipsesrsquo means lsquocausally or physically preventsrsquo In many cases likethat of being in London rather than Boston a person explains why thefact that p rather than q iff that person explains how or why the factthat p eclipsed or prevented the occurrence of the fact that q Theadditional information needed is sometimes minimal indeed oftenquite trivial In the case of the presence of the stone in London ratherthan Boston it is merely the information that its being in Londonphysically prevents its being in Boston in the course of the same lateevening

Peter Lipton argues that cases of choice surprise and discriminationcannot be handled by my lsquoeclipsingrsquo analysis The paresis casediscussed above is a case of explaining a discrimination Jonesrsquos gettingparesis does not eclipse and is not physically inconsistent with Smithrsquosgetting paresis as well In this case to explain why the fact that prather than q is to show that a causally necessary circumstance for thefact that p was absent in the circumstances that led up to the non-occurrence of the fact that q For example untreated latent syphilispresent in Jonesrsquos case was absent in Smithrsquos The analysis of thisexample will not replace my eclipsing analysis but will have to beadded to it

I do not assert that the cases of eclipsing and of the presence-absenceof causally necessary circumstances between them exhaust the contentof all genuinely contrastive fact explanations (I have not discussed casesof choice and surprise also mentioned by Lipton) There may be othersBut if these two are indicative I think that we can say that there are noexplanations of irreducibly contrastive facts These facts are reducibleto (perhaps relational) non-contrastive facts the fact that one thing

44


prevented another or the fact that something was present in one casebut absent in another It seems to me that such explanations can all behandled by techniques available to standard (non-contrastive) theoriesof explanation So I stick to the traditional terminology in which todiscuss explanation

45

CHAPTER II

Plato on Explanation

In one sense the whole of Platorsquos theory of the Forms can be read asan extended discussion of the requirements for explanation Howeverwhat Plato has to say explicitly about explanation is mainly to be foundin the Phaedo 95ndash1071 and in some remarks at the end of the Theaetetus201ndash82 to which I shall turn at the end of this chapter This chapter onPlato is on balance less well integrated into the main lines of argumentof the book than are the other chapters In the main I use Platorsquosstrictures against the explanation of and by opposites as a way in whichto pose the question of (what is usually called in the literature)probabilistic explanation This is not a concept that Plato would havebeen prepared to accept but I do not think it has been generallyappreciated how his explicit remarks on explanation depend on thatnon-acceptance I do not find in Plato many other insights aboutexplanation which I wish to export from his text for my own use

In chapter I I ascribed to Plato a non-argument determinative theoryof explanation My interim account of a determinative theory ofexplanation (to be revised in chapter VII) is that it is one which assertsthat the explanation of a particular is always by way of the deterministicor necessitating cause of that particular However terms like lsquonecessitatingrsquoand lsquocausersquo did not mean for the Greeks what they mean for us and someaccount needs to be taken of this in attributing such a view to Plato

In spite of Vlastosrsquos spirited attempt to read a distinction between logicaland physical necessity back into Platorsquos text I do not believe that the text willbear the distinction3 Vlastos rests his case on the Phaedo 97a 2ndash5 but it

46


does not seem to me to bear out the distinction that he wants I agree withEvan Burgersquos judgement4 lsquoWhat is not made clear [in Platorsquos Phaedo] is thedifference between different kinds of necessity in particular the differencebetween logical and physical necessityrsquo (Burge 19718) When I attribute aview about necessity or necessitation to Plato I think it should be understoodas an undifferentiated idea of necessity covering (what we would call) logicalor mathematical necessity metaphysical necessity and physical necessity

The Greek term aitiai in spite of its being standardly translated aslsquocausersquo had for the Greeks and hence for Plato and Aristotle a muchwider sense than it has for us For us a cause is the efficient cause thatwhich moves something or puts into motion some event process orwhatever As Vlastos reminds us all of the following are for the ancientGreeks statements giving somethingrsquos aitiai the Persians invaded Atticabecause the Athenians raided Sardis this statue is heavy because it ismade of bronze he is taking an after-dinner walk because of his healththis angle at the semicircle is a right angle because it is equal to the half oftwo right angles (the examples are Aristotlersquos) When Plato uses lsquoaitiairsquoin the Phaedo what does he mean by it I return to this question aftersketching an outline of his discussion of explanation

Finally Plato introduces three ontological lsquolevelsrsquo as it were into hisdiscussion things like physical objects and numbers which I representby lsquoxrsquo lsquoyrsquo lsquozrsquohellipetc the Forms by lsquoArsquo lsquoBrsquo lsquoCrsquohellip(Tallness Coldness)and the individual instances of properties in the object by lsquofrsquo lsquogrsquo lsquohrsquo (mytallness the rockrsquos coldness) The distinction between things and theirproperties is clearly drawn at 103bndashc

Then we were talking about the things which possess the oppositescalling them by the same name as the opposites themselves havebut now we are talking about those opposites themselves which bytheir presence give their names to the things called after themhellip

At 102dndashe (and elsewhere) Plato distinguishes between Tallness initself and lsquothat tallness which is in usrsquo that is Tallness and Shortnessfrom the tallness of Phaedo and the shortness of Socrates Letrsquos callthese latter lsquoindividual characteristicsrsquo as distinct from the Formslsquohellip not only is the Form itself entitled for ever to the name that isgiven to it but also something else which while not the same thingas the Form nevertheless in every instance presents the manifestationof itrsquo (103e) Individual characteristics are that something else

47


The Phaedo

The Phaedo takes the form of a dialogue between Socrates and Cebesand in what follows I identify the ideas expressed by Socrates in thatdialogue as those of Plato himself although this identification mightbe controversial or mistaken in some dialogues5

The discussion falls into three parts In the first (95endash99d) Socratestells of his earlier attraction to explanations in terms of the physical causesof things which were offered by various pre-Socratic philosophers andfor a variety of reasons some of which I shall examine in detail later inthe chapter he found these in the end to be unacceptable Let us say thathe rejects physical explanation Notice that he did not find them less thanfully adequate and in need of supplementation Rather he says they areentirely unacceptable he knows that he doesnrsquot want to follow this allegedmethod of explanation at all (97bndashc)

He then tells Cebes that he turned with high hopes to the sorts ofexplanations offered by Anaxagoras which were supposed to be in terms ofthe best Although Socrates does not offer an example of such an explanationone assumes that they would be akin to what Aristotle would call lsquofinalexplanationsrsquo or what we would call lsquoteleological explanationsrsquo Indeed wemight think of them as lsquosuperlative-final explanationsrsquo for they explain thingsnot just by the good at which they aim but in terms of the best

Socrates makes clear that such final explanations remain for him thepreferred type of explanation to which one should aspire but he expressesdisappointment at Anaxagorasrsquo practice which Socrates claims departedfrom his stated intention since Anaxagoras reverted to the sorts oflsquoexplanationrsquo in terms of physical cause that Socrates had already rejectedAgain he rejects the sorts of physical lsquoexplanationsrsquo which Anaxagorasin fact offers as lsquoquite absurdrsquo (99a) The sorts of things physical explainerscite are at best necessary conditions for what is being explained but theyare not aitiai in any sense of that term lsquoFancy not being able to see thatthe real cause is very different from the mere sine qua nonhellip Yet that iswhat most peoplehellipseem to call ldquocauserdquo using a name that doesnrsquot belongto itrsquo (99b) On the other hand Socrates is clear that explanation in termsof the best if there is such a thing is genuine explanation involving thecitation of an aitiai

In the second part (100bndash103a) Socrates introduces his own admittedlysecond-best approach to explanation in terms of the Forms I call theselsquoFormal explanationsrsquo Socrates first gives an account of the lsquosafersquo version

48


of such explanations As Plato has Socrates saying at 100c lsquoIt seems tome that if there is anything else beautiful beside the Beautiful Itself it isso purely and simply because it partakes of that ldquoBeautifulrdquorsquo Notice thatthese explanations since they employ all three levels mentioned abovealthough safe are not as obviously trivial as one might fear Theexplanations do not just have this form x is F because it participates in F-ness Rather it is this x is f because it participates in A-ness An examplemight be Why is this rock cold It has its coldness (f) because of theForm of Coldness (A) in which it participates

Forms can never become characterized by their opposites The Formof Beauty can never become ugly the Form of Tallness can never beshort Similarly for the individual characteristics in things Individualtallnesses can no more admit the short individual beautifulnesses no moreadmit the ugly than can the Forms themselves lsquohellipwhat is more hellipthattallness which is in us never admits the short and will not be overcome byithellipeither it flees and beats a retreat whenever its opposite the shortapproaches it or else when that comes it has perishedrsquo (102e) A thingx which becomes short does so only by the departure of its individualtallness which is forced to flee the arrival of its shortness

hellipnot only do these opposites [the Forms and the individualcharacteristics in us] refuse to admit each other but also thosethings which are not opposite to each other but always containthe opposites will not admit that character which is opposite tothe character that they containmdashinstead when it attacks theyeither perish or retreat

(Phaedo 104c)

In the third and last part of the discussion (103andash106a) Socratesintroduces a more informative less safe version of this kind of Formalexplanation A thing which is f is so not just because it participatesin the Form A-ness but because it participates in some other FormB-ness and its being B compels it (Platorsquos term at 104d) to participatein A and hence be f Why is this rock hot Because it participates inthe Form of Fire and in virtue of this it is compelled to be hot (103cndashd) and indeed also compelled not to be cold lsquofire [brings with it] theopposite of the coldrsquo (105a) As Socrates says Threeness will lsquocompelrsquo(104d) anything which admits it to be Odd prevent it from admittingthe Even

49


One way in which to capture some part (but I doubt that it is all) ofwhat Plato intended by this metaphoric talk of Forms compelling andforcing others retreating and perishing is with the idea of necessity TheForm of Fire necessitates or determines something to be hot necessitatesor determines the thing not to be cold Plato is in some sense committinghimself to a determinative theory of explanation the aitiai of a thingrsquosbeing thus-and-so determines it to be as it is Formal explanations includesome element of necessitation

Otherwise though the interpretation of this extended discussion that Ihave quoted is far from simple and uncontroversial If we permit ourselvesfor the moment Aristotelian terminology Plato has introduced three typesof (at least purported) explanation physical explanation (the rejectedlsquoexplanationsrsquo offered by the physicists and Anaxagoras) final explanation(explanations in terms of the best) and Formal explanation (explanationsin terms of the Forms) Efficient causal explanation that closest to ourmodern conception of causal explanation is arguably a more generalcategory than just physical explanation for the former leaves open thepossibility of non-physical efficient causes But this wider notion ofefficient causation has not been introduced by Socrates as a distinct typeall that we have been offered by him is physical explanation which isperhaps a particular kind of efficient causal explanation Physicalexplanations let us say are efficient causal explanations in which theefficient cause is something physical

What kind of explanation is Formal explanation for Plato and howdoes it relate to efficient and final explanation What is clear is that Platorejects physical explanation (explanation in terms of physical things) asany kind of genuine explanation But if there is a notion of efficientexplanation wider than that of explanation by physical causes then perhapsPlato thought that the Forms were another type of efficient cause in placeof physical substances

Vlastos argues against this6 That is he claims that Plato draws adistinction between logical and causal explanation and saw thatexplanation by Forms was a species of the former rather than the latterlsquoWhat Socrates is telling us put into modern language is that the reasonwhy the group of ten is more numerous than the group of eight is simplythat it satisfies the logico-metaphysical conditions of greaternumerousnessrsquo (Vlastos 1969314ndash15) lsquohellipPlatohellipuses the ldquosaferdquo aitiaito explode pseudo-problems which arise when the categorial differencebetween logical and physical aitiai is ignoredrsquo (ibid 325)

50


It is a modern doctrine that abstract (non-spatiotemporal) items canplay no efficient causal role The argument for this is often in terms ofchange whatever causes or is caused must undergo real change andabstract objects are not capable of real change7 The question that thenarises is this did Plato think of the Forms in the same way as we think ofabstract objects

At several places in 102ndash5 Socrates speaks of Forms doing things toparticulars8 He says that Forms approach and take hold of particularsand compel or force them to have certain qualities They depart whenother Forms approach lsquoNot only do opposite Forms refuse to stand firmat each otherrsquos attackhelliprsquo (104c) Unless these are mere metaphors (perhapsthey are but this too needs detailed argument) Forms do not seem to bemuch like abstract objects in the modern sense In spite of what Vlastossays I cannot see any clear evidence in the text that Plato is distinguishingbetween two kinds of explanation at all Whether he ought to have doneso is of course a different matter

Another point in favour of minimizing the difference between PlatorsquosFormal and his efficient causal explanation is that Socrates says that thosewho offer physical explanations (the physical explainers) were attemptingto explain lsquogeneration and destruction in generalrsquo (95endash96b) Platoregarded Formal explanations as answers to the same questions that thephysical explainers were unsuccessfully trying to explain if so Formalexplanations attempt to explain why things come into and pass out ofexistence in terms of the Forms and this seems clearly to be a kind ofefficient causal explanation

Vlastos attempts to find a sharp distinction between two kinds ofexplanation efficient causal and logical in Platorsquos text I agree with thejudgement of Julia Annas to the contrary9 In offering Formal explanationslsquothere is no recognitionrsquo by Plato that this lsquois something totally distinctfrom offering causal explanationshellip Plato has failed to see that he isconfusedly treating together very different kinds of explanationhellip Platoshows no sign of any such grasprsquo that there is a distinction between thetwo kinds of explanation (Annas 1983324ndash5) Given that Plato also wantsForms to be changeless he should have grasped that Formal explanationwas not a kind of efficient causal explanation but there is no sign that hedid fully appreciate this

Are Formal explanations final or teleological It would seem not sincePlato tells us that he is still lsquodeprivedrsquo of teleological explanations andhas taken Formal ones as second best If it was a complaint against physicalexplanation that they lacked any teleological element Formal explanations

51


in both the safe and the informative versions would at first glance seemto fare no better on this score However it is not at all clear how differentFormal explanations and explanations in terms of the best really areVlastos dismisses the possibility that Formal explanation is a type of finalexplanation or contains an element of teleology Taylor argues that allPlatonic explanations are meant to trace back to the Form of the Goodand hence be final or teleological in that sense10 Cresswell followingRS Bluck reminds us that Plato speaks of the particulars as wantingstriving and desiring to be like the Form itself (74dndash75b) which if takenseriously might indicate that Plato did see the relationship betweenparticular and Form on the model of an agent and a goal11 Again thepoint seems inconclusive it is not obvious what relationship Platorsquos Formalexplanation bears to final or teleological explanation

These are however not matters that I wish to pursue further I am muchless interested in questions of pure scholarship concerning Platorsquos systemthan I am in finding whether there is anything in what he says that can be ofvalue in producing a viable theory or account of explanation What I wishto concentrate on then is what Plato says is inadequate about the physicalexplanations advanced by his predecessors I do not discuss all of thesepurported difficulties for there are some which I think need not detain usWhat I do wish to look at are his remarks about opposites explaining thesame thing and the same thing explaining opposites

Platonic explanantia and explananda

The purpose of my discussion in this chapter (unlike my purpose inthe other two historical chapters of the book) is not to set out accuratelyin detail Platorsquos thoughts I do think that there are interesting ideascontained in what he is saying and hence the justification for thischapter But in order to get at these ideas I must read Platoanachronistically by importing into the discussion a number ofcontemporary distinctions and insights that were not available to himI do so shamelessly When I come to look at his examples oflsquounacceptablersquo physical explanations below I shall not be interestedin the detailed examples themselves but in the general message whichhendash perhaps wronglymdashextracted from them

What sorts of entities does Plato think of as being explanantia andexplananda In truth I doubt whether Plato thought much about this

52


question Certainly he thought of Forms or The Best as explanatoryand these seem to be particulars or individuals of some sort Plato islsquopredisposed by his most frequent syntactical usages to regard a requestfor an aitiai as a request for giving an explanation by naming some entityrsquo12

On the other hand as we shall see he surely sometimes thinks ofexplanantia as occurrences or states of affairs a division and a bringingof two things together It is easy to fit the example below of lsquothe headrsquointo this latter pattern it isnrsquot the head that explains but there being ahead difference between the two men and this latter seems to be a state ofaffairs It will suit me in what follows to foist onto Plato an ontology ofexplanation that at least includes states of affairs and occurrencesAlthough a criticism that Aristotle makes of Plato that we shall look at inchapter III seems to depend on taking the Forms as explanatory inthemselves one might insist that it is not the Forms themselves which areexplanatory but the particular or individualrsquos participating in some Formwhich is what explains why the particular is the way it is

In raising Platorsquos problems about physical explanation two words willbe essential the lsquosamersquo and the lsquooppositersquo (explanans or explanandum)In so far as we ask questions about the opposite of what actually happenedour questions can be usefully phrased as questions about what happens inother possible worlds that does not happen in the actual world

There are two different ways in which one might raise Platorsquos problemsabout sameness The first way (a) uses token identity across possibleworlds the second way (b) needs only sameness of type of two non-identical tokens

Consider some token event e that actually occurs We can ask (a) inother possible worlds what would explain that very same token event eor what would that very same token event e have explained If we raisePlatorsquos problems in this first way we must hold all other causally relevantcircumstances constant as we move from possible world to possibleworld13 After all it isnrsquot really just the matchrsquos striking which fullyexplains its lighting but only its striking-in-the-presence-of-oxygen-when-dry etc For ease of exposition I speak as if it is simply a token event likee that explains or is explained but the reader must understand these andsimilar claims in such a way that all other causally relevant circumstancesare implicitly assumed to be co-present with the token event in otherpossible worlds in which it occurs

Alternatively we can ask (b) what would explain or be explained by adifferent token instance ersquo which is of the relevantly (for the purposes offull causal explanation) same type as e Since I use possible worlds in any

53


case to get at Platorsquos idea of opposites I also get at the idea of sameness inthe first way using possible worlds and a single token event But for readerswho are happy with possible worlds but unhappy with token identity acrossthem everything I say about sameness using (a) could be translated intothe second way of speaking (b) (b) achieves what (a) does in terms of thesameness of fully explanatory type to which tokens belong rather than byholding all causally relevant circumstances fixed as one moves from possibleworld to possible world in which the same token event (re)occurs (I assumethe idea of a full explanation here as a primitive notion Platorsquos question iswhether physical explanations could ever count as full explanations)

Problems for the physical explainers

Let me begin by quoting from two passages First 96cndash97b

I had formerly thought that it was clear to everyone that a [man]grew through eating and drinkinghelliponly then did the mass whichwas small become large and in the same way the small manbighellip I used to think that I was justified in my conclusionwhenever a big man standing by a short one appeared to be tallerlsquojust by the headrsquomdashand a horse taller than a horse in the sameway and there are still clearer examples of thismdashten seemed tome to be greater than eight because of the addition of two andthe two-cubit measure to be greater than the one-cubit becauseit exceeded it by half its own lengthhellip I am very farhellipfromfrom thinking that I know the explanation of any of thesethingshellipif you cut one thing in half I can no longer be convincedthat this the division has been the explanation of the generationof lsquotworsquo for there is a cause of the generation of lsquotworsquo oppositeto that of the former instance First it was because they werebrought together alongside of each other and one was added toanother and next it was because one was taken away andseparated from another

The second passage occurs at 100endash101b

So you too wouldnrsquot accept the statement if anyone were to saythat one person was taller than another by a head and that the

54


shorter person was shorter by reason of the same thinghellip Youwould be afraidhellipyou might come up against an opponent whowould say that the taller is then taller and the shorter is shorterby reason of the same thinghellip You would be afraid to say thatten is more than eight by twohellipand that the two-cubit length isgreater than the cubit by a halfhellip Then you would beware ofsaying that when one is added to one the addition is theexplanation of the two or that when one is separated off fromone the division is the explanationhellip

Plato regards these and other features of giving an explanation interms of physical cause as grounds for rejecting this sort of explanationaltogether Let me try to state what some of these features are Thepurported explanations (which in the end he will reject as being bonafide explanation at all) that Plato has in mind with explanans andexplanandum identified are these

(1) Explanandum an instance of two things having come into beingwhere previously there had been only one thingPurported explanans a dividing of that one thing in half

(2) Explanandum one person t being taller than another sPurported explanans by a head or in virtue of a head

(3) Explanandum tenrsquos being more than eightPurported explanans on account of two

(The cubit-measure example seems to repeat whatever point it is that(3) makes)

The gist of Platorsquos objection to these three purported explanations isthis Letrsquos take (1) first If an explanation like (1) were acceptable thensince lsquothere is an explanation of the generation of two opposite to that ofthe formerrsquo then an explanation like (1) would in the appropriatecircumstances be equally acceptable

(1) Explanandum an instance of two things having come into beingwhere previously there had been only one thing Purported explanansan adding of a second thing to that first thing

Platorsquos argument is that if we accepted (1) we might have to accept(1) as well But we cannot accept both (1) and (1) for there cannotbe two lsquooppositersquo explanations of the same thing So it follows thatwe can accept as an explanation neither (1) nor (1) The kind of

55


explanation the physical explainers offer which commits them to therebeing opposite explanations for the same thing is not a kind ofexplanation that we should accept

How could this really be a problem Are the explananda of (1)and (1) about one token instance or two Surely there can belsquooppositersquo explanations for two different tokens But one can seewhat bothered Plato if my remarks in the previous section arerecalled Either Plato is asking whether there could be an explanationof the same token instance the two things coming into existence inanother world in which it occurs (and holding constant all the causallyrelevant circumstances) but an explanation in terms of an addingrather than the dividing or he is asking whether there could be anexplanation of another token instance of what he takes to be therelevantly same (for the purposes of full explanation) type anexample of two things having come into existence where previouslythere had been only one but an explanation in terms of an addingrather than a dividing

We shall have to go along with Platorsquos example and pretend Eitherwe shall have to pretend that types such as an adding of two thingstogether or two things having come into existence where previouslythere had been only one are types under which a particular fullyexplains or is explained or we shall have to pretend that the onlycausally relevant information that has to be held constant across theworlds in which the token adding dividing and generating of twothings occur is for instance that previously there had only been onething Both are utterly implausible but I ask the reader to make thepretence in this and Platorsquos other two examples for I think the lessonhe draws from such admittedly awful examples is worthy of seriousinterest

A similar pattern of argument concerns (2) and (2) and (3) and(3) with the difference that Plato extends his argument to cover thecase of the same explanation for opposite occurrences as well asopposite explanations of the same thing If we were to accept (2) and(3) we would also sometimes have to accept purported explanationssuch as

(2) Explanandum a person s being shorter than another person tPurported explanans by a head or in virtue of a head

(3) Explanandum eightrsquos being less than tenPurported explanans on account of two

56


But it cannot be the case that lsquothe taller is then taller and the shorteris then shorter by reason of the same thingrsquo (101a) So neither (2) nor(2) is acceptable (and similarly for (3) and (3))

Since I have chosen (a) rather than (b) as providing the vocabulary forthe discussion of Plato we can express Platorsquos Principles as

(PP1) Two opposites cannot explain the same thing(PP2) The same thing cannot explain two opposites

The two principles make claims about the explanations there couldbe for and by the same token thing We have already said quite a bitabout what counts as lsquoa thingrsquo for the purposes of the principles Butwhich things are opposites I now turn to the task of elucidating (PP1)and (PP2)

Some terminology

There are a few additional questions of terminology on which I shouldlike to be clear before I begin the discussion proper of the principleson which Plato is relying in making these claims about theunacceptability of some (pairs of) explanations Some of histerminology is merely a historical curiosity as far as I am concernedother points need developing before we can extract anything of interestfrom this

(1) Although I will try to reconstruct Platorsquos arguments using modernnotions of contrariety and contradictoriness in fact the primary sense ofopposition for Plato is the opposition between two Forms like additionand subtraction tallness and shortness more and less Other oppositions(between physical things events bits of language) can only be understoodin virtue of their participation in Forms which are opposed

Moreover for the Greeks there are Forms which are opposite whichwould not seem so to us addition and division for example were oppositesfor them For us there is no opposition in any interesting sense betweenTallness and Shortness as such that is between trsquos being taller (than s)and srsquos being shorter (than t) For Plato there is opposition here simplybecause the two states of affairs include opposite Forms It is clear thatfor Plato opposite states of affairs can indeed (sometimes) mustsimultaneously exist At one and the same time t can be taller than s andshorter than r if t is taller than s it follows that s must be shorter than t

57


I think we will make headway by imposing on Plato the distinctionbetween contrary and contradictory statements in the usual sense whetheror not so doing permits us to remain faithful to the full intention of his textThe terminology is not his and indeed it is clear that the distinction doesnot capture all of his examples of opposites But I think that the modernterminology will help us state perspicuously at least the salvageable core ofwhat Platorsquos Principle is asserting In what follows I speak of events orstates of affairs as being contrary or contradictory but this can easily becashed out as statements about them being contrary or contradictory

If we do impose on Platorsquos text our ideas of contrariety andcontradictoriness the assumption which he made that two lsquooppositersquothings can exist at the same time must be rejected If an event occurs itfollows that its contrary or contradictory (in our sense) cannot haveoccurred For example if x is blue all over at t x cannot be red at t and xcannot fail to be blue all over at t But of course we can still ask aboutwhat the explanation of the contrary or contradictory (merely possible)event xrsquos being red at t or xrsquos failing to be blue all over at t would havebeen if counterfactually it had occurred It is this terminology that I shalluse in what follows

Two events are contrary or contradictory only as described or onlyrelative to specific descriptions Two events can be contradictory whendescribed in one way but display no sort of opposition or incompatibilitywhen described in another Further two events can be contraries whendescribed in one way contradictories when described in another Supposea ball that is blue all over at t had not been blue all over at t Consider theworld in which it is red at t Its being red at t is merely contrary to itsbeing blue all over at t It could also have been yellow at t However if itis red at t it follows that it is not blue all over at t but its not being blue allover at t is contradictory to its being blue all over at t

In chapter V I develop the idea that explanation of an event is only ofan event as described or conceptualized talk of contradictory or contraryevents here should always be understood as being relativized to somespecific description of them A modern doctrine has it that if c causes ethen it is a truth whatever true descriptions of c or e are used in the statementof causality Clearly when I speak of one event causing and explaininganother in this chapter this is not the sense of lsquocausersquo I am using Thereader may think of my use of lsquocausersquo in this chapter as shorthand forlsquocausally explainsrsquo One event could causally explain a second whendescribed in one way but could fail to causally explain the second whendescribed in another way

58


(2) It would be an anachronism to enquire about Platorsquos views on theplace of laws in explanation He had as far as I can see no explicit viewabout this at all To some extent I have brought laws in by the back doorby including all of the information from the appropriate law as part of theconstant context for the particular token (or alternatively by insistingthat the relevantly similar type in fact be the type which would occur inthe statement of the appropriate law) Later I will introduce a furtherreference to laws The historical Plato notwithstanding we shall not getvery far unless laws find some place in the statement of Platorsquos Principles(although it need not be the same place given them by other theories ofexplanation

(3) The middle period Plato of the Phaedo had not as yet sorted outthe distinction between relational and non-relational properties It is onlythe Plato of the late dialogue the Sophist who is able to draw thisdistinction14 For us a person t being taller than another person s and aperson s being shorter than t are two different descriptions for the samestate of affairs For Plato who would read all this non-relationally sincebeing taller and being shorter are opposites trsquos being taller (than s) andsrsquos being shorter (than t) are not only not the same state of affairs but areopposite states of affairs Similarly Plato regards 10rsquos being more (than8) and 8rsquos being less (than 10) as two opposite states of affairs It is notworth our while to follow Plato in this tangle the examples I employ indeveloping Platorsquos Principles will be non-relational in our sense

Platorsquos Principles

My interest in the ensuing discussion will be to see what might besaid in favour of Platorsquos Principles Plato to be sure used hisprinciples to discredit physical explanation altogether The logic ofhis argument was that if one accepted any such explanation then onewould have to accept the opposite explanation and so since oneshould not accept both one should accept neither For Plato anacceptable explanation is one such that there is no possibility of therebeing the opposite explanation at all and he thought that onlyexplanations in terms of the Forms (and presumably final explanationsas well) but never physical explanations could meet this requirementA more plausible use of (PP) might be to assume that some physicalexplanations are acceptable but if they are then the opposite physical

59


explanations are unacceptable It is the latter use of (PP) in which Ishall be interested

Plato himself can be construed as using (PP2) in this more plausibleform in his discussion in The Republic of the tripartite nature of the soul(436andash441c) Plato begins the argument by obtaining Glauconrsquos agreementto the following principle lsquoIt is obvious that the same thing will never door suffer opposites in the same respect in relation to the same thing andat the same timersquo (436c) If we were to replace lsquodo or suffer oppositesrsquo bylsquoexplain oppositesrsquo we obtain something very much like (PP2) but withthe idea of lsquoone thingrsquo in lsquoone thing cannot explain oppositesrsquo morefinely sharpened to include a specification of time respect and relation

The use that Plato then makes of the principle stated at 436c is indeedconcerned with the requirements of explanation in particularpsychological explanation He considers the case of a man acting in variouscontrary ways (lsquocontraryrsquo in his sense but not in ours) the man who bothdesires and refuses to drink (439c) the man who desires to see corpsesand at the same time is repelled by the idea (440a) Platorsquos moral is thatone cannot explain both of the contrary desires in each pair by the sameexplanans his soul moves him Such an explanation in terms simply ofthe Soul would explain the opposites of desiring and refusing desiringand being repelled by the same thing The solution is to refuse to treatthe soul as simply one thing One faculty of the soul reason is one thinganother faculty of the soul high-spiritedness is another thing In each ofthe contrary pairs one desire must be explained by one thing one part ofthe soul the other by a different thing a different part of the soul

Even if we take into account all the remarks about terminology that Imade above Platorsquos Principles are still not expressed very precisely Letlsquodrsquo refer to a token dividing in half of something and lsquogrsquo to a tokengenerating of two things where previously there had been only one andlsquoDrsquo and lsquoGrsquo to the appropriate (adequate for the purposes of fullexplanation) descriptions or types respectively I let lsquo~drsquo lsquo~grsquo lsquo~Drsquo andlsquo~Grsquo refer to tokens and types of these actions or events failing to occurIf d is a token event ~d is an event omission drsquos failure to occur at aspecific time and place I assume that at least sometimes failures andomissions can both cause and explain As I indicated before I assumethroughout that although both d and ~d occur in two different possibleworlds the worlds agree in respect of all other causally relevantcircumstances

We might think that we could represent Platorsquos Principle (PP1) as

60


(4) If d and ~d are contradictories and if d explains g ~d does notexplain g

(5) If d and e are contraries and if d explains g e does not explain g

(4) and (5) cannot be what we need because they are trivially true Ifd explains anything then d occurs If d occurs it follows that neither~d nor e occurred (if d and e are contraries then at most only one ofthem can occur if d and ~d are contradictories one but only onemust occur) But since what did not happen does not explain anythingit follows that if d explains g then neither ~d nor e does explain g Ifthere being two things where there had been one is caused andexplained by a dividing that same instance of there being two thingswhere there had been one is not also caused and explained by anadding This surely cannot be all Plato is trying to tell us

The above argument relies on the premiss that what does not happendoes not explain anything Is this really true Canrsquot we sometimes explainthings on the bases of lacks failures and other sorts of absences Ofcourse we can but the sense in which we can does not constitute acounterexample to my claim For example suppose my failure to cometo the party explains why the party was a bore The occurrent token eventwhich has explanatory force is my failing to come to the party My failureto come is what did happen What does not occur viz my not failing tocome to the party (viz my coming to the party) is what has no explanatoryforce

So we need a rendering of (PP1) which has its consequences in thesubjunctive mood If d and ~d are contradictories or if d and e arecontraries then if d explains g trivially neither ~d nor e do explain thesame token g nor indeed do they explain anything else for that mattersince it is impossible for either to occur if d does The right question isnot do ~d or e explain but rather could they have explained As Iclaimed before if we use the terminology of (a) in which to express thePlatonic puzzle the right question must be about an explanation therecould or could not have been The first of Platorsquos Principles is expressiblein some form such as this (this is not the final and ultimately acceptableversion)

(6) If d and ~d are contradictories and if some token event d explainssome token event g then there is no possible world in which ~d occursand in which ~d explains g

(7) If d and e are contraries and if d explains g then there is no possibleworld in which e occurs and in which e explains g

61


Platorsquos point is this if in one world d occurs and does explaing then although there are many other worlds in which ~d or eoccurs rather than d in none of them does either ~d explain g ore explain g

I have laboured this point a bit because I think that it is relativelyeasy to miss this subjunctive consequence requirement in formulating aclaim like Platorsquos Principle Indeed I think that Hugh Mellor implicitlymisses the point in setting out his argument against a low dependencytheory of explanation15 Thus far we have been trying to formulate aversion of (PP1) but Mellor and the theories he is attacking areconcerned with (PP2) In fact Mellor is arguing in effect for (PP2)Mellor argues by reductio against a low dependency theory ofexplanation in the following way Suppose d did explain g which isimprobable given d (let grsquos low probability be p) But if d did explain git could just as well have explained ~g since ~g will be highly probable(1ndashp) given d and hence has at least as good a claim to be explained byd as g has (Let us take for granted that a theory of explanation whichsaid that we could explain improbable but not probable events would beimplausible) Low dependency theories are committed to the view thatsome explanans d could explain both g and ~g and since that isimpossible Mellor argues that d cannot explain g which is improbablegiven d Mellorrsquos argument here relies on the premiss that no explananscould explain both g and ~g and this is in substance (PP2) But whyshould we accept that no explanans could explain both g and ~g Hisargument for this (with lsquogrsquo and lsquo~grsquo substituted for his lsquoqrsquo and lsquo~qrsquo) isas follows

(a) Explananda must be true(b) No theory of explanation is acceptable if the criteria it proposes for a

successful explanation lsquoare indifferent to the explanandumrsquos truthvaluersquo

(c) An explanans that could relate as well to a false as to a trueexplanandum is no explanation at all

(d) Therefore nothing explains g that lsquowould by the same tokenrsquo explain~g

(Mellor 1976237)

Mellorrsquos argument as set out above relies on a crucial modal ambiguityWhat (b) rules out is d explaining ~g if d explains g since it is not possibleto explain what does not happen any more than it is possible that what

62


does not happen explains something But if Mellorrsquos argument is to cutagainst low dependency theories it must establish something muchstronger modally speaking It must show that if d does explain g d couldnrsquothave explained ~g had ~g occurred instead of g

We can make this point in possible worlds terminology What (b)says is that if d explains g in a world then it is not possible for d toexplain ~g in that same world (because in that world lsquo~grsquo is false)What Mellor needs in order to dismiss low dependency explanation isthat if d explains g in a world then there is no other possible world inwhich ~g rather than g occurs and in which d explains ~g Mellorrsquosargument certainly cannot show this Considerations about the truthof the explananda will surely rule out drsquos explaining g and ~g in thesame world but could not rule out d explaining g in a world in whichg occurs and d explaining ~g in some other world in which ~g occursAfter all in the other world in which ~g occurs lsquo~grsquo will be truerather than false unlike in the first world So low dependency theoriesdo not lsquorelatersquo an explanans to a false explanandum in the sense thatthe explanandum can be false in the world in which it gets explainedby the explanans

In the argument above if (d) is read in an indicative sense it followsfrom the conjunction of (a) (b) and (c) Nothing explains both g and~g in the same world But in the indicative interpretation (d) isconsistent with low dependency theories of explanation If (d) is readin the subjunctive sense that nothing that explains g in one worldexplains ~g in any other world it is inconsistent with low dependencytheories but the premisses Mellor adduces go no distance in showingthat (d) is true So Mellorrsquos argument for (PP2) neglects the indicativesubjunctive mood distinction that we have found crucial in formulatingboth parts of (PP)

(6) and (7) do not provide a plausible formulation of (PP1) for thefollowing reason Even if token event d explains token event g surelythere must be some logically possible world in which ~d explains g andanother in which e drsquos contrary explains g (And this is so even whenthe worlds share all causally relevant circumstances) If Platorsquos Principledenied this it must be wrong If we allow all the logically possible worldsthen some of them will differ from our world namely the one in which dexplains g in respect of their laws In the actual world in which d explainsg there may be a law that Ds cause (and hence let us suppose explain)Gs Even so there is a logically possible world with the law that ~Ds (orEs) cause (and explain) Gs and in that world ~d (or e) explains g If we

63


allow unlimited changes in natural laws there is no difficulty in allowingpossible worlds in which ~d or e explains g even when d explains g in theactual world

In order to get Platorsquos Principles correctly formulated we need toidentify a subset of the logically possible worlds namely those with thesame laws that hold in the actual world We need to select those worldswith the laws of nature fixed as they are in our world and in this way yetanother consideration of laws must enter into a formulation of (PP1)Letrsquos call this subset of the logically possible worlds lsquothe nomos-identicalpossible worldsrsquo (nomologically identical possible worlds) or the n-possible worlds for short So (PP1) should be formulated as the conjunctionof (8) and (9)

(8) If d explains g and if d and ~d are contradictories then there is no n-possible world in which ~d explains g

(9) If d explains g and if d and e are contraries then there is no n-possible world in which e explains g

(8) and (9) capture I claim what is salvageable in (PP1) (PP2)concerns itself with the explanation of contrary and contradictoryexplananda by the same thing Taking our cue from (8) and (9) (PP2)should be formulated as the conjunction of (10) and (11)16

(10) If d explains g and if g and ~g are contradictories then there isno n-possible world in which d explains ~g

(11) If d explains g and if f and g are contraries then there is no n-possible world in which d explains f

It is important to include this implicit reference to laws for anotherreason Couldnrsquot a pair of lsquoexplanationsrsquo by opposites confer highprobability on both g and ~g Suppose in the actual world d explainsg and g has a probability of p given d Suppose further contrary to(10) that there is some other possible world call that world lsquowrsquo inwhich d explains ~g It is true that in the actual world ~grsquos probabilityof occurring (it never of course occurred) given d was 1ndashp Are weentitled to assume that in w ~g will still have a probability of 1ndashpgiven d As we switch possible worlds couldnrsquot the conditionalprobabilities of events change Couldnrsquot it be the case that g givend has the probability p in the actual world but that it is ~g (ratherthan g) given d that has probability p in w the same probability thatg given d has in the actual world

64


In general of course it is true that the conditional probabilitiesassignable to events will change across logically possible worlds Butrecall that we are only interested in a subset of those possible worldsnamely the n-possible worlds These are worlds which have the samelaws If worlds are deterministic they will share deterministic laws ifworlds are nondeterministic they will share stochastic laws In virtue ofstochastic nomos-identity whatever grsquos probability conditional on d is inthe actual world it will have the same probability conditional on d in alln-possible worlds

Consider then two possible worlds with the same laws and suppose doccurs in both In world w the conditional probability of g given d is pIs it true that the conditional probability of ~g given d in any other worldwith the same laws must be 1ndashp There might after all be the followingtwo laws which obtained in both of the nomosidentical worlds (1)whenever a D but not an H the probability of a token event of type G is p(2) Whenever a D and an H the probability of a token event of type ~G isp There could of course be two such laws But this will not provide acounterexample because of the requirement of stability of causallyrelevant circumstances across worlds It is not true that all the causallyrelevant circumstances are the same in both worlds In one world d occursin H-ish circumstances in the other d occurs and there are no H-ishcircumstances

Platorsquos (PP2)

What might be said whether by Plato or more generally in favour of(8)ndash(11) as formulations of Platorsquos two principles Letrsquos take thesecond principle as expressed by (10) and (11) first Mellor in theargument I cited earlier was arguing for a high dependency theory ofexplanation against a low dependency theory in effect he wasarguing for (10)17 It is (10) that has received the most attention in thecontemporary literature

What for example would be involved in the rejection of (10) Supposed does cause and explain g and suppose further that there were some n-possible world (a world which shared all its laws with the actual world) inwhich d causes and explains ~g In one of the worlds suppose that bringingan atom to a certain level of lsquoexcitementrsquo causes and explains its decayand that in another possible world bringing the atom to the same level of

65


lsquoexcitementrsquo causes and explains its failure to decay (and of course holdingall other causally relevant circumstances fixed in the two worlds)

A necessitating or determining cause is let us say sufficient18 in thecircumstances for its effect One thing is clear it is inconsistent with thesupposition above that d is a necessitating or determining full cause of gIf d is the sufficient or determining cause of (and explained) g in oneworld d must be the cause of (and explain) g in all n-possible worlds inwhich it occurs (and in which all other causally relevant circumstancesare the same) This is just what lsquosufficiencyrsquo means In particular d wouldhave to cause g in that n-possible world in which it occurred and was alsothe necessitating or determining full cause of (and explained) ~g

It is impossible that there be a world in which both g and ~g occurTherefore if d causes and explains g in one world and causes and explains~g in some other n-possible world (and all other relevant causalcircumstances are the same in both worlds) such causation cannot bedeterministic and such explanation cannot be accounted for by adeterminative theory of explanation Such causes cannot be sufficient fortheir effects except in the weak and uninteresting sense of materialsufficiency

So if we reject (10) we must accept explanations employingnondeterministic causation and the n-possible worlds will includeprobabilistic or stochastic laws of causation A rejection of (10) willcommit the rejector to a non-determinative theory of explanationIndeed the rejector of (10) is committed to a low dependency theoryfor d causes and explains (in different possible worlds) g and ~g oneof which must have a low objective probability given dContrapositively acceptance of a determinative theory of explanationcommits one to (10) It should come as no surprise that Platorsquosadherence to (10) is coupled with and indeed underpinned by hisdeterminative theory of explanation19

How do matters stand with (11) Two statements are contraries if notboth can be true (although they might both be false) Suppose that dexplained g in the actual world but could explain f in some other n-possible world and that f and g were contraries For example suppose inone world the emission of a certain particle from a source causes andexplains its landing at position p1 on a photographic plate and in anotherworld the emission causes and explains its landing at position p2 on theplate (and of course holding all other causally relevant circumstancesfixed between the two worlds)

66


An application of the previous argument will show that a rejection of(11) also implies rejection of a determinative theory of explanation forotherwise there would be a possible world in which both f and g occur If(11) is surrendered it is also the case that some explaining causes will notbe sufficient for their effects In this way we can see why Plato wouldhave been led to embrace (11) as well as (10) which I have jointly referredto as (PP2)

If we unlike Plato were willing to reject a determinative theory ofexplanation would we then be free to accept that one explanans can explaintwo contrary explananda If f and g are contraries then ~(fampg) Theprobability calculus tells us that in this special case in which p(f+g)=0p(fvg)=p(f)+p(g) Since all probabilities are less than or equal to 1 p(fvg)=1Substituting p(f)+p(g)=1 Therefore it follows that p(f)=1ndashp(g)20

So if the probability of either one of f and g is high the probability ofthe other is low If d explains both it must be able to explain anexplanandum with a low probability A high dependency theorist cannotaccept the explanation of two contrary explananda by one explanans justas such a theorist could not accept the explanation of two contradictoryones by one explanans This is something only a low dependency theoristcan accept Only a low dependency theorist can reject (10) or (11)

Platorsquos (PP1)

What of (8) and (9) which I have jointly referred to as (PP1) (8)and (9) cash out the idea of the unacceptability of oppositeexplanations of the same thing rather than the idea of theunacceptability of the same thing explaining opposites Unlike thesame explanation for opposites (10) and (11) there has been little orno discussion of Platorsquos (PP1) in the contemporary literature

There are two lines of argument that might be tried in order to arguefor (8) and (9) The first focuses on the necessity of a cause for its effectthe second addresses the intuition that such explanations are empiricallyempty Letrsquos take (8) first If (8) is rejected then there could be two n-possible worlds otherwise identical with regard to causally relevantcircumstances and in one of which d occurs and causes (and explains) gin the other ~d occurs and causes and explains g

Just as the rejection of (10) commits the rejector to a form of non-deterministic causation in which a cause is not in the circumstances

67


sufficient for its effect rejection of (8) commits the rejector to causationin which a cause is not in the circumstances necessary21 for its effectAfter all if d was necessary for g in the first world and since the secondworld is just like the first in point of both laws and other causally relevantcircumstances g will not occur in the second world unless d does But thesupposition is that in the second world g occurs in spite of drsquos failing tooccur (~d occurs) So if d is a cause of g in the first world and if ~d is acause of g in the second neither d nor ~d can be a cause which is necessaryin the circumstances for its effect

Can there be causes which are not necessary in the circumstancesfor their effects There is disagreement about this and I have no desireto take sides in the dispute but only to point out how Platorsquos Principleslink up with certain ideas about causation In chapter I I mentionedthat John Mackie insists that a cause be strictly necessary in thecircumstances for its effect although it need not be sufficient22 Anondeterministic cause according to Mackie is not sufficient in thecircumstances for its effect but even a nondeterministic cause isnecessary in the circumstances for its effect David Lewis on the otherhand speaks of chancy causation if the cause had not been then theeffect would have been less likely to occur but might still haveoccurred23 On the Lewis account a cause is not even necessary in thecircumstances for its effect So a rejector of (8) needs causes whichare not necessary in the circumstances for their effects if such therebe those unwilling to accept this (like Mackie) could not consistentlyreject (8)

We can also show that the rejection of (9) commits the rejector tocauses which are not necessary in the circumstances for their effects Thetwo worlds are alike in all relevant respects (laws and circumstances)save this one in one world d occurs and causes and explains g and in thesecond world e drsquos contrary occurs and causes and explains g In thatsecond world since e occurs d cannot occur (they are contraries) So inthe first world d cannot be necessary in the circumstances for g becauseg can occur without d (as g does in the second world) If d causes andexplains g in the first world but e does so in the second neither d nor ecan be a cause which is necessary in the circumstances for its effect

The second argument for (8) concerns the apparent empiricallsquoemptinessrsquo of an explanation for g in terms of d if there could have beenan explanation (in otherwise the same circumstances and with the samelaws) of g in terms of ~d In truth the same intuition sometimes informsarguments in favour of retaining (10) John Watkins for example says

68


this about explanations of lsquooppositersquo results in terms of the sameinformation

if d can lsquoexplainrsquo g given that g turned out to be true then dcould have explained ~g at least as well had ~g turned out to betrue [according to a low dependency theory]hellip Thus d as wellas lsquoexplainingrsquo the occurrence of the event depicted by g couldequally well have lsquoexplainedrsquo its non-occurrence I hold that sucha dual purpose lsquoexplanationrsquo that will serve whichever way thingsgo does not provide a genuine explanation of the way thingsactually went24

Suppose d causes and explains g and contrary to (8) suppose thatthere is an n-possible world in which all causally relevantcircumstances are the same and in which ~d causes and explains g Itwould be tempting to suppose that this supposition means that g wouldbe lsquocausedrsquo and lsquoexplainedrsquo lsquowhichever way things gorsquo and thereforethat the lsquoexplanationrsquo would be empirically empty Within thesepossible worlds which have the same laws and the same fixed causallyrelevant circumstances g will be lsquoexplainedrsquo whatever happens sohow can the explanation of g depend or be contingent upon anything(This argument assumes the Law of the Excluded Middle but we cantake that as uncontroversial for the purposes of this argument)

The above argument from empirical emptiness is more complicatedthan might at first seem to be the case The empirical emptiness argumentpresupposes something that has not yet been made explicit Let lsquosrsquo be thatsubset of logically possible worlds which have the same laws and thesame fixed causally relevant circumstances as does the world in which dcauses and explains g The argument from empirical emptinesspresupposes that occurrences of d and ~d in s have the same explanatoryimpact on g from which it is concluded that the causal or explanatoryimpact of either d or ~d on g in s-worlds must therefore be nil Withinworlds in s it cannot matter to g whether d or ~d

However d and ~d are irrelevant to the explanation of g in the worldsin s only if they both explain g to the same extent or with the same impactin all the worlds in s On a determinative theory this requirement ofsameness of explanatory impact is automatically met If d explains g in aworld in s then grsquos occurrence is necessary in any s-world given d (d issufficient for g) if ~d explains g in a world in s then grsquos occurrence is

69


necessary in any s-world given ~d (~d is sufficient for g) But d and ~dbetween them exhaust the possibilities so grsquos occurrence isunconditionally necessary in any world in s The empirical emptiness ofthe purported explanations relates to the fact that neither d nor ~d seemsreally to make any difference to grsquos occurrence in s Since Plato holds adeterminative theory we can also see why he would have held (8)

We can similarly show that on a low or high dependency theory if dand ~d confer the same likelihood on grsquos occurrence then the lsquoexplanationrsquois empirically empty because d and ~d make no difference to the likelihoodof grsquos occurrence in any s-world

The argument for this last contention is as follows Suppose that it isclaimed that d explains g since given d g is probable to some degree(whether the probability is high or low) and also that ~d explains g sincegiven ~d g is probable to some degree (whether high or low) In thespecial case in which the probability of g given d and the probability ofg given ~d is the same we can reject the supposition that both d and ~dexplain g by means of the following argument If the probability is thesame then d is statistically irrelevant to the probability of g That is

But since the two dependency theories of explanation conjoinexplanatory power with dependency d is explanatorily irrelevant tog as well On the determinative account the supposed explanation ofg by d and ~d collapsed because of the determinative irrelevance of dand ~d to g g was unconditionally necessary in s In the case of ahigh and a low dependency account of explanation in the specialcase in which d and ~d confer the same probability on grsquos occurrencethere is a parallel dependency irrelevance of d and ~d to g in any s-world g has an unconditional probability p in s-worlds (notconditional at any rate on d or on ~d)

But there is no reason why any dependency theorist who wanted toargue for the explanation of g by d and by ~d within s would have toassume that the likelihood or chance of occurring conferred on g by d andby ~d was the same As far as the requirements of the probability calculusgo if d confers probability n on g and ~d confers probability m on g mand n might not be and indeed it would be exceptional if they wereequal Both m and n might have any value between 0 and 1 Both m andn might be high or both might be low or one might be high and the other

70


low There could be two stochastic laws both of which held in the worldsin which d causes and explains g and in the worlds in which ~d causesand explains g if a D-type event then there is a probability n that a G-type event if a D-type event fails to occur (~D) then there is a probabilitym that a G-type event (mn) Intuitively whether d or ~d is not irrelevantto grsquos occurrence

If d explains g and ~d explains g what do d and ~d do to grsquosunconditional or prior probability We can show that if d raises theprobability of g from whatever probability it had then ~d must lower thatprobability The argument runs as follows

(1) p(g)=p(gd) p(d)+p(g~d) p(~d) [follows from the probabilitycalculus the definition of conditional probability and additivity]

(2) Suppose that p(gd)gtp(g) and that p(g~d)=p(g)(3) Then p(g)gtp(g) p(d)+p(g) p(~d) [by substitution in (1)](4) p(g) p(d)+p(g) p(~d)=p(g) (p(d)+p(~d)) [by factoring](5) p(d)+p(~d)=1 [the probability calculus](6) p(g) p(d)+p(g) p(~d)=p(g) [by substitution in (4)](7) p(g)gtp(g) [by substitution in (3)]

Whether we assume a high or a low dependency theory this result isthe same If we assume that both ~d and d raise the probability of gfrom some unconditional or prior probability (or even that one raisesthat probability and the other keeps it the same) we can derive acontradiction So if both d and ~d cause and explain g within s thenif one of them raises grsquos probability the other must lower grsquosprobability

Imagine that grsquos unconditional probability is 097 Suppose d raises itto 098 ~d might only lower the probability of g to 096 so the probabilityof g on both d and ~d might be very high both after the raising and afterthe lowering The requirement that one of the contradictory pair lowerthe probability of the explanandum is consistent with grsquos probability beinghigh or low given d or given ~d However although the rejection of (8) istherefore consistent with both high and low dependency theory assumingthat d and ~d do not confer the same probability on g explaining anexplanandum by contradictories is consistent only with the forms of thesedoctrines which permit an explanation actually to lower the probabilityof what is being explained

Finally what of (9) the idea that contraries cannot explain the sameexplanandum We cannot use the argument from empirical emptiness inthe case of contraries because contraries do not exhaust the possibilities

71


in the way in which contradictories do Suppose we have a machine thatsorts through balls of three different colours balls that are red all overballs that are blue all over and balls that are green all over and eitherunfailingly or with a certain probability rejects balls if and only if theyare either red or green The machine unfailingly or with a certainprobability accepts blue balls

Suppose there is some specific ball b that is red and which the machinerejects If asked to explain why the machine rejected ball b I can replythat it is because b was red But if a contrary state of affairs had obtainedbrsquos being green brsquos being green could have just as well explained brsquosrejection So we could have explained the same explanandum ball b isrejected by a contrary explanans b is green if b had been green brsquosbeing red and brsquos being green are only contraries and do not jointly exhaustthe relevant empirical possibilities which in the light of the machinersquoslaws of working explain the machinersquos behaviour It is this fact that savesthe explanation from emptiness or non-contingency So the rejection of(9) is well-motivated and is consistent with all theories of explanationand we might therefore wonder why Plato subscribed to (9) We couldexplain the same explanandum by means of two contrary explanantiawhatever our theory of explanation might be

There are however cases in which we dislike explanation by contrariesSuppose some psychological theory explains a piece of behaviour by citingthe agentrsquos inferiority complex Suppose further that had the agent had asuperiority complex the theory would have explained the same piece ofbehaviour by citing the superiority complex (I assume that the twoexplanations will have some other differences in what they explain forotherwise they will be empirically equivalent) We intuitively feel thatthis sort of explanation is empty Strictly speaking lsquoAgent a has asuperiority complexrsquo and lsquoAgent a has an inferiority complexrsquo are merelycontraries because both cannot be true but both might be false and wehave already seen that there is nothing intrinsically wrong with explanationby contraries

The two statements are strictly contraries because lsquoAgent a has neither aninferiority nor a superiority complexrsquo is non-empty It is empirically possiblethat both lsquoAgent a has a superiority complexrsquo and lsquoAgent a has an inferioritycomplexrsquo be false Agent a could have had a lobotomy or be dead or be in apermanent catatonic trance But although these are possibilities they are notthe sorts of possibilities up for consideration by the theory in the way inwhich the possibility of the ballrsquos being blue as well as red or green was upfor genuine consideration in the preceding example This is to say I think

72


that we often relativize the possibilities to the theory at hand and what are infact technically contraries often get thought of as if they were contradictoriesas if (~damp~e) was empty for all the practical purposes the theory is intendedto cover For the purposes of the psychological theory being assessed havinga superiority complex and having an inferiority complex between them exhaustthe relevant possibilities

I conjecture then that in those cases in which we are uneasy about thepossibility of explanation of the same explanandum by contraries it isbecause we think of the contraries as if they were contradictories (9) isfalse but there are interesting special cases like the one from psychologyin which a suitably restricted version of (9) would be true In those casesin which we think of the contraries as genuine contraries in which wethink of (~damp~e) as non-empty there is no reason to accept (9)

The Theaetetus

I now turn to some of Platorsquos remarks in the Theaetetus 201dndash210bIn that dialogue Plato is discussing knowledge but given the closerelationship between explanation and knowledge both of which areepistemic or at least quasi-epistemic concepts (or so I shall argue inchapter V) it is hardly surprising that some of the things he saysabout knowledge are or can be construed as being relevant toexplanation On several occasions Socrates speaks of explicabilityand inexplicability as well as knowability and unknowability

Towards the end of that dialogue Theaetetus suggests to Socrates thefollowing analysis of knowledge knowledge is true belief with the additionof an account (logos) (201d) In the contemporary epistemologicalliterature it is sometimes said that Platorsquos suggestion amounts to the claimthat knowledge is justified true belief but it is clear I think that lsquoanaccountrsquo means more than lsquoa justificationrsquo I may have a justified truebelief that p if I learn that p from a reliable authority but Plato wouldnever have allowed this as a case of having an account for p Plato heremeans by lsquoaccountrsquo something more like lsquoan explanationrsquo25 Theaetetusrsquosuggestion then is that knowledge that p is true belief that p plus havingan explanation for prsquos being so Aristotlersquos analysis of scientific knowledgeas we shall see includes a similar requirement

Plato addresses two questions about explanation in these passagesfirst the question of the regress of explanation whether explanation

73


ultimately comes to a stop with something itself inexplicable secondwhat it is to have an explanation for something the additional elementrequired for knowledge of that thing

Plato discusses the regress of explanation question with regard to aspecific theory about the stopping of that regress On that theory givingan account is tantamount to resolving something into its ultimateconstituents Complex entities can be analysed into their simpleconstituents The complexes are explicable by means of that analysis theultimate simples for which no further analysis or decomposition ispossible are therefore inexplicable So explanation stops at the inexplicablesimples

Plato presents the doctrine with a dilemma either the complexes arejust the sums of their elements or they are lsquoemergentrsquo new entities whichare more than their parts and therefore strictly speaking have no parts (orso Plato says) If the former then anyone who knows and can explain thewhole entity necessarily knows the elements lsquothe letters must be neithermore nor less knowable and explicable than syllables since we made outthat all the parts are the same thing as the wholersquo If the latter then ashaving no parts and hence no analysis the syllables are simples and asunknowable and inexplicable as the letters lsquoBut if on the other hand thesyllable is a unity without parts syllable and letter are equally incapableof explanation and unknowablersquo (205dndashe)

Note that on the first horn of the dilemma Plato is tacitly relying onthe following principle

(12) If x is a whole and p1hellippn are all its parts and if x is just p1 hellippn thenx is explicable iff each of p1hellippn is explicable

The idea is that one cannot get an explicable something frominexplicable parts On the first horn of the dilemma the whole is justthe sum of its parts so the whole is explicable if and only if each partis explicable Of course if the whole is just the sum of its parts thenif the whole is explicable the sum of the parts must be That followsfrom the law of the indiscernibility of identicals and is uncontentiousBut Plato is not just arguing this He is arguing that if the whole isexplicable then not only must the sum of the parts be explicable buteach part must be (and conversely)

If this is meant to follow from a general principle that says that awhole has a property iff each of its parts has the property then thisseems to be a clear case of the fallacy of composition (or decomposition)One can get a vanishing Indian tribe from braves and squaws no one of

74


whom is vanishing Of course Plato might argue for (12) not as aninstance of a general principle but on its own in virtue of the meaningof explicability In that case Plato needs to supply the missing specialcase argument

On the second horn of the dilemma that Plato offers the wholes aremore than just the sums of their parts and indeed as he argues thereforestrictly speaking qua emergent entities they have no parts On this hornthey would in fact be new entities unities A consequence of this wouldbe that like the elements they would be neither knowable nor explicablefor there would be no parts through which by analysis they might beexplained or known To repeat lsquoifhellipthe syllable is a unity without partssyllable and letter likewise are equally incapable of explanation andunknowablersquo (205e) The doctrine under examination was that partlesssimples from which explicable and knowable complexes are composedare themselves inexplicable But if the so-called complexes becomepartless emergent entities or lsquounitiesrsquo then they too becomeinexplicable His conclusion then is that lsquowe must not accept thisstatementmdashthat the syllable can be known and explained the lettercannotrsquo (205e) Plato accepts that both the basic constituent elementsor parts and the composite entities formed from them which are studiedby an ordinary science (his examples at 206andashb are music and grammar)are knowable and explicable

Although there is no mention of the Forms in this passage from theTheaetetus no doubt Plato would say that it is only with the Forms thatthe regress of knowability and explicability can come to an end for theForms are self-explanatory the high point in the process of dialectic suchthat there is no vantage point still higher by which to explain or throughwhich to know them One way in which to construe the famous ThirdMan Argument in the Parmenides is that it shows that the Forms must beself-explanatory26 Alcibiades is certainly beautiful and his participationin the Form of Beauty explains why he is beautiful his participation inthat Form compels his beauty The Form of Beauty is beautiful too infact supremely beautiful How can we explain the beauty of the Form ofBeauty If in terms of another beautiful form in virtue of which the Formof Beauty is beautiful we shall clearly be involved in an infinite regressThe Form of Beauty must self-explain its own beauty So Plato does notdeny that there is an end in the chain of explanation Indeed there mustbe What Plato insists upon is that the regress of explicability does notcome to an end at the level of the analysis of an entity in terms of its partsor constituents

75


As for the second question what is it to provide an explanation or givean account Plato canvasses three attempts at answering it none of whichis found to be satisfactory and only two of which are of any interest to usFirst lsquogiving an account of xrsquo might mean lsquoenumerating xrsquos elementsrsquo(207a) Second lsquogiving an account of xrsquo might mean lsquobeing able to namesome mark by which the thing one is asked about differs from everythingelsersquo (208c) rather I suppose like offering a thingrsquos individual essenceThe second suggestion is dismissed as being circular for one not onlywould need to have a notion of some distinguishing mark of x but alsowould need to know that the mark was distinctive of x This shows ifPlato is right not that lsquoexplainingrsquo cannot mean lsquoknowing what isdistinctive about a thingrsquo but rather that no member of such a tight littlefamily of interrelated concepts can be illuminating in analysing any othermember

The first suggestion is dismissed as being insufficient for explanation(or knowledge) Plato argues that I might know all the elements thatcompose x and still lack an explanation of x Platorsquos argument is that Icannot be said to know the syllable lsquoThersquo if when I write lsquoTheaetetusrsquo Iwrite lsquoThersquo but when I write lsquoTheodorusrsquo I write lsquoTersquo If I am in thissituation then concerning the lsquoThersquo in lsquoTheaetetusrsquo I can give an accountie the letters that compose the syllable in addition to having true beliefbut I do not yet know lsquoThersquo

I find Platorsquos argument against this first suggestion difficult toreconstruct in detail In a more general way though the interest of thesepassages at the end of the Theaetetus is that they further reflect the Platonicdisinclination to take physical explanation or (what Aristotle would call)material explanation seriously However the argument is to bereconstructed it is clear that its conclusion is that giving an account of athing canrsquot be the same as enumerating the thingrsquos elements In both theTheaetetus and in the Phaedo physical explanations or explanations interms of a thingrsquos matter have been canvassed by Plato and found lacking

Summary

What have we learned about explanation from the discussion of thischapter It is true that in comparison with the remaining historicalchapters there is little in this chapter that I shall want to carry forwardas a substantial contribution to the theory of explanation that I advancein chapters V VI and VII The main point of the chapter seems to be

76


of great interest none the less Plato was opposed to lsquophysicalexplanationsrsquo for the reason among others that they licensed theexplanation of opposites by the same thing When unpacked into myterminology what Plato is opposing is a low dependency theory ofexplanation However his own theory is a determinative theory ofexplanation which would disallow both high and low dependencyexplanations In chapter VII I develop a theory of explanation whichunlike Platorsquos is compatible with any of the three theoriesdeterminative high and low dependency theories of explanation

77

CHAPTER III

Aristotle on Explanation

Does Aristotle have a general account of explanation Richard Sorabjidenies that he has lsquoOf course it would have been satisfying if he[Aristotle] had been able to give a perfectly general account of whatexplanation is Since he does nothelliprsquo1 On the other hand JuliusMoravcsik takes Aristotle to be offering just such a general theorylsquoThe claim that Aristotlersquos theory of aitiai is a general theory aboutexplanation is further strengthened strengthenedhelliprsquo2 Who is rightPerhaps a proper answer to this question depends on what one countsas a lsquoperfectly general accountrsquo and two of the questions I shall wantto address in discussing Aristotle are what is a general account ofexplanation and how is one to justify one account over anotherThis last question returns us to a theme begun in chapter I IsAristotlersquos method for justifying an explication of explanation basedon language use or is it a lsquotechnicalrsquo account in the sense that I gaveto those expressions in chapter I If the latter what considerationswould Aristotle offer to justify his account

The doctrine of the four causes

If anything in Aristotle could count as a general account ofexplanation it would be his discussion of the four causes (aitiai) ofthings Almost every philosophy undergraduate knows that Aristotleheld that there were four causes of or explanations for things the

78


matter form goal or end and motion-originator of a thing I followJulius Moravcsik Max Hocutt Julia Annas and many othercontemporary scholars and take this doctrine to be about explanationrather than simply about causation as we understand that latter idea3

I refer to the traditional doctrine as lsquothe doctrine of the four causesrsquobut I sometimes reform quotations from older translations that referto the aitiai of a thing by changing lsquocausersquo to lsquoexplanationrsquo Thedoctrine of the four causes is about four explanatory principles Themodern conception of efficient causation is closest to but by no meansidentical with Aristotlersquos idea of the motion-originator which is onlyone of his four lsquocausesrsquo or explanatory principles

Is it possible to cite an efficient cause or indeed any other of the fouritems but in an unexplanatory way Aristotle develops a terminology inwhich one can do just this Later in the discussion I shall also need a termfor an Aristotelian cause in a non-explanatory sense (what Aristotle callslsquothe incidental causersquo) When there is any possibility of confusion I referto these Aristotelian incidental non-explanatory causes as lsquocausesrsquosimpliciter and the causes that do explain as lsquoexplanatory causesrsquo It mustbe remembered that these causes both explanatory and non-explanatoryinclude not only causes in (what is close to) our modern sense but alsomatter form and end or goal

The doctrine of the four causes is set out in at least two differentplaces which I quote below First there is the following long passagein Physics II chapter 3 (compare also Physics II chapter 7) which isrepeated almost verbatim in Metaphysics V chapter 2 1013a25ndash1014a254

Knowledge is the object of our inquiry and men do not think that theyknow a thing till they have grasped the lsquowhyrsquo of ithellip In one sensethen (1) that out of which a thing comes to be and which persists iscalled lsquoexplanationrsquohellip In another sense (2) the form or thearchetypehellipand its genera are called lsquoexplanationsrsquohellip Again (3) theprimary source of the change or coming to resthellip Again (4) in thesense of end or lsquothat for the sake of whichrsquo a thing is donehellip

This then perhaps exhausts the number of ways in which the termlsquoexplanationrsquo is usedhellipAs the word has several senses it follows that there are severalexplanations of the same thinghellip Further the same thing is theexplanation of contrary results For that which by its presence bringsabout one result is sometimes blamed for bringing about the contrary

79


by its absence Thus we ascribe the wreck of a ship to the absence ofa pilot whose presence was the cause of its safety

(Physics II chapter 3)

Aristotle holds that one does not have the fullest type of knowledgeabout a thing unless one possesses an explanation for it In this hefollows Plato who as we saw held that knowing something involvesbeing able to give an account of it I return to Aristotlersquos analysis ofknowledge later in the chapter

A second passage in which the doctrine of the four causes is set out isin the Posterior Analytics Book B chapter 11 Aristotle again refers tohis doctrine of the four causes and asserts that lsquothere are four explanationsrsquoFormal final and change-initiator explanations are clearly mentionedbut where we expect material explanation Aristotle speaks instead of lsquoifcertain things hold it is necessary that this doesrsquo Aristotle says elsewherethat the premisses of a deduction are the matter or material explanation ofits conclusion Jonathan Barnes argues convincingly in my view andpace Ross that in fact this sort of explanation is material explanationlsquounder a non-canonical descriptionrsquo5

As the opening sentence of the first quotation above makes clearAristotle is thinking of explanation as whatever appropriately occurs inresponse to a why-question He repeats this in Physics II chapter 7 thereare as many kinds of explanation as there are things lsquocomprehended underthe question ldquowhyrdquorsquo Aristotle appears to put considerable weight on thisgrammatical point How seriously should we take this In my view notvery seriously

In the contemporary literature there are other sorts of questions withwhich why-questions are contrasted eg what something is or howsomething was done What kind of explanation is it if I explain what iswrong with my car or how the fight started Surely there might be nodifference between asking what is wrong with my car and asking why mycar wonrsquot operate normally or between asking how the fight started andasking why the fight started On the other hand some what-questions andhow-questions are not interchangeable in this way with why-questionsNeither lsquoWhat are the rules of chessrsquo nor lsquoHow does one greet theQueenrsquo are convertible to lsquoWhy [anything]rsquo

Could we say that an explanation is an answer to a why-question or ananswer to a question which can without loss of sense be transformedinto a why-question But even this wonrsquot do First some why-questionscan be understood as requests for justification or defence rather than

80


explanation why did the framers of the American constitution insist on asystem of checks and balances6 An appropriate answer might be ajustification in terms of the arguments the framers might have given forsuch a system whether or not it was those arguments which actually movedthem to include such a system in the constitution Not all answers to why-questions are explanations We require a prior sense of what an explanatoryrequest is in order to distinguish between why-questions which arerequests for justification and those which are requests for explanation

Second Aristotle understands the idea of a why-question in such awide sense that one already needs a concept of explanation to see whichquestions are why-questions That Aristotle thought of why-questions ina much wider sense than we do is confirmed by looking at Physics IIchapter 7 198a15ndash20 Aristotle says that regarding things which do notinvolve motion a why-question relates to the what of a thing eg to thedefinition of a straight line or of commensurability But can the lsquowhat ofa thingrsquo be converted into a question with the form lsquoWhyhelliprsquo

Aristotle does insist that an answer to a why-question can be given interms of a thingrsquos essence and in the case of things which do not involvemotion this will be the only sort of appropriate answer to a why-questionIt may be that Aristotle was thinking that the question lsquoWhat is a straightlinersquo was transformable into the why-question lsquoWhy are these lines straightlinesrsquo Whatever he was thinking I do not think that one can pick outwhich questions he thought were why-questions in the wide sense heintended unless one already had in mind his doctrine of the four explanatoryprinciples Aristotlersquos remarks on why-questions should be taken merely asa heuristic device for picking out the area in which he is interested not asan adequate philosophical criterion offering necessary and sufficientconditions for when an answer to a question is an explanation

Aristotle also asserts that since the word lsquoexplanationrsquo can be used inmore than one way there can be more than one explanation of the samething Some translations have Aristotle speaking of ways in whichlsquoexplanationrsquo can be used others (like the one I quoted above) have himspeaking of lsquosenses of explanationrsquo In his discussion of the Physics passageWieland says that according to Aristotle lsquoCause has several meanings inordinary usagehellip Strictly speaking therefore we are dealing here not withfour causes but with the four senses in which we speak of causesrsquo7

Wieland is simply wrong about this Nothing in the text would justify usin attributing to Aristotle a full-blown semantic point to the effect thatlsquoexplanationrsquo is equivocal and can bear four distinct senses Aristotle isoffering a classification of different kinds (the Greek word here is topoi) of

81


explanation and it does not follow that aitiai bears four different sensesOne can classify animals into various species and genera it does not followfrom this that lsquoanimalrsquo is ambiguous as applied to humans ducks and slugsI know of nothing in the text to suggest that Aristotle is doing anythingmore than offering such a classification (Wieland also notes that Aristotleis offering a classification of explanations but does not distinguish withany care between that and holding that the word lsquoexplanationrsquo is equivocal)I therefore stick to the less semantic-involving formulations thatlsquoexplanationrsquo can be used in four ways or with regard to four differentelements of things or that there are four different types of explanation

What are these four different types of explanation Sometimes referringto the material from which a thing is made sometimes referring to itsform sometimes to that which initiates change and sometimes to that forthe sake of which it acts will be an appropriate reply to a why-questionIndividuals capable of coming into being (ie generable primarysubstances) have explanations in all four senses

When one inquires into the cause of something one shouldsince lsquocausesrsquo are spoken of in several senses state all thepossible causes eg what is the material cause of man Shallwe say lsquothe menstrual fluid What is the moving cause Shallwe say lsquothe seedrsquo The formal cause His essence The finalcause His end

(Metaphysics VIII 4 1044a33ndash7)

Aristotle is not committed though to the view that things of everytype or category or even every particular of the type substance hasan explanation in each of the four senses Both things that are notsubstances (like the examples of straight line and commensurabilitymentioned above) and substances which are eternal may fail to havea material explanation

Regarding the substances which are natural and generable if thecauses are really these and of this number and we have to learnthe causes we must inquire thus if we are to inquire rightly Butin the case of natural but eternal substances another account mustbe given For perhaps some have no matterhellip Nor does matterbelong to those things which exist by nature but are not substances

82


their substratum is the substance Eg what is the cause of eclipseWhat is its matter There is none the moon is that which sufferseclipse What is the moving cause which extinguished the lightThe earth The final cause perhaps does not exist

(Metaphysics 1044b3ndash14)

Also at Metaphysics Beta 995b 22ndash35

For how can a principle of change or the nature of the good existfor unchangeable thingshellip So in the case of unchangeable thingsthis principle could not exist nor could there be a good-itselfThis is why in mathematics nothing is proved by means of thiskind of cause nor is there any demonstration of this kindmdashlsquobecause it is better or worsehelliprsquo

As the passage from the Physics (and its repetition in Metaphysics) quotedearly in this section makes clear Aristotle obviously believed that animportant consequence of the doctrine of four causes was the rejection ofPlatorsquos view that we discussed in the previous chapter namely that thesame thing (the pilot) cannot explain contraries Or so I interpret the passageAristotle asserts that the same thing can explain contrary results Aristotleseems anxious to make the point it does not follow naturally from thedoctrine of the four different types of explanation which immediatelyprecedes it in the text I conjecture that Aristotle has Plato in mind in thediscussion To repeat the relevant portion of the Physics passage

As the word has several senses it follows that there are severalexplanations of the same thinghellip Further the same thing is theexplanation of contrary results For that which by its presencebrings about one result is sometimes blamed for bringing aboutthe contrary by its absence Thus we ascribe the wreck of a shipto the absence of a pilot whose presence was the cause of its safety


His view is that even if the pilot explains why the ship reached portsafely the pilot could also have been the explanation of why the shipdid not reach port safely if the ship had failed to reach port

83


Of course Aristotle sees that the explanation is not just lsquothe pilotrsquo butrather in one case it will be the pilotrsquos steering the ship in the other case thefailure of the pilot to steer the ship Aristotle often speaks of a thing or substanceas efficient cause or motion-originator Polyclitus as the cause of the statuethe doctor of health sperm of the man These are for him formulations ofpotential causes only He distinguishes between actual and potential motion-originators (Physics II chapter 3 195b15ndash25) A full formulation makes clearthat an explanation is in terms of the actual cause the substance-as-doing-something this healing person and that housebuilding man8

To what extent is Aristotlersquos criticism of Plato fair if this is whatAristotle had in mind Not very fair at all Aristotle does not himselfreally think that the pilot could be the explanatory cause of both shipwreckand ship safety because he does not think that the pilot could be the fullactual cause of either The pilot on his own is only a potential cause

As I remarked in chapter II Plato does not really address himself tothe question of the ontology of explanation Certainly sometimes Platospeaks at least as far as the Forms are concerned as if they tout courtexplain But it would not take a great leap of imagination to see that whatis really explanatory for Plato is some event or process or whatever theparticular participating in the Form or the Form compelling the particularParaphrasing Aristotle Plato could say the Forms are only a potentialcause it is only the Form-as-doing-something (or the particular-as-participating-in-the-Form) which is an actual cause

This is certainly his general strategy In The Republic in the case in whichit seemed as if it was a single substance the soul that explained contraries onreflection he held that there were further distinctions to be made The soul atone time and in one respect and in one part might explain something but thesoul at another time or in another respect or in another part could explain itscontrary So Plato could easily say that the pilot in respect of his steering theship explains its safe arrival the pilot in respect of his failing to steer the shipcould explain its being wrecked if it had been wrecked

Does Aristotle have a general account of explanation

Letrsquos return then to the question with which we began Does Aristotlehave a general theory of explanation We might put Aristotlersquos view inthis way

(E) Something can be explained only by either its matter or itsform or its end or its change-initiator

84


Is (E) at least the kernel of a general theory I said that this questionis important There is a way to trivialize the answer so that it has noimportance and I want now to rule out this trivialization What countsas one theory or analysis of something Suppose we have an analysisof p by r and of q by s If we allow unrestricted disjunction can wenot say that we have a single theory or analysis of pvq namely byrvs I rule out artificial analyses formed by the ad hoc technique ofdisjunction although I do not say that there could be no genuinenon-artificial case of a disjunctive analysis

Now (E) is a disjunction of conditions for explanation and as I havejust argued the existence of a disjunctive set of conditions might not byitself count as a single general theory But is (E) just an ad hoc disjunctionformed from unrelated disjuncts Aristotle appeared to have some reasonfor thinking that these four modes of explanation were exhaustive of thesorts of explanation there are lsquoIt is evident then even from what wehave said before that all men seek the causes named in the Physics andthat we cannot name any beyond thesersquo (Metaphysics I 10 993a12ndash15)If any theory of explanation is not to be ad hoc it must be based onsomething But on what In chapter I I distinguished two broadapproaches the language usersrsquo approach and the technical approach

Does Aristotle take the language usersrsquo approach and base his theoryon how the term lsquoexplanationrsquo or lsquoscientific explanationrsquo was used eitherby everyone in classical Greece or by classical Greek scientists or perhapseven just by classical Greek philosophers As we saw earlier this was theview of Wieland against which I have already argued lsquoThe doctrine ofthe four causes does not consist of a recondite theory of fundamentalmetaphysical principleshellipbut of something m simpler Herehellipwe are infact confronted with the results of an analysis of linguistic usagersquo (Wieland1975147) It is also the view of Peter Achinstein Peter Achinstein assertsthat Aristotlersquos view that explanation must be in terms of either formend matter or change-initiator was based on semantic considerationsabout the meaning of the term9 As far as I can see Achinsteinrsquos view likeWielandrsquos rests on a rendering of the Greek text at Metaphysics 1013a25as lsquomeans semanticallyrsquo for which as I have already said there is nowarrant

In support of the Wieland-Achinstein view one might point toAristotlersquos remarks under the topic of final cause in Physics II chapter 3where he seems to justify the category by an appeal to ordinary usagelsquoWhy is he walking about we say ldquoto be healthyrdquo and having said thatwe think we have assigned the causersquo Although it is true that Aristotle

85


refers to what we commonly say in his remarks on final causation thestyle of his discussion of the four causes seems overall to be noticeablydifferent from his discussions of many other topics where he demandsthat his exposition lsquosave the appearancesrsquo by being in line with commonopinion on the subject10 This is a line of argument that he appears toavoid with explanation

The alternative would be to interpret Aristotle as subscribing to thetechnical approach and as introducing a special concept of explanationwhatever linguistic usage may or may not be like If so we still have toconfront the question of how Aristotle would justify his analysis overother possible competing explications of explanation According to JuliusMoravcsik the Aristotelian theory of explanation is ultimately groundedon and to be justified in terms of metaphysics11 I agree with him

To see how this works Moravcsik offers the following Aristoteliandefinition of substance a set of elements with a fixed structure that movesitself towards self-determined goals The four factors in this definitionare element structure motion originator and goal These correspond toand justify the four types of explanation Since everything else that canbe said to be is an aspect of substance the four types of explanation areboth non-arbitrary and exhaustive If Moravcsik is right Aristotle doeshave (at least the kernel) of a general theory of explanation grounded onmetaphysical considerations and (E) tells us what that kernel of a theoryis Aristotlersquos technical approach introduces a very special and distinctiveidea of explanation and Aristotlersquos metaphysics provides the justificationfor so doing It may be that this concept of explanation was to a greateror lesser degree reflected in ordinary or specialist Greek speech butwhether or not it was is irrelevant Its defence is metaphysical notlinguistic

In some ways the contrast between metaphysics and linguistic usageas an anchor for a view of explanation is not well-expressed as it standsSuppose it turned out that what lay behind linguistic usage was itself acertain metaphysical view of things (indeed how could this fail to be thecase although it might of course be several incompatible metaphysicalviews which informed that usage) I take it that Wieland and Achinsteinare arguing that linguistic usage is where Aristotle stops that there are nofurther metaphysical principles which one could uncover that lie behindthat usage for otherwise the contrast explicitly drawn by Wieland wouldbe pointless

On the Wieland-Achinstein view Aristotle really does not have whatmight be called a general theory of explanation at all There is just this

86


four-fold linguistic usage with no more general principles by which tounify or explicate the classification The disjunction in (E) would be simplythat a disjunction of the ways in which the term is actually used (byeveryone or by some) On this view since based on linguistic usage (E)would not be exactly ad hoc but it is no more than a disjunction of fourseparate elements Wieland attempts to block this unacceptableconsequence of his view lsquothe formal unity of these distinct meanings isestablished through a functional element namely through the questionldquoWhyrdquorsquo Is Wielandrsquos attempt successful

I have already expressed my reservations about the usefulness of thewhy-question approach but in any case the approach offers no real solutionto the ultimately ad hoc disjunctiveness of Aristotlersquos concept ofexplanation on Wielandrsquos account If lsquoexplanationrsquo were four-wayambiguous for Aristotle and if all explanations were answers to why-questions then lsquowhyrsquo ought to be four-way ambiguous as well To asklsquoWhyrsquo would not be to ask a question that was unequivocal Supposethat Aristotlersquos four senses of lsquoexplanationrsquo not only differed in meaningbut also had no part of their meaning in common Then the unity providedby lsquowhyrsquo could only be an uninteresting kind of unity lsquowhyrsquo would besyntactically a single word which also bore four different non-overlappingsenses12

Aristotle at least as interpreted by Moravcsik as rejecting this languageusersrsquo approach has supplied an answer to the question of how we mightground or justify a technical approach to the explication of explanationAlthough explanation being an epistemic concept must suit our epistemicneeds and capacities it can do so only by fitting what we think the worlditself is like13

It may be that the concept of explanation that we actually use isoutmoded it has evolved over a long period of time and it may reflecterroneous or even incompatible beliefs about reality It may no longerfit what we currently think the world is like It may be so outdated thatconceptual tidying-up is no longer sufficient If so concept replacementis the order of the day If possible a concept of explanation should beadopted that fits what we think the world is like14 How we conceiveof what the world is like what its constituents are and how it workswill justify (at least in large measure) choice of concept of explanationWe have seen how Aristotlersquos account of explanation fits hismetaphysics What I shall want to explore in chapter VII is what ourconception of explanation should be like given what we know orbelieve about reality

87


I do not necessarily presume that the metaphysics relevant to an accountof explanation will be composed only of a priori or metaphysicallynecessary truths There is room in a conception of metaphysics for themost general and abstract truths of contemporary science and these maybe a posteriori and contingent Aristotle may have thought of hismetaphysics as (in some sense other than logically) necessary but certainlyit was for him a posteriori I think that Aristotle would have beensympathetic to the methodology of Wesley Salmon

what constitutes adequate explanation depends crucially upon themechanisms that operate in our world In all of this there ishellipno logicalnecessity whateverI have not been trying to lay down conditions that must be satisfied byall admissible scientific explanations in all possible worldshellip My aimhas been to articulate contingent features of scientific explanations inthis world as we presently conceive it15

The nature of metaphysics is not an issue that need detain us hereclearly what I have to say is compatible with any view concerningthe epistemic and logical status of metaphysics the reader might wishto adopt

Incidental and per se causes

Polyclitus is a (potential) cause (the efficient cause or motion-originator) of the statue However Aristotle distinguishes between asubstance as described in an explanatory way (Polyclitus quasculptor) and as not so described (Polyclitus qua pale man) Aristotlecalls Polyclitus qua sculptor the per se cause of the statue Polyclitusqua pale man the incidental cause of the statue For Aristotle per secauses are explanatory causes incidental causes are non-explanatorycauses lsquoAgain we may use a complex expressionhellipand say egneither ldquoPolyclitusrdquo nor ldquosculptorrdquo but ldquoPolyclitus sculptorrdquorsquo(Physics II chapter 3 195b5ndash12) The same must be true for thematerial from which something is made and the end or goal for whichsomething strives It is not just the material that explains but thematerial as described in one way rather than another It is not just thething that is in fact the goal that explains but the thing described inthe terms under which it is desirable to the agent

88


Aristotle says that the same is true of the thing explained lsquoSimilardistinctions can be made in the things of which the causes are causeseg of this statue or of statuehelliprsquo (Physics II chapter 3 195b7ndash9) Whatgets explained by Polyclitus qua sculptor is the statue qua statue and notqua a bronze object Under the latter conceptualization it presumably isexplained by Polyclitus qua brazier To use modern terminologyexplananda as well as explanantia are only such when conceptualized inan appropriate way

Aristotle gives us a definition of the incidental or accidental inMetaphysics16

lsquoAccidentrsquo means (1) that which attaches to something and can be trulyasserted but neither of necessity nor usuallyhellipfor neither does the onecome of necessity from the other or after the other norhellipusuallyhellipAnd a musical man might be pale but since this does not happen ofnecessity nor usually we call it an accident Therefore since there areattributes and they attach to subjects and some of them attach only ina particular place and at a particular time whatever attaches to a subjectbut not because it was this subject or the time this time or the placethis place will be an accident

(Metaphysics V 30 20ndash5)

Aristotle is distinguishing two senses of aitiai only one of which canproperly be tied to explanation It is true that nothing in the textcommits Aristotle to the view that this distinction provides twodifferent senses of lsquocausersquo there is no more textual evidence to ascribea semantic point to him here than there was in his discussion of thefour-fold typology of causes But I think that if what Aristotle sayshere is true then there are two different senses of lsquocausersquo

In the first lsquoaccidentalrsquo or lsquoincidentalrsquo sense the concept of causationis not logically tied to the concept of explanation Aitiai do not necessarilyexplain that for which they are the aitiai In the second per se senseexplanation and causation are tied and therefore a cause in this sensenecessarily explains what it causes That difference amounts to a differencein the meaning or sense of the two usages of the term aitiai

Suppose the sculptor was a pale man In the incidental and non-explanatory sense it is true that the pale man caused the statue to comeinto existence In the explanatory and per se sense (or as some translationshave it the lsquoin virtue of itselfrsquo sense) it is true that the sculptor caused thestatue to come into existence In this latter sense of aitiai although the

89


sculptor caused the statue to come into being and even if the sculptorwas a pale man it does not follow that the pale man caused the statue tocome into being This per se sense qualifies as non-extensional in at leastone meaning of that term because substitution of singular terms salvaveritate fails for that sense Things (causally) explain and are explainedonly as conceptualized or described in an appropriate way

Aristotle then is marking what we would call an extensional and anon-extensional sense of aitiai only the latter of which is explanatory AsI indicated at the beginning of this chapter where there is any chance ofconfusion I use lsquocausally explainsrsquo or lsquoexplanatory causersquo for Aristotlersquosper se sense of aitiai and simply lsquocausesrsquo or lsquocausersquo for the incidental oraccidental sense

In the passage quoted above Aristotle also tells us how we are todistinguish between accidental and per se descriptions of the cause Thecriterion for a description of a cause being a per se description rests onthe existence of suitable laws Suppose we have an assertion with theform the F caused the G We want to know whether the F asconceptualized is an incidental or a per se cause of the G asconceptualized The text quoted above began lsquoldquoAccidentrdquo means (1) thatwhich attaches to something and can be truly asserted but neither ofnecessity nor usuallyhelliprsquo Aristotle in effect is telling us

(A) The F is the per se or explanatory cause of the G iff the F causes the Gand lsquoFrsquo and lsquoGrsquo occur as (at least part of) the antecedent and the consequentrespectively in the statement of a deterministic or a stochastic law

The deterministic law covers the case in which things like that happennecessarily or anyway always the stochastic law the case in whichthings like that happen usually or for the most part In the simplestbut unlikely case the law will be Fs cause Gs The final clause lsquoatleast part ofrsquo is meant to cover the more complicated case in whichmore descriptive content must be added to lsquoFrsquo or lsquoGrsquo or both to obtaina true universal law For Aristotle lsquothe F causally explains the Grsquo canbe true only if the F and the G are linked by a law under the samedescriptions that occur in the explanatory singular assertion namelylsquothe Frsquo and lsquothe Grsquo

Is there an explanatory cause for everything that happens Is there anincidental cause for everything that happens Aristotlersquos answers to thesetwo questions are complicated In a related set of discussions at PhysicsII chapters 4ndash6 Metaphysics V 30 VI 2ndash3 and XI 8 Aristotle asks

90


whether chance is a cause The answer is tied in with the discussion ofaccidental or incidental causes One example of his is this Consider thosethings which do not happen always in the same way or even for the mostpart in the same way Some of these things are the results of choice anddeliberation and some are not (like the musical man being pale) but letus restrict our discussion to examples of the former sort Notice thatAristotle in this passage need not be taken as asserting that no outcomescan be linked to choices by always-or-for-the-most-part laws but onlythat some choices and outcomes cannot be joined by such laws

Suppose a man who is busy collecting subscriptions for a feast goesto the market to buy food While there by chance he stumbles upon aman from whom he collects money for the feast Had he known that theman was there he would have gone to the market and collected the moneybut that wasnrsquot why he went there He went there to buy food

Does the collectorrsquos meeting the subscriber have a cause Aristotlethinks it does and that the cause is chance lsquoThings of this kind thenwhen they come to pass incidentally are said to be ldquoby chancerdquorsquo (PhysicsII chapter 5 196b24) In which of the two senses that we have indicateddid Aristotle think that chance was a cause lsquoChance is an incidentalcausehelliprsquo (197a5) He adds lsquostrictly it is not the causemdashwithoutqualificationmdashof anything for instance a housebuilder is the cause of ahouse incidentally a flute-player may be sorsquo (10ndash15) That is in theincidental and non-explanatory sense chance caused him to meet thesubscriber and collect the money

Is there any explanatory cause of his collecting the money even ifchance was not it There can be no explanatory cause of his collectingthe money because there is no universal or for-the-most-part law thatlinks wishing to buy food and collecting of subscriptions Accidents donot have explanatory causes in the explanatory sense there was nocause of his collecting the money Nothing causally explains hiscollecting the money lsquoEvidently there are not causeshellipof the accidentalof the same kind as there are of the essentialhelliprsquo (Metaphysics XI 71065a7ndash9)

Letrsquos return again to the incidental cause Which is the incidental causeof collecting the money chance or the collectorrsquos wanting to buy foodAristotle seems clear that the correct reply is chance He seems loth tocount the collectorrsquos desire to buy food as the incidental cause sincelsquothere is no definite cause of an accident only a chance cause ie anindefinite onersquo (Metaphysics V 30 1025a23) In the case of the collectorof subscriptions Aristotle argues

91


And the causes of the manrsquos coming and getting the money (when hedid not come for the sake of that) are innumerable He may have wishedto see somebody or been following somebody or may have gone tosee a spectaclehellip Hence to conclude since causes of this kind areindefinite chance too is indefinite

(Physics II chapter 5 197a15ndash20)

In a similar case in which a sailor finds himself in Aegina because hewas carried away by a storm (Metaphysics V 30) Aristotle assertsthat since the storm was an accident there was only an indefinitechance cause of the man getting to Aegina and also that the storm adefinite event if ever there was one was the cause of his lsquocoming to aplace for which he was not sailingrsquo Aristotle makes a clear distinctionbetween two different descriptions of the same occurrence lsquocomingto a place for which one was not sailingrsquo and lsquocoming to Aeginarsquo

If we describe the effect incidentally as a coming to Aegina (comparecollecting the subscription) the effect has only an incidental and indefinitecause namely chance After all there is no law that joins storms andcoming to Aegina (or wanting food and collecting subscriptions) Theeffect qua a coming to Aegina has no per se cause and therefore nothingcausally explains the sailorrsquos coming to Aegina as so described

On the other hand if we describe the same effect per se as a coming toa place for which one was not sailing then there is a definite per se causeof it viz the storm The storm is the per se cause of coming to a place forwhich one was not sailing and hence causally explains coming to a placefor which one was not sailing since there is presumably a law to theeffect that storms often or for the most part cause sailors to arrive at placesother than that for which they were sailing

Note the further evidence here for the non-extensionality of the per seexplanatory causal context On Aristotlersquos view even though the stormwas the per se cause of and therefore causally explains his getting to aplace for which he was not sailing and his coming to a place for which hewas not sailing=his coming to Aegina it does not follow that the storm isper se cause of and hence causally explains his coming to Aegina

For Aristotle then (a) lsquochance is the incidental and indefinite cause ofthe sailorrsquos coming to Aeginarsquo and (b) lsquogetting blown off course by thestorm is the per se cause of the sailorrsquos coming to a place for which hewas not sailingrsquo are both acceptable assertions (compare lsquochance is theincidental cause of his collecting the subscriptionrsquo and lsquothe wish to buyfood is the per se cause of his buying foodrsquo)

92


What is not acceptable is (c) lsquogetting blown off course by the storm isthe cause of his coming to Aeginarsquo (or (d) lsquothe wish to buy food was thecause of his collecting the subscriptionrsquo) in either sense of lsquocausersquo (c)and (d) cannot be true in the per se sense because of the close connectionthat Aristotle draws between per se cause explanation and law Nor can(c) and (d) be true in the incidental sense since in that sense only chanceis a cause

Letrsquos accept for the sake of argument that (c) and (d) cannot be truein the per se sense But why canrsquot they be true in the incidental sense oflsquocausersquo Why does Aristotle insist that only indefinite chance can be acause in the incidental sense Aristotlersquos argument quoted above isdreadful



From the fact that any one of a large and indefinite number of causesmight have led to his coming and getting the money it does not followthat there was anything indefinite about the cause that actually didoperate on this occasion If he came to buy food there seems to bedespite what Aristotle claims a definite incidental cause of his gettingthe money namely his desire to buy food

Aristotle is certainly ready to countenance definite incidental causeswhen he is not discussing specifically the nature of the accidental

Another mode of causation is the incidental and its genera eg in oneway lsquoPolyclitusrsquo in another lsquosculptorrsquo is the cause of a statue becauselsquobeing Polyclitusrsquo and lsquosculptorrsquo are incidentally conjoined Also theclasses in which the incidental attribute is includedhellip

(Physics II chapter 3 195a35ndash195b3)

At Metaphysics 198a5ndash7 he says that lsquospontaneity and chance arecauses of effects which though they might result from intelligenceor nature have in fact been caused by something incidentallyrsquo whichappears to say that chance qua indefinite incidental causepresupposes or supervenes upon some definite incidental cause Why

93


his insistence at least in some passages that accidents have no definiteincidental cause

Necessitation and laws in explanation

Aristotle treats incidental causes in the case of accidents differentlybecause of their implications for necessitation17 He is keen to avoidthe view that all things happen by necessity which he regards asobviously false (see Metaphysics VI 3) lsquohellipall all things will be ofnecessity if there has to be a cause non-accidentally of what goesthrough a process of beginning or ceasinghelliprsquo (Sorabjirsquos translationof Metaphysics VI 3 1027a30) Definite causes necessitate theireffects and explain them Aristotle lsquoconcedes that an effect isnecessary given its [definitemdashmy addition DHR] causersquo18

Explanatory causes necessitate their effects On this basis it is fair toascribe to Aristotle a determinative theory of explanation (although Ishall indicate some contrary evidence below)

On the other hand a mere indefinite cause like chance neither necessitatesnor explains its effect since lsquothe cause of the [accidental] is indefinitersquo(Metaphysics VI 3 1028a) the accidental is not necessary and the chainof necessitation is broken In order to introduce contingency into hismetaphysics Aristotle introduced accidents which lack any definite causeIf there is no definite cause of an accident then there is no possibility of itsbeing necessitated What follows the accident may then be necessary giventhe accident but the non-necessitated occurrence of the accident hasintroduced a contingency in the subsequent necessary unfolding of events

Given the close connection between per se or explanatory causationand law one might wonder whether Aristotle is committed in this generalexposition of explanation (which is to be distinguished from his view ofscientific explanation in the Posterior Analytics) to an account ofexplanation which requires the presence of a law in every full explanationI think that he is not so committed The concept of law figures here onlyas a criterion for distinguishing between per se and incidental descriptionsof causes As far as this account goes lsquothe F caused the Grsquo might be a fullexplanation of why the G assuming that the conceptualizations thereinare per se and without the explicit addition of any laws to the explanationitself The existence of the appropriate laws is what makes theseconceptualizations per se rather than incidental but it does not followthat those laws must be a part of the explanation Explanations might

94


work not because they include laws but because the descriptions they useare derived from laws In such a case let us say that explanations arebacked by laws but do not include them

What sorts of laws did Aristotle believe backed the explanations ofnon-accidental actions ie ones done for the sake of something andalso done in accordance with deliberate intention Such laws must bethe laws of practical science Aristotlersquos remarks at Metaphysics VI 2lsquono sciencemdashpractical productive or theoreticalmdashtroubles itselfrsquo withthis category of the accidental It is only accidents for which there canbe no science lsquoThat a science of the accidental is not even possible willbe evident if we try to see what the accidental really isrsquo (Metaphysics1064b30) Practical science explains actions by means of practicalsyllogisms eg why (C) a man who desired food went to the marketnamely because (P1) he desired food and (P2) he knew that the marketwas where the food was and (P3) whoever desires something andbelieves that some action is the way to get what he desires does thataction19 So Aristotle holds that explanations of human actions are backedby laws like (P3) which we might call lsquoaction lawsrsquo and whichpresumably have the same epistemic status as natural laws

Aristotlersquos views on laws and necessitation are somewhat morecomplicated than the above account would so far suggest Aristotle believedas we saw that there are some laws that hold only for the most part Herepeatedly informs us that often a predicate will belong to a specific kindonly for the most part (De Generatione Animalium 727b29 770b9ndash13772a35 777a19ndash21 De Partibus Animalium 663b28 Prior Analytics 25b1432b4ndash13 Metaphysics VI 2 1027a20ndash5) In the passage from Metaphysicslisted above Aristotle even argues that the proof of there being accidentaloccurrences rests on the fact that lsquothe majority of things are only for themost partrsquo In the Posterior Analytics itself he says

Some occurrences are universal (for they are or come to be what theyare always and in every case) others again are not always what theyare but only as a general rule for instance not every man can grow abeard but it is a general rule

(Posterior Analytics II 12 96a8ndash19)

If he is to be taken at his word that is if there really are fundamentallyfor-the-most-part laws and not just universal laws knowledge of whichis sometimes incomplete it would have been open to him on thisbasis to say that a manrsquos reaching puberty (remember this will be

95


the lsquowidersquo event which is the full cause and so includes all of thecausally relevant circumstances) caused him to grow a beard or madehis growing a beard more likely without necessitating him to growit His growing a beard when he did could depend on his attainingpuberty without being determined by it Aristotle could have therebyintroduced contingency into his system without introducing accidentswhich fail to have definite causes by denying that all (definite) causesdetermine or necessitate But he does not seem to have seen thispossibility Aristotle seems to have had the materials available withwhich to deny that causes always necessitate but not to have takenthe additional step and deny that they do

Aristotle on scientific explanation

The topic we have been discussing in this chapter has been Aristotlersquosgeneral theory or account of explanation Nothing so far has beensaid about scientific explanation There is a lengthy discussion byAristotle of explanation in the Posterior Analytics20 It is clear thatthis is his account of explanation in the sciences How does thatdiscussion fit into the exposition that I have already offered

To begin with the topic of the Posterior Analytics is knowledge Thisis sometimes translated as lsquoscientific knowledgersquo but the Greek word isepisteme and is sometimes translated with the qualification lsquoscientificrsquoand sometimes without What prompts translators to add the qualificationis clear enough Aristotlersquos paradigm for knowledge at least here isscientific knowledge

Aristotle accepted that there were kinds of knowledge other thantheoretical or scientific knowledge namely practical and productiveknowledge There is the productive knowledge of a craftsman and moregenerally the productive knowledge knowledge-how pursued for the sakeof making something There is also practical knowledge knowledgepursued for the sake of acting and represented by the ability to engage inpractical reasoning In the Nichomachean Ethics he talks of knowledgeof or a science of the good but it is evident that such a practical sciencewould be very different epistemologically from the sciences he speaks ofin the Posterior Analytics (Aristotle himself points this out in theNichomachean Ethics Book I chapter 3)

Even though there are these other sorts of knowledge scientificknowledge for Aristotle deserves special consideration So Aristotle is

96


restricting his discussion in the Posterior Analytics to the kind ofknowledge found in the physical and biological sciences and his remarksthere on explanation are similarly so restricted

Aristotle delimits a separate sphere of scientific explanation as distinctfrom explanation in general and imposes special requirements orconditions on scientific explanation that may not be appropriate forexplanation in other contexts or spheres This contrasts with Platorsquos viewsince for him ordinary explanation if it is to withstand philosophicalscrutiny must pass the same requirements as explanation in science oranywhere else

I have raised the problem of how a philosopher is supposed to justifythe requirements he sets for explanation Arbitrary stipulation fidelity tolinguistic usage sensitivity to metaphysics I followed Moravcsik inclaiming that it was the latter that Aristotle used in his general expositionof explanation But in his discussion of scientific explanation newadditional requirements are imposed on explanation From whence dothey arise For Aristotle special requirements for explanation in sciencearise from considerations about the nature of scientific knowledge and itsobjects

First the link between scientific knowledge and explanation is madein Posterior Analytics

We suppose ourselves to possess unqualified scientific knowledge of athing as opposed to knowing it in the accidental way in which thesophist knows when we think that we know the cause on which thefact depends as the cause of that fact and of no other and further thatthe fact could not be other than it ishellip

(PA I 2 71b8ff)

Aristotle distinguishes two kinds of knowledge knowledge of thebare fact and knowledge of the reasoned fact Knowledge of the barefact is knowledge that Knowledge of the reasoned fact is knowledgewhy which Aristotle calls lsquounqualified scientific knowledgersquo We shallsee this distinction at work later Aristotle can account for knowledgeof the reasoned fact in terms of knowledge of the bare fact andexplanation The view in the above quotation then is this

(A) x knows the reasoned fact that p (knows why p) iff(1) for some q x knows the bare fact that q is the explanation of p and(2) (x knows that) ~p is impossible

97


It is ambiguous in Aristotlersquos text whether lsquox knows thathelliprsquo shouldprecede the lsquo~p is impossiblersquo in clause (A2) But since the point ofexplanation is epistemic it makes better sense of Aristotlersquos intentionsto include the additional requirement An analysis of knowledge-why(knowing the reasoned fact) presupposes a prior grasp of the idea ofknowledge that (knowing the bare fact)

Aristotlersquos view is that all scientific explanations are demonstrations Iclassify him therefore as holding an argument theory of explanation(but only as far as scientific explanation goes not in his general accountdiscussed at length above) Aristotlersquos theory of the demonstration is asketch of what we must possess in order to have understanding in hissense demonstrations must be such that they permit us to meet theconditions for understanding set out in the two clauses of (A) One cansee why Aristotle was led into thinking that explanations in science had tobe demonstrations when one considers what he took to be the nature andobjects of scientific knowledge

Aristotlersquos (A2) commits him to the view that one can only havescientific knowledge of that whose contradictory is impossible (lsquothe factcould not be other than it isrsquo) Aristotle believed that the laws of naturealthough (as we would say) a posteriori were necessary and hence thattheir denials were impossible (Aristotlersquos necessity and impossibility areof course weaker than logical necessity and logical impossibility) lsquohelliptheobject of scientific knowledge can not be other than it isrsquo (PA I 6 74b5)lsquoSince the object of pure scientific knowledge cannot be other than itishelliprsquo (PA I 4 73a21)

Laws are therefore the only suitable candidates for being the objectsof scientific knowledge Normally one would assume that there can bescientific knowledge and explanation of both laws and particular factsbut there is no attempt by Aristotle in the Posterior Analytics to extendthe discussion to include the latter It is true that Aristotlersquos scientist issometimes interested in explaining particular facts (see for example PAII 11 94a36ndashb8) but Aristotle shuns a discussion of such knowledge inthis treatise on scientific knowledge

Scientific knowledge is not possible through the act of perception helliponemust at any rate actually perceive a lsquothis somewhatrsquo and at a definitetime and place but that which is commensurately universal and true inall cases one cannot perceivehellip Seeing therefore that demonstrationsare commensurately universal and universals are imperceptible weclearly cannot obtain scientific knowledge by the act of perceptionhellip

98


So if we were on the moon and saw the earth shutting out the sunrsquoslight we should not know the cause of the eclipse we should perceivethe present fact of the eclipse but not the reasoned fact at all since theact of perception is not of the commensurate universalhellip

(PA I 31)

If we set out to understand and hence explain a law of science therequirement that the explanation take a demonstrative form followsnaturally from two of Aristotlersquos views namely that the objects ofscientific knowledge must be necessary and must be known to be so(or so I interpreted the second clause of the definition of knowledge)Aristotle held that if the conclusion is to be known as necessary itmust follow necessarily from premisses themselves known to benecessary First each step in the inferential chain must be necessarybeginning with the initial premisses lsquothe truth obtained bydemonstrative knowledge will be necessary And since demonstrativeknowledge is only present when we have a demonstration it followsthat demonstration is inference from necessary premissesrsquo (PA 73a22ndash4) and lsquoBut when the middle term [of a demonstration] is fromnecessity the conclusion too is from necessity just as from truth it isalways truersquo (PA 75a4ndash6)

Moreover the connections between each step in the chain must alsobe necessary connections only deductively valid demonstrations areproductive of knowledge lsquohellipdemonstrative knowledge must knowledgeof a necessary nexushellipotherwise its possessor will not knowhellipthe factthat his conclusion is a necessary connexionhelliprsquo (PA I 75a12ndash18) lsquoSinceit is impossible for that of which there is understanding simpliciter to beotherwise what is understandable in virtue of demonstrative understandingwill be necessaryrsquo (PA A4 73a22ndash5) Although Aristotle agrees that theremay be some other kind of knowing he concludes

What I now assert is that at all events we do know by demonstrationBy demonstration I mean a syllogism productive of scientificknowledge a syllogism that is the grasp of which is eo ipso suchknowledge Assuming then that my thesis as to the nature of scientificknowing is correcthellip

(PA I 2 71b17ndash20)

Perhaps one can non-deductively infer a necessary truth from anecessary truth Aristotle nowhere as far as I know explicitly rules

99


this out However even if I know that the premiss in such a non-deductive inference is true and necessary Aristotle would be I thinkloth to allow that I thereby could know rather than just have reasonto believe that the conclusion is true and necessary even if it is soIt is only deduction that ensures knowledge of necessity-preservationfrom premisses to conclusion The deductive requirements ofscientific explanation follow from the very high demands Aristotlemakes on scientific laws (that they are necessary) and on scientificknowledge (to know the reasoned fact that p entails being certainthat ~p is impossible) Aristotle holds a deductivist theory ofexplanation

Since Aristotle held that some laws hold only for the most part howcould there be a demonstration of them Aristotle discusses the form thata demonstration of such a stochastic generalization might take In thePosterior Analytics Book II chapter 12 Aristotle says

In the case of such connections the middle term too must be a generalrule [a rule-for-the-most-part]hellip But we have assumed a connectionwhich is a general rule consequently the middle term B must also bea general rule So connections which embody a general rulehellipwill alsoderive from immediate basic premisses

(PA II 12)

In the Posterior Analytics Book I Aristotle explicitly tells us that wecan have scientific knowledge of what happens for the most part

There is no knowledge by demonstration of chance conjunctions forchance conjunctions exist neither by necessity nor as generalconnectionshellip Now demonstration is concerned only with one or otherof these two for all reasoning proceeds from necessary or generalpremisses the conclusion being necessary if the premisses arenecessary and general if the premisses are general Consequently ifchance conjunctions are neither general nor necessary they are notdemonstrable

(PA I 30)

He seems to be contemplating deductive syllogisms(lsquodemonstrationsrsquo) with lsquofor the most partrsquo premisses and a lsquofor themost partrsquo conclusion although he is not likely to be successful inconstructing valid deductions with this form21

100


Is there any evidence that he might be willing to contemplate anon-deductivist argument theory of explanation Certainly Aristotlehas an account of induction (epagoge) lsquoThus it is clear that we mustget to know the primary premisses by inductionhelliprsquo (PA II 19 A100b5ndash15)22 But these particular instances cannot provide the explanationfor the ultimate principles of a science indeed it would be closer tothe truth to say that it is the ultimate principles which explain theparticular cases23

There is also no doubt that Aristotle recognized something which hewas prepared to call lsquoinductive argumentrsquo He mentions it in Book Ichapter 1 of the Posterior Analytics where he discusses the Socratic ideathat one must know something before one can learn it lsquothe two forms ofdialectical reasoning syllogistic and inductivehelliprsquo He treats it again brieflyin Book I chapter 12 of the Topics In the latter he asserts that of thetwo forms of dialectical argument induction is even more convincingand clearer than deduction All of this suggests although doesnrsquot quitesay that an inductive argument might constitute an explanation of itsconclusion

Also since Aristotle does assert that there are generalizationswhich hold for the most part then if he were to shift from his officialview and consider the possibility of the scientific explanation ofparticular events then any such explanation of a particular eventwhich used a lsquofor the most partrsquo generalization would have to bean inductive or probabilistic explanation since no deductiveinference could capture an explanation with those features Aristotletoys with this thought in one place Poetics 10 the actions lsquoshouldeach of them arise out of the structure of the plot itself so as to bethe consequence necessary or probable of the antecedentsrsquo InRhetoric I 2 1357a25ndash38 and II 25 1402 15ndash1403a15 Aristotleintroduces something which he calls lsquoargument by examplersquo whichis a form of analogical and certainly non-deductive argumentAlthough in what follows I count Aristotle as a deductivist as hecertainly was concerning the explanation of laws whether holdinguniversally or for the most part there is some textual evidence thatsuggests that Aristotle might have been willing to consider adifferent view

101


Aristotlersquos demonstrations

The idea of a demonstration gives content to the two conditionsAristotle requires for knowledge of the reasoned fact What is ademonstration Not just any deductively sound argument is ademonstration (A) states the two conditions required for knowledgeof the reasoned fact that p one concerns the impossibility of ~p theother knowledge of the explanation of p So a demonstration mustdo at least two things (A1) it must provide the explanation of whatwe know (A2) it must lead to knowledge of the necessity of what weknow For Aristotle therefore a demonstration is a deduction that isable to accomplish these two things

In order to meet (A2) Aristotle insists that a demonstration must bea syllogism with necessary premisses and hence a necessary conclusionWhat further conditions does Aristotle lay down to ensure that thesyllogism accomplishes (A1) What has to be the case in order that forsome q one knows that q is the explanation of p Not just any necessaryq that entails a necessary p will do (where of course p and q are bothuniversal generalizations) Suppose p and q are logically equivalent Ifso then a necessary p will entail a necessary q and a necessary q willentail a necessary p yet surely at most only one of them explains theother We assume that the explanation relation is asymmetrical (oranyway non-symmetrical which is enough for the case at hand) for thecases in which we are here interested if p explains q q does not explainp How shall we account for this asymmetry (or non-symmetry) ofexplanation24

As far as I can see all of the remaining six conditions that Aristotleimposes save truth (which I take as implied in the necessity condition inany case) are intended to introduce the requisite asymmetry (or non-symmetry) of explanation Here is his own summary of the additionalrequirements each of which is later developed by a fuller discussion

Assuming then that my thesis as to the nature of scientific knowing iscorrect the premisses of demonstrated knowledge must be trueprimary immediate better known than and prior to the conclusionwhich is further related to them as effect to causehellip Syllogism theremay be indeed without these conditions but such syllogism not beingproductive of scientific knowledge will not be demonstrationhellip


102


A demonstration is not only a deductively valid syllogism fromnecessary premisses to a necessary conclusion Aristotle adds that ademonstration is a special sort of such a syllogism viz one thatmeets the following further six requirements first the premissesmust be true second and third they must be primitive andimmediate Fourth they must be prior to the conclusion drawn fromthem Fifth they must be explanatory of the conclusion which itselfmust be true Sixth they must be more familiar (in nature and to us)than the conclusion Requirements (4)ndash(6) relate to features of thepremisses relative to the conclusion requirements (1)ndash(3) concernthe premisses per se

It is not worthwhile to move through these conditions one by oneThey are not conceptually independent and at least two are equivalent25

In what follows I remark on some of the requirements that are of interestin a rather ad hoc way However the fifth requirement is extremelyimportant and we shall pause to look at it in more detail than the others

Of the six conditions placed on syllogisms that lead to knowledge ofthe conclusion the first Aristotle tells us is that the premisses of anexplanatory demonstration must be true lsquoNow they26 must be truehelliprsquo(PA A2 71b26ndash7) Aristotlersquos argument seems to be that one can come toknow the explanandum conclusion only on the basis of premisses whichone already knows But a necessary condition for knowing the premissesis that the premisses be true So says Aristotle the premisses of anexplanatory demonstration must be true

Is Aristotle right to require the truth of (as we should say) the explanansThis is a requirement which almost every philosopher who has written onexplanation has adopted I will accept without argument this Aristotelianrequirement That easy acceptance requires only the distinction betweenan explanation and a potential explanation (or the explanation that therewould have been ifhellip) It is possible for false empirical statements toexplain potentially It is this sort of thing we have in mind when we saythat some false astronomical theory explained for example the motionof the planets What we mean is that the theory would have explained themotion had it been true

Second and third Aristotlersquos remarks on the immediacy andprimitiveness (the non-demonstrability) of the premisses cannot ofcourse apply as a requirement to every scientific explanatorydemonstration Primitiveness can only apply to the first principles ina scientific chain of such explanatory demonstrations which constitutesthe form that a finished science takes in Aristotlersquos view If contra

103


suppositione such first principles were non-primitive i edemonstrable they could not be the first principles of a science Quiteapart from Aristotlersquos particular theory of science this requirement isinteresting

There is the following trilemma about explanation (there is ananalogous trilemma about epistemic justification) either explanationsregress ad infinitum or there is some circularity in explanation so thatsomething can be part of the explanation for itself or there must besome ultimate explanans which is itself inexplicable or self-explanatoryWe attributed to Plato the view that the Forms are ultimate and self-explanatory the third lemma of the above trilemma Aristotle has thisto say

Now some think that because one must understand primitives there isno understanding others that there is but that there are demonstrationsof everything Neither of these [views] is either true or necessary Forthe one party supposing that one cannot understand in another waymdashthey claim that we are led back indefinitely on the grounds that wewould not understand what is posterior because of what is prior ifthere are no primitives and they argue correctly for it is impossible togo through indefinitely many things And if it comes to a stop andthere are principles [they say] these are unknowable since there is nodemonstration of them which alone they say is understanding but ifone cannot know the primitives neither can one understand whatdepends on them simpliciter or properly but only on the suppositionthat they are the case The other party agrees about understanding forit [they say] occurs only through demonstration But [they argue that]nothing prevents there being demonstration of everything for it ispossible for the demonstration to come about circularly andreciprocally

But we say that neither is all understanding demonstrative but inthe case of the immediates it is non-demonstrablehellip

(PA A3 72b5ndash20)

The first party Aristotle rejects is the party of sceptics who acceptthe first horn of the trilemma and construe it as showing thatunderstanding anything is impossible Explanation they say requiresan infinite regress of explanation and since this is impossibleexplanation is itself impossible The second party accepts the secondcircularity lemma of the trilemma

104


Aristotlersquos theory like Platorsquos embraces the third lemma of thetrilemma There is such a thing according to him as lsquonon-demonstrableunderstandingrsquo Ultimate explanantia (there will be more ultimateexplanantia than there are ultimate sciences for every ultimate sciencewill have to have several such ultimate explanantia) are self-explanatoryIf Aristotle and Plato are right explanation is not an irreflexive relationthere can be things that explain themselves

It may be as Aristotle suggests in the very last chapter of the PosteriorAnalytics that we come to these first principles by means of a processof induction (epagoge) from particular instances (the preciseinterpretation that should be put on Aristotlersquos doctrine of epagoge iscontroversial) But there still will be no explanatory demonstration ofthem As I said before these particular instances cannot provide theexplanation for the ultimate principles of a science indeed it would becloser to the truth to say that it is the ultimate principles which explainthe particular cases

Notice that the idea of the self-explanatory is different from the ideasof both the a priori and the self-evident (I suppose that whatever is self-evident is a priori but not conversely) Whatever is self-evident is self-evidently true but it does not follow that one knows any explanationfor the truth one has thus grasped not even that it is its own explanationOne might see that something is true merely by thinking about orattending to it and this may provide only knowledge of the fact ratherthan knowledge of the reasoned fact What is self-evident may not beself-explanatory

The first principles of science in spite of being self-explanatory (andnecessary) certainly cannot be a priori Indeed if as Aristotle says weobtain them by means of the process of epagoge they cannot be a prioriAristotlersquos claim is that the first principles of a science must be self-explanatory once we have them they explain themselves But he doesnot assert that they are a priori that we could come to know them insome way other than via their instances

The third condition immediacy is a relation that holds between twoterms A and B iff there is no middle term C such that all A are C and allC are B For Aristotle in the finished setting out of a science eachgeneralization should be immediate each generalization should followimmediately from its predecessor in the inferential chain If it does notthen there are some further premisses on which its truth depends orthrough which its truth is mediated such that those premisses have notyet been incorporated into the science

105


Fifth and sixth the premisses in an explanatory syllogism must bemore familiar than and prior to that which they explain Barnes takesthese two requirements priority and familiarity to be equivalent

Let me return to the fourth condition which I omitted The fourthcondition is stated by Aristotle in the following way lsquothe premisses mustbe the explanatory causes of the conclusionrsquo (PA I 2 29) lsquoDemonstrationis syllogism that proves the causehelliprsquo (PA 85b24) Aristotle introducesthe need for this fourth condition at PA A13 78a23ndash78b15 The passageis lengthy but I reproduce it in full because a great deal of my discussionin chapters VI and VII will depend on the insights it contains

Understanding the fact and the reason why differ first in the samesciencemdashand in that in two ways in one fashion if the deductiondoes not come through immediates (for the primitive explanation isnot assumed but understanding of the reason why occurs in virtue ofthe primitive explanation) in another if it is through immediates butnot through the explanation but through the more familiar of theconverting terms For nothing prevents the non-explanatory one of thecounterpredicated terms from sometimes being more familiar so thatthe demonstration will occur through thisEg that the planets are near through their not twinkling let C be theplanets B not twinkling A being near Thus it is true to say B of Cfor the planets do not twinkle But also [to say] A of B for what doesnot twinkle is nearhellip So it is necessary that A belongs to C so that ithas been demonstrated that the planets are near Now this deduction isnot of the reason why but of the fact for it is not because they do nottwinkle that they are near but because they are near that they do nottwinkleBut it is also possible for the latter to be proved through the formerand the demonstration will be of the reason whymdasheg let C be theplanets B being near A not twinkling Thus B belongs to C and A toB so that A belongs to C And the deduction is of the reason why forthe primitive explanation has been assumedAgain [take] the way they prove that the moon is spherical through itsincreasesmdashfor if what increases in this way is spherical and the moonincreases it is evident that it is spherical Now in this way the deductionof the fact comes about but if the middle term is posited the other wayabout [we get the deduction] of the reason why for it is not becauseof the increases that it is spherical but because it is spherical it getsincreases of this sort Moon C spherical B increase A

106


But in cases in which the middle terms do not convert and the non-explanatory term is more familiar the fact is proved but the reasonwhy is not

(PA A13 78a23ndash78b15)

The same point is made at PA II 16 98b4ndash24 A plant is deciduousiff it has broad leaves but it is deciduous because it is broad-leavedand not vice versa (Jonathan Barnes tells me that poor Aristotle didnrsquotknow about the larch which is deciduous but not broad-leaved) Ifwe know that all vines are broad-leaved we can infer that vines aredeciduous if we know that vines are deciduous we can infer thatthey are broad-leaved Since lsquodemonstration through the cause is ofthe reasoned fact and demonstration not through the cause is of thebare factrsquo one who knows the broad-leavedness of vines through thedeciduousness lsquoknows the facthellipbut not the reasoned factrsquo Such aperson does not know why the vine is broad-leaved he only knowsthat it is

The lesson of these examples is this To use Aristotlersquos second examplefrom the long quotation above assuming that things increase in a certainway if and only if they are spherical compare the following twodeductions

(1) Things increase in a certain way iff they are spherical (2) The moon increases in just that way

(3) The moon is spherical

(4) Things increase in a certain way iff they are spherical (5) The moon is spherical

(6) The moon increases in just that way

Aristotle claims that (4) and (5) explain (6) whereas (1) and (2) do notexplain (3) If we have two convertible terms (lsquoArsquo and lsquoBrsquo areconvertible terms iff all As are Bs and all Bs are As) we can oftenconstruct deductions that meet all of his other conditions for ademonstration yet fail to be productive of lsquoknowing the reason whyrsquoThe premisses might be immediate more familiar (to us at least)necessary universal true and deductively imply the conclusion SoAristotle feels compelled to impose a further requirement on thesyllogism in virtue of which it can count as productive of understandingwhymdashnamely the premisses must be lsquoexplanatory of the conclusionrsquo

107


lsquoAnd the deduction is of the reason why for the primitive explanationhas been assumedrsquo (from the long quotation above)

Aristotlersquos example of the moonrsquos shape and increase does not employonly laws in both premisses and conclusion which is what he is officiallymeant to be discussing but the example of the vines does and in any caseit is not difficult to construct many similar examples having the followingform using only generalizations let and lsquo(x) (Qx Rx)rsquo be

the premisses and be the conclusion in a deduction It follows

that this will also be a deduction let and lsquo(x) (Qx Rx)rsquo be

the premisses and be the conclusion One of the deductions

may be explanatory if so typically the other would not beLetrsquos recall the explication of knowledge with which we began


In explicating (A1) Aristotle tells us that we require a demonstrationthat meets six conditions It might seem that Aristotle is going tooffer us a lsquoreductiversquo explication of knowledge of the reasoned fact(understanding) in terms that refer to ideas such as demonstrationnecessity and so on However one of the crucial conditions for anargumentrsquos being a demonstration is that the premisses must beexplanatory of the conclusion

We have not then in any sense lsquoeliminatedrsquo the idea of explanationfor Aristotle has used the idea of explanation in accounting for the firstclause of (A) Can we further eliminate this final reference to explanationor is it simply to be taken as a primitive

Baruch Brody27 sketching what he calls lsquoan Aristotelian theory ofexplanationrsquo claims that the point of the above discussion by Aristotle isthat a certain disjunctive condition typically omitted in modern theoriesof explanation must obtain in order for a deduction to count as anexplanation

a deductive-nomological explanation of a particular event is asatisfactory explanation of the event when (beside meeting all ofHempelrsquos requirements) its explanans contains essentially a descriptionof the event which is the [efficient] cause of the event described in theexplanandumhellip [Further] we can set down another requirement for

108


explanation as follows a deductive-nomological explanation of aparticular event is a satisfactory explanation of that event when (besidemeeting all of Hempelrsquos requirements) its explanans containsessentially a statement attributing to a certain class of objects a propertyhad essentially by that class of objects (even if the statement does notsay that they have it essentially) and when at least one object involvedin the event described in the explanandum is a member of that class ofobjects

(Brody 197226)

On Brodyrsquos account of Aristotlersquos theory of scientific explanationone knows why only if inter alia one knows the efficient cause orthe essence of what it is that one is trying to explain But there is noreason to think that Aristotle himself is limiting lsquocausersquo to efficientcauses (or motion-originator as I have preferred to put it) andessences Aristotle in the long passage I quoted has in mind the aitiaiin any of his four permitted senses Further he says

We think we have scientific knowledge when we know the causeand there are four causes (1) the definable form (2) anantecedent which necessitates a consequent (3) the efficientcause (4) the final cause Hence each of these can be a middleterm of a proofhellip

(PA II 11)

Aristotlersquos theory of scientific knowledge (understanding)presupposes and makes use of (E) his account of explanation ingeneral but adds further requirements to it His account of scientificknowledge requires (E) his general account to spell out what isinvolved in explanation in a non-circular way28

Summary

What lessons has Aristotle taught us about explanation that we shouldcarry forward to later chapters I think there are at least four Firstthe connection he sees between a theory of explanation andmetaphysics provides a methodological alternative to what I called

109


lsquothe language usersrsquo approachrsquo I return to this theme in chapter VIISecond his insight into per se causation offers the beginnings of atheory of how our conceptualization or view of things makes adifference to explanation This forms the basis of my discussion inchapter V Third Aristotle believes that all explanations are argumentsand that laws have an especially central role to play in explanationChapter VI returns to these themes Finally Aristotlersquos requirementthat no argument can be an explanation unless it mentions the causeof what is to be explained in the premisses suggests that anyacceptable theory of explanation must be in some sense a causaltheory of explanation I examine this question as well in chapter VII

110

CHAPTER IV

Mill and Hempel on Explanation

Carl Hempel in his 1948 article lsquoStudies in the Logic of Explanationrsquoclaims that at least part of the account of explanation that he developshas been defended by several previous writers lsquoThe account givenabove of the general characteristics of explanation and prediction inscience is by no means novel it merely summarizes and statesexplicitly some fundamental points which have been recognized bymany scientists and methodologistsrsquo1 Among those precursorsHempel lists John Stuart Mill and offers the following two quotationsin support of this claimrsquo lsquoAn individual fact is said to be explainedby pointing out its cause that is by stating the law or laws of causationof which its production is an instancersquo and lsquoa law or uniformity ofnature is said to be explained when another law or laws are pointedout of which that law is but a case and from which it could bededucedrsquo2 It would seem that Mill subscribed to a deductivist accountof explanation for Mill all explanations are a subset of the set ofdeductively valid arguments namely those which meet additionalrequirements to be specified Hempel agrees that some (although notall) explanations conform to the deductive model of explanation thatJohn Stuart Mill outlines

On the other hand Mill holds a peculiar account of deduction lsquoIt mustbe granted that in every syllogism considered as an argument to provethe conclusion there is a petitio principiirsquo (II III 2) Deductive inferenceaccording to Mill is in some sense circular and is in fact founded uponsome sort of non-deductive inference

111


Now there is certainly no formal contradiction in Millrsquos holding botha deductivist theory of explanation and a lsquoreductivistrsquo account of deductionas founded on a special kind of non-deductive inference However evenif formally consistent the conjunction of the two views seems odd andsuspiciously unmotivated The thought behind Millrsquos rather murky doctrineabout deduction is that one cannot learn something new via a deductionDeduction cannot advance knowledge One would have imagined thatthis epistemic down-grading of deduction would have carried over to Millrsquosviews on explanation which would have provided a natural extension ofthe doctrine After reading Mill on deduction one might reasonably expecta non-deductive view of explanation

But Mill remained a deductivist about explanation In none of hisremarks about explanation does Mill return to his view of deduction orremark upon how his deductivist account of explanation fits with thatview The oddity in this conjunction of views is noted by Alan Ryan inhis book on Mill but he does little to dispel the worry that the views donot fit together well3 One thing that I shall do in this chapter is to discussMillrsquos views first on explanation and then on deductive inference to seeif we can find better clues for why he might have held these views intandem In fact I think that there is a natural explanation for why Millthought that these two views fit together harmoniously

This chapter offers an account of Millrsquos and Hempelrsquos views onexplanation which are at any rate superficially very similar Millrsquos viewson explanation will be taken from his remarks in A System of Logic Hempelhas written extensively on explanation but I will limit myself to two ofhis articles lsquoStudies in the Logic of Explanationrsquo (first published in 1948)which I sometimes refer to as lsquothe early articlersquo and lsquoAspects of ScientificExplanationrsquo (1965) which I sometimes refer to as lsquothe later articlersquo4

These two articles contain the essentials of his views and are the startingpoints for any contemporary discussion of the nature of explanation

Mill is part of the empiricist tradition in philosophy From Hobbes andBacon through Locke to Berkeley and Hume there is an increasinglycritical philosophical rejection of concepts or ideas which cannot be traceddirectly to experience Substance matter essence or form the self andcausation are just some of the concepts about which various empiricistphilosophers expressed doubts and reservations None as far as I knowhad much if anything to say directly about the concept of explanationBut it is easy to see why the Aristotelian or Aristotelian oriented scholastictraditions of explanation would have made them suspect explanation hadthey turned their attention to it

112


As we saw in the last chapter explanation for Aristotle had been tiedto such ideas as form or essence matter goal or end and efficient causeEach of these ideas is challenged or found perplexing in some way by atleast one of Millrsquos empiricist predecessors Substance and matter arecriticized by Berkeley efficient cause by Hume both essence and finalcause by Hobbes All of these ideas seem to transcend all possibleexperience and hence to present a problem for the empiricist Either theymust be rejected or it must be shown that despite appearances they donot transcend experience after all

In many ways it is surprising that no empiricist philosopher beforeMill turned in an explicit way to the scrutiny of the concept of explanationwhich hadmdashgiven its connections with these other suspect notionsmdasheveryappearance of being experience-transcendent Of course many empiricistphilosophers held views which have consequences for a theory ofexplanation For example much of what Bacon says is pertinent to atheory of explanation5 Lockersquos belief in the external world can beconstrued indeed has been construed as an example of inference to thebest explanation Berkeleyrsquos philosophy of science Humersquos variousscepticisms all of these topics will have important implications forexplanation But Mill is as far as I know the first empiricist philosopherto have explicitly addressed himself to the question of the nature ofexplanation and it is this fact that I find surprising

As I suggested above there are two reactions possible for an empiricistto any concept that appears experience-transcendent First the philosophercan confirm that the concept not only appears but is experience-transcendent and therefore that he wishes to reject or eliminate the notionExamples of this strategy include Berkeley on material substance Hobbeson immaterial substance and final cause and Hume on the continuingand independent existence of objects Second the philosopher can holdthat the appearance of experience-transcendence is misleading that areconstruction of the concept or notion can be offered such that on thatreconstruction the concept can be shown to be directly tied to experienceExamples of this second strategy include Hume on causation Berkeleyon objects like tables chairs and trees (including the one in the quad)and Hobbesrsquos linguistic construal of essence

The same choice of strategies is available to an empiricist in a discussionof explanation If explanation invokes experience-transcending elementsit can be eliminated or rejected from sound philosophy and science Anexample of this strategy is adopted by Pierre Duhem in his The Aim andStructure of Physical Theory6 Duhem defines lsquoto explainrsquo as lsquoto strip

113


reality of the appearances covering it like a veil in order to see the barereality itselfrsquo Since lsquoThe observation of physical phenomena does notput us into relation with the reality hidden under the sensible appearancesrsquoDuhem has little difficulty in showing that explanation as he understandsit is a lsquometaphysicalrsquo (ie experience-transcendent) idea If the purposeof physical theory were to explain physical theory would be subordinateto metaphysics

There is for Duhem an alternative way to understand the purpose ofphysical theory namely that the aim of physical theory is merely tosummarize and classify logically a group of experimental laws lsquowithoutclaiming to explain these lawsrsquo Having rejected explanation as a legitimateaim of science Duhem claims that lsquoA physical theory is not an explanationIt is a system of mathematical propositions deduced from a small numberof principles which aim to represent as simply as completely and asexactly as possible a set of experimental lawsrsquo (Duhem 197719) Sinceexplanation is connected for Duhem with a non-empirical conception ofreality it has no place in science

Duhem then represents one empirically minded strategy for dealingwith explanation that of rejecting explanation as an experience-transcendent and hence illegitimate (at least for science) notion or ideaJohn Stuart Mill represents the other empirically minded strategy theattempt to reconstruct or reconstrue explanation as an empiricallyacceptable idea Mill goes out of his way to stress that his theory ofexplanation is a case of making explanation acceptable to the empiricisthe eschews any idea of explanation as unravelling the deeper mysteriesof nature

The word explanation is here used in its philosophical sense What iscalled explaining one law of nature by another is but substituting onemystery for another and does nothing to render the general course ofnature other than mysterious we can no more assign a why for themost extensive laws than for the partial ones

(Mill 1970310)

For Mill explanation has none of the mystery attributed to it by Duhemor other philosophers We need only the ideas of a law of naturecause and causal law and deduction in order to explicate the idea ofexplanation We do not need Platorsquos Forms or Aristotlersquos final causesand essences or Duhemrsquos non-sensible reality

114


Mill thinks that the ideas of a law of nature and a causal law are safefor empiricists He has previously explained a law of nature as ageneralization to the effect that lsquoa certain fact invariably occurs whenevercertain circumstances are present and does not occur when they are absentrsquo(Mill 1970206) Such uniformities are among either simultaneous orsuccessive phenomena and causal laws are of the latter kind lsquoThe law ofcausationhellipis but the familiar truth that invariability of succession is foundby observation to obtain between every fact in nature and some other factwhich has preceded ithelliprsquo (p213) A particular causal law is merely aspecific invariability of succession between facts of two kinds

Mill explicitly rejects any non-empirical idea of causation asmetaphysical

The notion of causation is deemed by the schools of metaphysics mostin vogue at the present moment to imply a mysterious and most powerfultie such as cannot or at least does not exist between any physicalfact and that other physical fact on which i t is invariablyconsequenthellipand thence is deduced the supposed necessity of ascendinghigher into the essences and constitutions of thingshellip

(Mill 1970213)

Mill has thereby rendered both lsquolaw of naturersquo lsquocausersquo and lsquocausallawrsquo acceptable for an empiricist And since explanation is built outof these concepts (and deduction) it is acceptable as well

Mill admits that lsquoexplanationrsquo has an ordinary meaning as well as thelsquoscientificrsquo one that he proposes to give it In lsquocommon parlancersquo anexplanation often replaces the unfamiliar by the familiar but Mill notesthat in science just the reverse is usually the case

it resolves a phenomenon with which we are familiar into one of whichwe previously knew little or nothinghellip It must be kept constantly inview therefore that in science those who speak of explaining anyphenomenon mean (or should mean) pointing out not some morefamiliar but merely some more general phenomenon of which it is apartial exemplificationhellip

(Mill 1970310ndash11)

Mill contrasts the meaning of lsquoexplanationrsquo in ordinary parlance andthe meaning he will attach to it (and what those who use it in sciencelsquoshould meanrsquo by it) In science typically the unfamiliar explains the

115


familiar in the ordinary sense the familiar explains the unfamiliarHempel is part of the tradition that we can trace back to Mill of making

explanation metaphysically safe for the empiricist philosopher As far asI know Hempel does not say this explicitly but I claim that it is implicitin the way in which he executes the project of explicating explanation Aswe shall see explanation for Hempel can be explicated via the conceptsof deductive and statistical inference truth empirical content and lawlikegeneralizations All of these concepts are for Hempel comprehensiblewithin the terms of empiricist philosophy although he notes that the ideaof lawlikeness presents difficulties which have lsquoproved to be highlyrecalcitrantrsquo (Hempel 1965338)7

Millrsquos account of explanation laws of coexistence and succession

In one sense an explanans and an explanandum are sentences Butas I have claimed sentence explanation is parasitic on the idea ofnon-sentence explanation In the non-sentence sense for Mill whatsorts of entities are explanantia (do the explaining) and explananda(are explained)

Mill sometimes says that facts explain facts lsquoAn individual fact is saidto be explained by pointing out its causersquo (Mill 1970305) In the verynext sentence Mill says lsquoa conflagration is explainedhelliprsquo and presumablya conflagration is an event Events are not I assume facts

Explanations need laws on Millrsquos theory Laws are uniformities anduniformities are patterns of events or some such However on oneoccasion Mill calls a uniformity lsquoa factrsquo If we distinguish between singularfacts like the fact that some particular conflagration has some feature orproperty and universally general facts like the fact that for all objects ifthey have property P then they have property Q we could think ofuniformities or laws as universally general facts Mill also tells us thatlaws are explained by laws from which the former can be deduced But itis sentences which are deduced from sentences Mill must be thinking oflaws in this last context as sentences that express or state such uniformitiesof nature

Mill switches freely between talk of events and facts as what explainand are explained without much attention to the matter This is the firstwe have seen of facts other than the brief mention of them by Salmon inchapter I facts did not figure in either Platorsquos or Aristotlersquos ontology ofexplanation Whatever facts are they are not events or even patterns of

116


events although there is the fact that some event occurred or the fact thatsome law or pattern of events obtains I return to these questions aboutthe ontology of explanation in the next chapter

Millrsquos definition of explanation which I quoted on page 110 and repeatmore fully here is put rather simply and baldly It is intended to coverboth the case of the explanation of particular matters of fact and theexplanation of general laws

The word lsquoexplanationrsquo occurs so continually and holds so importanta place in philosophy that a little time spent in fixing the meaning ofit will be profitably employed

An individual fact is said to be explained by pointing out itscause that is by stating the law or laws of causation of which itsproduction is an instance Thus a conflagration is explained whenit is proved to have arisen from a spark falling into the midst of aheap of combustibles and in a similar manner a law of uniformityof nature is said to be explained when another law or laws arepointed out of which that law itself is but a case and from which itcould be deduced

(Mill 1970305)

Let me enlarge on my earlier remarks about these lsquoempirically safersquoideas of law and causal law Mill distinguishes between uniformitiesof coexistent phenomena and uniformities of successive phenomenalsquoThe order of the occurrence of phenomena in time is either successiveor simultaneous the uniformities therefore which obtain in theiroccurrence are either uniformities of succession or of co-existencersquo(Mill 1970377) As the names imply the first kind of uniformity isof two sorts of things or events that happen at the same time thelatter of two types that occur at successive times

Uniformities of succession which are causal are invariable andunconditional regularities of experience The regularity of night and dayis a good example of a uniformity of succession However much the night-day sequence might be a uniformity of succession it is not anunconditional uniformity of succession and hence not a causal uniformityWe can see that this uniformity is conditional on other things Shouldthese other things (eg the rotation of the earth) cease there might beperpetual day unsucceeded by night or perpetual night unsucceeded byday The uniformity is a causal uniformity if and only if it is a uniformityof succession which is unconditional and invariable

117


Millrsquos causation is what I have called deterministic causation To repeatlsquoThe law of causationhellipis but the familiar truth that invariability ofsuccession is found by observation to obtain between every fact in natureand some other fact which has preceded ithelliprsquo (Mill 1970 213) Mill arguesthat (1) there is a cause for everything that happens (2) that every suchcause is a determining cause That is he denies both nondeterminism andindeterminism

Given a cause an effect of the appropriate sort invariably follows AsI claimed in chapter I this commitment to deterministic causation willlead Mill to hold some form of an epistemic certainty model of explanationAnd this is indeed what he does hold since he believes that an explanationis always a deductive argument whose conclusion is the statement of thefact to be explained

Mill holds that explanation of a particular fact or event is always byway of citing the invariable law of succession (ie the causal law) onwhich the lsquoproductionrsquo of that fact or event depends Presumablyalthough he does not say it we can take it that he means that the fullexplanation involves both the law of invariable succession and anoccurrence of a token of the type mentioned in the antecedent clause ofthat law Explanation of a universally general fact (a uniformity) is byway of pointing to the more general law or uniformity of which it is aspecial case (lsquofrom which it could be deducedrsquo) Explanation of bothkinds of singular facts and of uniformities requires only invariabilityof succession8 of kinds or types9 and deduction

Mill draws a contrast between ultimate and derivative laws lsquoFrom alimited number of ultimate laws of causation there are necessarilygenerated a vast number of derivative uniformities both of successionand co-existencersquo (Mill 1970339) There can be both uniformities of co-existence and non-causal uniformities of succession (like that of nightand day) at the level of derived laws Sometimes we know on which lawsderived laws depend in other cases we presume that these uniformitiesare derived but we have not actually been able to discover on whichmore fundamental laws they depend These latter are what Mill callslsquoempirical lawsrsquo

It is implied therefore in the notion of an empirical law that it is notan ultimate law that if true at all its truth is capable of being andrequires to be accounted for It is a derivative law the derivation ofwhich is not yet known

(Mill 1970338)

118


Mill says that lsquoFrom a limited number of ultimate laws of causationthere are necessarily generated a vast number of derivativeuniformities both of succession and of co-existencersquo (Mill 1970339)In some cases we can explain the derivative uniformities on the basisof fundamental laws alone But in other cases we need also initialparticular information about lsquothe collocation of some of the primevalcauses or natural agentsrsquo or the lsquomode of co-existence of some of thecomponent elements of the universersquo This information is anomic itis a brute fact that there is just this distribution of things in the universeor that particular causes exist in just the number or distribution thatthey do We can explain derivative uniformities of both kinds(coexistence and non-causal succession) by ultimate laws of causationsometimes in conjunction with ultimate facts about the distributionof natural causal agents But so far Mill seems to say that forwhatever is explainable ultimately a causal law is part of theexplanation for it

Do laws of coexistence as well as laws of succession have anyultimate explanatory value for Mill That they can at least sometimesbe explained is not open to doubt the question is whether they can beused to explain anything either ultimately or lsquoin the interimrsquo If theanswer to the above question is lsquonorsquo then there is a sense in whichfor Mill all explanation is causal explanation If the answer is lsquoyesrsquothen there is room in science for ultimately non-causal explanationsof things explanations which do not rely upon causal laws For thesake of convenience I adopted in chapter I the assumption that allexplanation of particulars is causal explanation Perhaps Millrsquostreatment of this question will help us see whether this assumption isat all plausible

Mill generally down-grades uniformities of coexistence He explicitlyconsiders two sorts of cases First some of these will be the result of theoperation of a single law of causation as when a single cause invariablyhas two effects lsquoIn the same manner with these derivative uniformities ofsuccession a great variety of uniformities of coexistence also take theirrisersquo (Mill 1970378) Suppose that As cause Bs and As cause Cs andthat as a consequence there is a derivative regularity of coexistence tothe effect that Bs iff Cs Such a regularity is nomic and Mill is quitehappy to call the statement of it a law lsquoThe only independent andunconditional co-existences which are sufficiently invariable to have anyclaim to the character of laws are between different and mutuallyindependent effects of the same causehelliprsquo (p227)

119


However even though the regularity is nomic it is not explanatoryOne cannot explain the occurrence of a B by the occurrence of a C orvice versa So nomic uniformities of coexistence that owe their origin tothe fact that some single cause has more than one effect will not themselveshave any explanatory value

The second case of a law of coexistence that he considers is this Supposethere is a uniformity of coexistence that arises from the fact that differentlsquoprimevalrsquo causes happen to coexist there being no further causal explanationwhy this should be so As a matter of brute fact about the distribution andnumber of causal agents in the universe there is a B iff there is a C As wewould put it it is only an accidental generalization that Bs iff Cs Mill himselfmakes the point by denying that the universal generalization is lsquounconditionalrsquo(he even sometimes denies that it is lsquouniversalrsquo but he must mean by thisonly that it is not lsquounconditionalrsquo) Such an accidental generalization is notnomic or lawlike at all and one cannot therefore explain the occurrence of aB by the occurrence of a C or vice versa10

Such accidental generalizations could arise in either of two ways Bsand Cs might themselves be primeval causes or natural agents or Bsmight be the effects of one such type of agents and Crsquos the effect of adifferent type If there is a uniformity of coexistence between two lsquoprimevalcausesrsquo or if the uniformity of coexistence is the effect of two differentlsquoprimeval causesrsquo then the uniformity of coexistence is about two typesof occurrence which coexist lsquocasuallyrsquo and not universally as Millmisleadingly puts it

Since everything which occurs is determined by laws of causation andcollocations of the original causes it follows that the co-existenceswhich are observable among the effects cannot be themselves thesubject of any similar set of laws distinct from the laws of causationUniformities there are as well of co-existence as of succession amongeffects but these must in all cases be a mere result either of the identityor of the co-existence of their causeshellipit follows that (except in thecase of effects which can be traced immediately or remotely to thesame cause) the co-existence of phenomena can in no case be universalunless the co-existences of the primeval causes to which the effectsare ultimately traceable can be reduced to a universal law but wehave seen that they cannot There are accordingly no original andindependent in other words no unconditional uniformities of co-existence between effects of different causeshellip

(Mill 1970227)

120


Such a co-existence of two primeval causes or the effects of twoindependent primeval causes cannot be unconditional it is merely alsquocasualrsquo (ie accidental or non-nomic) collocation lsquothere ishellipnouniformity no norma principle or rule perceivable in the distributionof the primeval natural agents through the universersquo (Mill 1970340)Uniformities of coexistence not resulting from the operation of a singlecause are not unconditional and hence do not deserve the title oflsquolawrsquo at all A fortiori they are not explanatory

Thus far the only genuine laws available to play any part in explanationwould seem to be causal laws This is however not the position that Millfinally adopts In his discussion of kinds and empirical laws (pp377ndash81) headmits ultimate laws or uniformities of coexistence not dependent on causationlsquothere must be one class of co-existences which cannot depend on causationthe co-existences between the ultimate properties of thingshellip Yet amongthese ultimate properties there are not only co-existences but uniformities ofco-existencersquo (p379) These ultimate uniformities of coexistence are lawlikeand hence not to be confused with the brute and inexplicable coexistence ofprimeval causes or collocations or the derivative effects of them

Millrsquos ultimate laws of coexistence presuppose the idea of natural kindslsquolaws of this type assert that there is an invariable concomitance of determinateproperties in every object that is of a certain kindrsquo Millrsquos examples includeblackness and being a crow woolly-hairness and being a negro11 He is thinkingof these claims viz all crows are black as claims about denotation and notabout connotation they are not verbal but real truths if they are true Hewarns us that it is hard to be sure that these coexistences are not just jointeffects of a single cause but he is willing to admit that there must be someuniformities of coexistence which are genuinely uniformities of coexistencebetween the ultimate properties of kinds and lsquoit is of these only that the co-existences can be classed as a peculiar sort of laws of naturersquo (Mill 1970380)

Uniformities of co-existence then not only when they are consequencesof laws of succession but also when they are ultimate truths must beranked for the purposes of logic among empirical laws and areamenable in every respect to the same rules with those unresolveduniformities which are known to be dependent on causation

(Mill 1970386)

Ultimate uniformities of coexistence are laws but lsquomust be rankedamong empirical lawsrsquo they are lsquoamenable in every respect to the

121


same rulesrsquo as empirical laws But he does not quite say that they areempirical laws But if they are not empirical laws why are theylsquoamenable in every respect to the same rulesrsquo

Mill does not place great reliance on these ultimate laws of coexistencewhich do not depend on causation mainly because he holds thatconcerning any uniformity that we hold to be such an ultimate uniformityof coexistence it can be overturned by the finding of a singlecounterexample

Universal propositions therefore respecting the properties ofsuperior kinds unless grounded on proved or presumed connectionby causation ought not to be hazarded except after separatelyexamining every known sub-kind included in the larger kindhellip Thusall the universal propositions which it has been attempted to laydown respecting simple substanceshelliphave with the progress ofexperience either faded into inanity or been proven to beerroneoushellip


It is this feature of them which makes them comparable to merelyempirical laws

But he did not deny that there were such laws whatever epistemicproblems there may be in knowing what they are and indeed his theoryof natural kinds presupposes that there must be for whatever ultimateconstituents of things that there are such laws of coexistence which donot depend on causation Mill is only wary of being able to identifycorrectly the coexistences which are the unconditional ultimate ones Soultimate laws of coexistence are not empirical laws because they are notderived laws But epistemologically the warrant we have for ultimatelaws of coexistence is like the warrant we have for empirical laws andtherefore he draws the comparison between them

Non-accidental laws of coexistence were also allowed by Ernest Nageland for reasons similar to Millrsquos12 Nagel says that this type of law restson the assumption that there are natural kinds of substances It is notclear from Nagelrsquos discussion precisely what such a law would assertbut something like this is what is suggested by his remarks lsquoRock salthas a melting point of 804 degrees Centigrade and a density of 2163rsquoNagel includes this in his list of laws lsquothat are used as explanatory premisesin various scienceshelliprsquo Mill does not assert like Nagel that we can usethese laws of coexistence in explanation But equally he does not say that

122


we cannot use them in order to explain something (of course we cannotexplain them for they are for Mill ultimate laws)

This then raises an interesting question for Mill although not one towhich he addressed himself clearly and explicitly Mill stresses theimportance of causal explanation (explanation by causes or bysubsumption under causal laws) Is there any reason why we cannot usethese ultimate coexistences between the properties of kinds or anycoexistences of properties dependent on them as the explanans in someexplanation These laws of coexistence might have explanatory power intwo ways First ultimate non-causal uniformities of coexistence mightexplain derivative non-causal uniformities of coexistence Second it isnot clear why Mill should limit as he does explanation of singular factsto their causes Suppose that it is an ultimate law of the uniformity ofcoexistences for the kind crow that all crows are black It would seementirely in keeping with the general thrust of his empiricist philosophy ofexplanation to argue that in such a case we could explain why a particularbird is black on the grounds that it is a crow and that it is an ultimate lawthat all crows are black This is certainly the spirit as we shall see inwhich Hempel develops the theory If we did develop Millrsquos theory ofexplanation in this way we could produce examples of the explanation ofa feature of a thing by one of its coexistent features and hence on Millrsquosaccount of causation examples of non-causal explanation

To whatever extent these ultimate laws of uniformity of coexistencemay have explanatory power it would be only a most reluctant admissionby Mill dragged from him unwillingly and tentatively There is no doubtthat he is happiest with causal laws Mill therefore turns his attention tothe discussion of explanation of and by causal laws laws of the invariableand unconditional succession of phenomena rather than to the possibilityof explanation by laws of the simultaneity or coexistence of phenomena

Mill spends some time in discussing the explanation of causal lawsand delineates three subspecies of such explanations lsquoThere are thenthree modes of explaining laws of causation or which is the same thingresolving them into other lawsrsquo (Mill 1970310) First there is the case oflsquoan intermixture of laws producing a joint effect equal to the sum of theeffects of the causes taken separately The law of the complex effect isexplained by being resolved into the separate laws of the causes whichcontribute to itrsquo (p305) Second there is the case in which a uniformitybetween two kinds of facts is shown to be the result of two uniformitiesone linking the first kind of facts with a new third kind and anotheruniformity linking the third kind with the second lsquobetween what seemed

123


the cause and what was supposed to be its effect further observationdetects an intermediate linkrsquo (p306) The third is the subsumption of alaw by another law lsquoThis third mode is the subsumptionhellipof one lawunder another or (what comes to the same thing) the gathering up ofseveral laws into one more general law which includes them allrsquo (p309)Mill remarks that in all three cases the explaining laws are always moregeneral than the laws to be explained so all three cases are examples ofthe explanation of the less general by the more general lsquoIt is of importanceto remark that when a sequence of phenomena is thus resolved into otherlaws they are always laws more general than itselfrsquo (p307)

Mill repeatedly uses the language of lsquoresolutionrsquo in all three cases Inat least one passage Mill speaks of the lsquoeliminationrsquo of the less generallaw by the more general (Mill 1970309) In more modern terminologywe might say that this type of explanation of a law by other laws is oneform of reduction The resolved or reduced law can be seen to be nothingmore than a particular instance or application of more general resolvingor reducing laws

Millrsquos account of explanation the symmetry thesis

Mill asserts what has come to be called lsquothe symmetry thesisrsquomdashthatis the claim that there is a symmetry of sorts between explanationand prediction This symmetry thesis should be distinguished from asecond and different question of symmetry that we shall be discussinglater the (controversial) claim that explanation is itself asymmetricthat if p explains q it follows that q does not explain p When I speakof the symmetry thesis I shall mean the question of the symmetrybetween explanation and prediction When I want to speak of thesecond question I shall speak of the symmetry (or asymmetry ornon-symmetry) of the explanation relation Mill does not addresshimself explicitly to this second question at all

Mill says lsquoAs already remarked the same deductive process whichproves a law or fact of causation if unknown serves to explain it whenknownrsquo (Mill 1970310) One and the same deduction can answer eitherof two questions lsquoGiven a certain combination of causes what effectwill follow and What combination of causes if it existed wouldproduce a given effectrsquo (p 303) In the first case we predict what willhappen (Mill speaks of proving what will happen rather than in termsof predicting) in the second we explain what we know to have happened

124


The symmetry thesis holds that there is only a pragmatic or epistemicbut no logical difference between explaining and predicting Explainingand predicting are human activities both of which involve the producingof a deduction The difference between these activities is only adifference in what the producer of the deduction knows just before thetime at which the deduction is produced Whether I fully explain why eoccurred or fully predict that e will occur the deduction produced willbe the same

Recall the distinction I drew in the first chapter between explanationas a process (or activity) and explanation as the product of such an activitySuch a product is according to Mill a deductive argument Millrsquossymmetry thesis can be expressed by making use of this distinction Inthe product sense explanations and predictions are identical One and thesame deduction is both an explanation (product) and a prediction (product)The difference between explanation and prediction is only between actsof explaining and acts of predicting

The symmetry thesis more generally is this (a) the informationproduced in a (successful) explanatory act could have been the informationproduced in a (successful) act of prediction (b) the information producedin a (successful) act of prediction could have been the informationproduced in a (successful) act of explanation In the process or activitysense explanations and predictions differ Consider a case of explainingthat p and a case of predicting that p I will know or believe cruciallydifferent things in the two cases and this will mean that what activity Iam engaged in in the two cases is different But in the product senseaccording to the symmetry thesis explanations and predictions of thesame thing do not differ at all

The plausibility of the symmetry thesis is closely tied to construingboth explanations and predictions as arguments If explanations andpredictions are both arguments it is perhaps not a large leap of faith tohold that the argument produced in an act of the one type of activity willbe identical to the argument that would have been produced in an act ofthe second type But suppose that explanations or predictions are notarguments Whether the symmetry thesis is held to be true will depend onthe details of the non-argument view But it would be open to such a viewto claim that the information content of explanations and predictions differand therefore that the symmetry thesis is false

125


Mill on ultimate explanations

We have already mentioned both Platorsquos and Aristotlersquos views onultimate explanations Both agree that the regress of explanation mustcome to some end Mill agrees with that view It is now time to lookat this question in some more detail First though I want to developa terminology in which to discuss this Using Millrsquos own terminologyof facts let me introduce the idea of an explanatory tree Explanatorytrees look like this

Each of the facts e h k n q t on this tree is explained by those factsto the left of it to which it is connected by an arrow (lsquothe arrow ofexplanationrsquo) The facts that explain might be universally general facts(Millrsquos uniformities like the fact that everything which is F is G) orsingular facts (like the fact that a is F) or existentially general facts(like the fact that there is something which is an F) or stochasticfacts (like the fact that many or most Fs are Gs) facts of identity (likethe fact that a=b) and conjunctions and combinations thereof (thislist is meant to be indicative but not necessarily exhaustive of thekinds of facts that there are) Hempel calls singular facts lsquoparticularfactsrsquo and I sometimes follow him in this I use lsquosingularrsquo andlsquoparticularrsquo interchangeably and mark no distinction by the twoexpressions In general however I prefer lsquosingularrsquo because inclassical logic a particular judgement has the form lsquoSome A is Brsquoand the fact that this expresses is an existentially general fact ratherthan a singular one

There are parallels between causal trees and explanatory trees buteven for Mill they are not the same thing Moreover whatever our viewsabout causation different columns cannot be taken to represent successivetimes Nor is the arrow of explanation the same as the arrow of causationGeneral laws or universally general facts at earlier times do not cause lessgeneral facts or laws at later times More general laws explain less generallaws without causing them and if they explain them they atemporally do

126


so So the tree employs the arrow of explanation not the arrows of timeor of causation although of course both time and causation will figure inthe tree where appropriate

This tree is only a portion of a much larger and more extensive treesince it goes back only to the explanation of e in terms of a b and c Themore extensive tree branches at each point as we travel backwards (to theleft) on it Presumably a b c d f g i and all the other facts abovewhose explanation is not accounted for in the tree fragment there presentedthemselves have explanations

Our question is this do such trees necessarily have initial starting pointson the left To put it another way are there any ultimately inexplicablefacts Millrsquos reply is that there must be such facts which have no possibleexplanation According to Mill there must be some inexplicableuniversally general facts

Derivative laws are such as are deducible fromhellipand mayhellipbe resolvedinto other and more general ones Ultimate laws are those which cannotWe are not sure that any of the uniformities with which we are yetacquainted are ultimate laws but we know that there must be ultimatelaws and that every resolution of a derivative law into more generallaws brings us nearer to them

(Mill 1970318)

Millrsquos view is that there must be an ultimate plurality of laws whichhave no further explanation even though we may be unsure whetherconcerning any particular law it be ultimate or derived We attributedto Plato and Aristotle the idea of self-explaining entities They didnot see how an ultimate inexplicable could in turn explain somethingelse so the ultimate points in the regress of explanation had for themto be self-explaining Since for Mill explanation is only by way ofderivation from something more general (unlike for Aristotle for Millthere is no such thing as lsquonon-demonstrable understandingrsquo) it followsthat for him these ultimate starting points of explanation must beinexplicable rather than self-explaining

I do not claim that I can see the difference between inexplicability andself-explicability but only that the philosophers under discussion seemto see a difference The real difference between ultimate inexplicabilityand self-explicability may be verbal rather than real It is certainly notobvious what of interest follows from one that does not follow from theother13

127


So far we have seen that Mill subscribes to the thesis that there areultimate inexplicable laws He also thinks that there are ultimateinexplicable singular facts concerning the number and distribution of theprimeval causal agents in the universe a topic we have already touched on

Derivative laws therefore do not depend solely on the ultimate lawsinto which they are resolvable they mostly depend on those ultimatelaws and an ultimate fact namely the mode of coexistence of some ofthe component elements of the universe The ultimate laws of causationmight be the same as at present and yet the derivative laws completelydifferent if the causes co-existed in any different proportionshellipthereishellipno uniformity no norma principle or rule perceivable in thedistribution of the primeval natural agents through the universe


So Mill is committed to a double ultimacy of inexplicables (1) thereare laws for which there is in principle no explanation (2) there areparticular brute facts for which there is in principle no explanation

In the last chapter Aristotle presented us with the obvious alternativesto the doctrine of ultimate inexplicability or self-explanation There areat first sight two First the trees might just extend indefinitely or infinitelybackwards with no stopping point We might refer to this as the doctrineof the infinite (or indefinite) regress of explanation Aristotle Plato andMill certainly assumed that such a regress if it existed would be viciousBut would it A regress is said to be vicious if for example in order tohave something there is always an additional something one is firstrequired to have In general in a vicious infinite regress one could neverbe in a position to have anything at all for the requirements for havingthe first or any additional thing could never be met

For example suppose that there is a tree of belief justificationanalogous to my explanation tree Each belief I hold can be justified interms of other beliefs which I do or could hold which themselves can bejustified in terms of still other beliefs which I do or could hold and so onad infinitum Suppose further that for any one of my beliefs to be justifiedI must actually possess a justification for all the beliefs which justify itThis requirement sets in motion a vicious infinite regress since theconsequence would be that no belief could ever be justified

However it isnrsquot just the fact that there is an infinity of justified andjustifying beliefs that makes this regress vicious The viciousness arisesfrom that additional further supposition that no justification of some belief

128


is possible until one actually has a justification for each of the beliefswhich justify it That there could be an infinitely long chain of beliefseach of which lsquoobjectivelyrsquo could justify its successor seems acceptableWhat is unacceptable is the repeated application of the thesis that forsome belief of mine to be justified I must actually be in possession of ajustification for each of the beliefs which justify it

Compare these two theses

(1) All beliefs can be justified in terms of other beliefs ad infinitum (orindefinitely)

(2) For any belief of a personrsquos to be justified he must actually be inpossession of a justification for each of the beliefs which justify it

with the following two theses about explanation

(3) All facts can be explained in terms of other facts ad infinitum (orindefinitely)

(4) For any fact to be able to explain another one must actually have anexplanation for it

(3 amp 4)mdashlike (1 amp 2)mdashinvolves a vicious infinite regress Ifeverything has an explanation and I couldnrsquot actually have anexplanation for anything until I had an explanation of everythingthat I used in the explanation then I could never have anexplanation of anything at all But (3) by itself does not requirethis I may be able to explain f by g even though I have noexplanation of g itself There may be objectively as it were suchan explanation of g in the sense that there is some h such that hwould or could explain g if I knew about h (3) asserts that this isso But I am not prevented from explaining f by g just because Ifail to know the h that explains g

Perhaps it is different with justification It might be plausible to holdthis I cannot justify f by g if I do not actually have a justification for g anunjustified belief cannot itself justify another belief This is controversialBut as far as explanation is concerned unexplained (by me) g can stillexplain f A fact itself unexplained can still explain another (3) by itselfrequires an entirely non-vicious infinite regress of explanation PlatoAristotle and Mill were wrong

The second possibility is that such trees are not really trees at all butin fact are loops If one travels far enough to the left of some (particularor universally general) fact one ends by being on that factrsquos right No

129


fact is inexplicable but sometimes facts are explained by facts which arefar to the right of them on the circular tree

In a tree of explanation as I have imagined it each fact is (part of) theexplanation of the fact immediately to its right to which it is joined by anarrow For example b is part of the explanation of e If e is part of theexplanation of h does it follow that b is part of the explanation of h Is theexplanation relation transitive as Robert Nozick asserts (he also says thatthe explanation relation is irreflexive and asymmetric)14 Intuitions differhere and so I introduce the terms lsquoexplanatory ancestorrsquo and lsquoexplanatorydescendantrsquo If b is part of the explanation of e and e is part of the explanationof h then b is an explanatory ancestor of h h an explanatory descendant ofb The relation of being an explanatory ancestor of is uncontroversiallytransitive whether or not the explanation relation is

Now if b and t figure in an explanatory loop then even though b is anexplanatory ancestor of t t may also be a explanatory ancestor of b Insuch a case b will occur at least twice over as we journey in the explanatoryloop once as an explanatory ancestor of t and once as an explanatorydescendant of t This would mean that the relation of being an explanatoryancestor of is not an asymmetric relation (it might be either symmetric ornon-symmetric depending on further details about the construction ofthese trees) Finally if being an explanatory ancestor of is transitive andsymmetric (or non-symmetric) and if there are these loops the relationwill also be reflexive (or non-reflexive) a fact must (or can) be anexplanatory ancestor of itself

The idea of an explanatory relation being the explanatory ancestor ofthat is transitive but fails to be asymmetric and irreflexive departsdrastically from the way in which we normally think of explanatoryrelations (I return to the question of the asymmetry versus non-symmetryof the explanation relation in chapters VI and VII) We may perhaps dismissthe idea of such explanatory loops altogether However even discountingexplanatory loops and Aristotle and Platorsquos thesis of self-explanationthere seems to be the very plausible idea of a non-vicious infinite regressof explanation Millrsquos conclusion that there must be ultimate inexplicablelaws and singular facts is too swift

Mill on deduction and explanation

I raised right at the beginning of the chapter the question of how wellMillrsquos theory of explanation and his views on deduction cohere His

130


views on deduction are not unambiguously clear in fact severaldifferent theses seem to be conflated into what Mill regarded as asingle thesis I summarize his view below to the extent that is requiredfor my discussion and without paying attention to the other strandsthat make up this ambiguous doctrine15 Nor am I much interested instating whatever if anything is plausible in his views My main aimis to see how one essential epistemic strand in his claims aboutdeduction could fit with his views on explanation

I stated the general point at the beginning of the chapter if deductioncannot advance knowledge it would have been natural for Millrsquos epistemicdown-grading of deduction to have carried over to his views onexplanation since explanation is surely or so anyway one might supposea way of advancing our knowledge But in none of his remarks aboutexplanation does Mill take any account of his general views aboutdeduction or explain how his epistemic down-grading of deduction fitswith his holding a deductivist theory of explanation

To infer Mill tells us is to reason in the widest sense Mill means byan inference lsquoa means of coming to a knowledge of something whichwe did not know beforersquo (Mill 1970120) That something knowledgeof which we acquire in making the inference is the inferencersquosconclusion It is crucial to see that his conception of an inference (lsquoareal inferencersquo) unlike ours is partly epistemic Millrsquos concept of areal inference cannot be explicated just by syntactic or semantic conceptsA real inference moves the inferer from a state of not-knowing the truthof the conclusion of an inference to knowledge of the truth of thatconclusion

Reasoning in the wide sense is commonly (but as we shall see nottruly) said to be of two kinds lsquoreasoning from particulars to generalsand reasoning from generals to particulars the former being calledInduction the latter Ratiocination or Syllogismrsquo (Mill 1970107) Reasonin the narrow sense is ratiocination of which syllogism is the generaltype Mill identifies in these passages deductive reasoning with syllogisticreasoning so that he speaks of all deductive inference as involving apassage from general to particular propositions I will follow him in thisand direct my remarks to syllogistic reasoning

Not all things taken to be inferences on the commonly accepted vieware real inferences and not all real inferences are taken to be such onthe commonly accepted view Mill has no doubt that induction reasoningfrom particulars to the general is a process of real inference in hissense

131


the conclusion in an induction embraces more than is contained in thepremisseshellip In every induction we proceed from truths which we knewto truths which we did not know from facts certified by observation tofacts which we have not observedhellip Induction then is a real processof Reasoning or Inference

(Mill 1970108)

Millrsquos assertion is that in inductive inference we move from knowntruths to truths hitherto unknown And the same will be true for whatMill calls lsquothe third species of reasoningrsquo unrecognized by thecommon view in which we move from particular truth to particulartruth without the aid of general propositions at all

Mill does not think that syllogistic reasoning can be a process of realinference in this same sense whatever the common view of the mattermight be

we have now to inquire whether the syllogistic process that of reasoningfrom generals to particulars is or is not a process of inference aprocess from the known to the unknown a means of coming to aknowledge of something which we did not know before

(Mill 1970120)

Mill argues that since a syllogistic inference would be invalid if therewere anything more in the conclusion than what was in the premisseslsquonothing ever was or can be proved by syllogism which was notknown or assumed to be known beforersquo (Mill 1970120) For Millevery syllogism if considered as an attempt to gain knowledge thatthe conclusion is true on the basis of knowledge of the premissesmust be a petitio principii No one could know that the premisses ofa syllogistic inference were true unless they already knew that theconclusion was true The idea of a syllogistic argument advancingknowledge about the truth of the conclusion is according to himepistemologically circular

I abbreviate lsquoknows that prsquo as lsquoK(p)rsquo and use the sign for logicalentailment In what follows lsquoxrsquo is an unbound variable lsquoarsquo is not a variablebut a name or definite description of an object The principle on whichMill relies stated for one type of syllogistic inference only seems to be

(1) [(lsquoAll F are Grsquo amp lsquoa is Frsquo lsquoa is Grsquo) amp K(all F are G amp a is F)] K(a is G)

132


In ordinary language someone who does not know that a is G couldnot come to know that a is G by deducing it from his knowledge thatall F are G and that a is F because if he fails to know that a is G andif there is this entailment then he must also fail to know either thatall F are G or that a is F

There is not much to be said in favour of Millrsquos principle (1) on hisprinciple there could be no such thing as a surprising conclusion of asyllogistic inference yet it is clear that some such inferences areinformative (even if it is difficult to see this in my specific example) Onemight try to defend (1) by arguing that if one does know that all F are Gand that a is an F one does know willy-nilly that a is G despite protests tothe contrary on the part of the person Itrsquos just that a person might not beaware of the fact that he knows that a is G and it may indeed come as asurprise to him that this is what he knows He knows but he doesnrsquotknow that he knows I shall not pursue the possibility of defending Millwith this sort of lsquoexternalistrsquo account There is a sense in which the defencewould not be of any help to him Deduction would on that defence be asort of real inference after all by deduction one could advance from astate of not knowing that one knew that a is G to a state of knowing thatone knew that a is G

How could one know that all F are G and that a is F and not know thata is G If one does not know that a which (as one knows) is F is G howis it possible to know that everything which is F is G The answer is thatone could know the generalization via some other route than via the onewhich goes through a which is F being G One might know lsquoAll F are Grsquoby deduction from a higher level principle or by induction from manyother cases of F-ish things being G but not including the case of a AsJohn Skorupski says

the conclusion [that syllogistic inference is epistemically circular]only follows if one assumes that any process of reasoning which canraise my confidence in the proposition that all men are mortal has toinclude a specific and separate assessment of the probability thatSocrates [my lsquoarsquo] is mortal Suppose on the other hand that there isa sound method of reasoning which can rationally raise my confidencethat all men are mortal without requiring me to consider the particularcase of Socrates Then from the general proposition together withmy knowledge that Socrates is a man I can infer that Socrates ismortal and thus without circularity I become more confident ofSocratesrsquo mortality There obviously is such a method of reasoningmdash

133


induction I can argue inductively from cases other than that ofSocrates to lsquoAll men are mortalrsquo and hence from lsquoSocrates is a manrsquoto lsquoSocrates is mortalrsquo16

Skorupskirsquos argument against Mill is expressed in terms of degree ofrational confidence mine (like Millrsquos) in terms of knowledge Perhapsone might try to reply on Millrsquos behalf by taking advantage of thisdistinction One can have it might be said a high degree of rationalconfidence in all F being G based on lots of Fs being G but withouthaving considered the case of a (I may know that a is F but simplyhave not considered whether a is G) But how the Millian mightcontinue could I know that all F are G and know that a is F withouthaving considered the case of arsquos G-ness

The answer to this question surely is that the criteria even for knowingthat a generalization is true cannot be set this high I can know that ageneralization is true without having per impossibile separately andspecifically assessed each instance of it If I could not know ageneralization in this way all generalizations except those grounded in aperfect induction would be unknowable

There is a principle worth considering that is more plausible than theone to which Mill actually commits himself

(2) [K(lsquoAll F are Grsquo amp lsquoa is Frsquo lsquoa is Grsquo) ampK(all F are G amp a is F)] K(a is G)

Suppose one knows as before that all F are G and that a is F butnow also knows that these two beliefs entail that a is G It isnrsquot justthat one knows that sentences with the form lsquox is Grsquo follow fromlsquoAll F are Grsquo and a sentence with the form lsquox is Frsquo Rather oneknows concerning a that lsquoa is Grsquo follows from lsquoAll F are Grsquo and lsquoais Frsquo One is not in the dark about the fact that arsquos G-ness followsfrom arsquos F-ness In a sense one has considered arsquos status (the factthat arsquos G-ness follows from arsquos F-ness) although not whether infact a is G

Does it then follow that one knows that a is G One might try arguingthat it does on the grounds that knowledge is closed under known logicalentailment One could then try to reconstruct Millrsquos view on syllogisticinference using this more plausible principle Robert Nozick rejectsthe principle that knowledge is closed under known logical entailmentso any defence of (2) would have to take account of his argument17

134


Mill had a deep appreciation of the triumphs of natural science Howone might wonder could his view of deductive inference be compatiblewith his knowledge of scientific advance His answer must be that scienceadvances by induction and inference from particular-to-particular It ispossible to summarize or describe the advance in a general way by use ofdeductive inference But the advance of science itself cannot be viadeductive inference at all

How if at all is Millrsquos deductivist account of explanation consistentwith his view of deduction The answer is contained in the aboveexplaining is not advancing Recall that Mill said lsquoAs already remarkedthe same deductive process which proves a law or fact of causation ifunknown serves to explain it when knownrsquo (Mill 1970310) Predictionand explanation are to be distinguished by the fact that they ask thefollowing distinct questions lsquoGiven a certain combination of causes whateffect will follow and What combination of causes if it existed wouldproduce a given effectrsquo (p 303) That the conclusion of an explanatoryargument is true is what one already knows before the explanation isproduced So there is no question of coming to know on the basis of theexplanatory deduction that the conclusion is true On Millrsquos viewexplanation does not involve gaining new knowledge about the truth ofthe conclusion of an explanatory argument

Explanations are surprising Whence comes the surprise in anexplanation An explanation must teach something new that was notknown before Of course the premisses of a deduction typically containmore information than what the conclusion by itself asserts In explainingan explanandum I might learn a great deal of this new information Imight be surprised to learn that the major premiss (all F are G) or theminor premiss (a is F) of the explanatory argument is true (On Millrsquosview knowing that all F are G is just to know that there is a real inferencefrom any sentence with the form lsquox is Frsquo to a sentence with the form lsquoxis Grsquo18)

But somehow that canrsquot be all there is to the surprise I might explainthe fact that a is G already knowing that a is F and all Fs are Gs After allI can actually set out the explanation as a deductive argument only whenI already know all the premisses and the conclusion What then doesputting the explanation in the form of a deductive argument do when itdoes not surprise me about the truth of any premiss or conclusion Why(to put the same question differently) must the law that all Fs are Gs beincluded in the explanation at all What motive could Mill have had forbeing a deductivist about explanation Mill does discuss the rationale for

135


setting out sciences in a deductive form (Mill 1970141ndash7) but theconsiderations he there adduces will not help us to understand what isgained by insisting that explanations are deductive arguments with at leastone law premiss

The problem here runs very deep One can sense how deep byconsidering Millrsquos views on the third species of reasoning which he saysis the ground for both deduction and induction immediate inference froma particular proposition to a particular proposition without the aid ofany generalization at all

If from our experience of John Thomas etc who once were livingbut are now dead we are entitled to conclude that all human beingsare mortal we might surely without any logical inconsequence haveconcluded at once from these instances that the Duke of Wellington ismortal The mortality of John Thomas and others is after all thewhole evidence we have for the mortality of the Duke of WellingtonNot one iota is added to the proof by interpolating a general propositionSince the individual cases are all the evidence we can possess evidencewhich no logical form into which we choose to throw it can makegreater than it is and since that evidence is either sufficient in itselfor if insufficient for the one purpose cannot be sufficient for the otherI am unable to see why we should be forbidden to take the shortest cutfrom these sufficient premisses to the conclusion and constrained totravel the lsquohigh priori roadrsquo by the arbitrary fiat of logicians


If we take seriously Millrsquos remarks on the conceptual priority of thistype of reasoning that goes from particulars to a further particularwe might again wonder why he holds a deductivist account ofexplanation Why to paraphrase his remarks above on inference canrsquotwe explain arsquos being a G on the basis of arsquos being an F and the factthat b c d etc which were F were also G What does a generalizationdo in a deductive explanation that could not just as well be done by anon-deductive lsquoexplanationrsquo from particular propositions to aparticular proposition without lsquointerpolating a general propositionrsquoDoes it just serve as a reminder about other explanations we wouldbe prepared to offer in the way in which Mill says that a deductionserves as a register or reminder about other real inferences we areprepared to make If that were the answer then although we could

136


sometimes give deductive explanations there ought to be a categoryof real explanation from particulars to particulars to parallel thecategory of real inference from particulars to particulars But Milloffers no such doctrine

I think there is a good answer to the question of why Mill retaineddeductive explanation and did not espouse a view of real explanationparallel to real inference It is true that in setting out an explanation indeductive form one might learn nothing new about the truth of any premissor of the conclusion The conclusion lsquoa is Grsquo I already knew was truethe premisses I either knew already or learned in order to be in a positionto set out the argument But there is something else that one learns aboutthe fact that a is G in setting out the deduction other than its truth namelyhow arsquos being G fits into the overall pattern of nature An analogy heremight be with a jigsaw puzzle One already has all the pieces what onelacks is the ability to fit them all together An explanatory deduction israther like a set of directions that show how those pieces of the jigsaw fittogether The directions do not give one any new pieces to the puzzleonly new information about how they mesh into a whole picture Puttingthe pieces together can be surprising One had all the pieces but wassurprised to learn that that is the picture that results when they areassembled

And so it is with deductive explanation No new particular piece ofknowledge of the sort one is likely to find in a premiss or a conclusion ofa syllogism must be gained all that typically will be gained in deductiveexplanation is new information about how all the same old pieces ofknowledge fit together in an overall grasp of what nature is like It isnrsquotnecessary that in explaining why a is G I learn that a is F or that all F areG (although I might learn one or both of these) What I may do is to bringall the information I already possess together to assemble it as it were toform an overall view of how my existing stock of information interrelateshow some pieces of it bear on other pieces And the new knowledge Iacquire about this pattern about these interrelations typically will itselfbe surprising information

Millrsquos view of what explanation does for us quoted earlier bears outthese remarks

The word explanation is here used in its philosophical sense What iscalled explaining one law of nature by another is but substituting onemystery for another and does nothing to render the general course ofnature other than mysterious we can no more assign a why for the

137


most extensive laws than for the partial ones(Mill 1970310)

Mill seems to be distinguishing in this passage between lsquoexplainingrsquoon the one hand and providing an answer to lsquowhyrsquo on the other IfMill had thought of explanation as providing some sort of deep answerto a lsquowhy-questionrsquo he might have accepted a Duhemian scepticismabout explanation But he does not think of explanation in this waywe never really ever know (in the lsquodeeprsquo and demystifying sense)why according to him All that we can do in explaining is to embedthe particular fact or law to be explained in this wider or more generalpattern without ever lessening the deeper mystery of the universeWe represent this fitting of facts into wider patterns by means ofdeductive arguments In so doing we can gain new knowledge of thepatterns in nature itself As Mill said lsquothose who speak of explainingany phenomenon mean (or should mean) pointing out not some morefamiliar but merely some more general phenomenon of which it is apartial exemplificationrsquo (Mill 1970310ndash11) Millrsquos epistemic down-grading of deductive inference fits well with his deductivist accountof explanation because he thought of his theory of explanation as asimilar epistemic down-grading of explanation from its non-empiricist pretensions

Millrsquos doctrines of explanation and deductive inference are both partof his overall project of making certain concepts lsquosafersquo for an empiricistAn empiricist can surely accept explanation as merely registering theincreasingly general patterns or structures into which all the particularsubstantive pieces of our knowledge are to be fitted Deeper insight thanthat if there be such is a mystery happily beyond the reach of soundempiricist philosophy and its acceptable doctrine of what explanation isall about

How could Mill make his view of deduction consistent with makingsurprising predictions This seems to me more difficult Unlikeexplanations predictions seem clearly to advance the predictor to newknowledge about the conclusion Had Mill been willing to accept realpredictions analogous to real inferences he could have explained howscience advances by means of such predictions from particulars toparticulars while retaining his doctrines of the epistemic circularity ofdeductive inference and of the deductive argument form of explanationBut he no more considers real prediction as a possibility than he doesreal explanation In any event such a move on its own would have been

138


inconsistent with the thesis of the symmetry between explanation andprediction

Hempelrsquos account of scientific explanation

We dealt at some length in chapter I with some of the features ofHempelrsquos methodology In this section of chapter IV I will introduceand describe some of the substantive features of his account ofexplanation In many ways Hempelrsquos account is a development andsophistication of what can already be found in Millrsquos theory ofexplanation

Hempel holds that there are lsquotwo basic types of scientific explanationdeductive-nomological [D-N] and inductive-statistical [I-S]helliprsquo19 (In somepassages there is a third deductive-statistical but I ignore that here) Inthis chapter and in chapter VI I will take Hempel to be offering adisjunctive list of conditions that cover the two cases the disjuncts beingjointly necessary and individually sufficient for the concept of the(scientific) explanation of particular events20 Hempel mentions the needfor a further condition which would rule out self-explanation21 This is aproblem that I shall not discuss and none of my criticisms turns on itsomission

Hempelrsquos requirements for a deductive-nomological explanation of aparticular event are these Let lsquoc1 c2 c3hellipcnrsquo be sentenc es describ singularfacts let lsquoL1 L2hellipLmrsquo be universally quantified sentences asserting certainlawlike regularities (these together constitute the explanans) Let lsquoersquo be asentence describing whatever fact is to be explained (the explanandum)The laws and singular facts described by the explanans sentences explainthe fact described by the explanandum sentence iff (1) e is a logicalconsequence of the conjunction of the explanans sentences (2) e doesnot follow from any proper subset of the explanans sentences (3) theexplanans sentences must have empirical content (4) the explananssentences must all be true Little or nothing in the analysis of D-Nexplanation goes beyond what can be found in Millrsquos position althoughof course Hempelrsquos presentation unlike Millrsquos is detailed careful andtechnically sophisticated

As Hempel and Oppenheim (with whom Hempel co-authored the earlyarticle) point out (3) is redundant Since the explanandum fact is anempirical fact and if as (1) requires the explanandum sentence that statesthat fact is derivable from the conjunction of explanans sentences (3) is

139


automatically fulfilled The explanans sentences are bound to have someempirical content in virtue of implying the empirical explanandumsentence itself (1) is trivially true since the kind of explanation beinganalysed by this disjunct of the analysis has been restricted to D-Nexplanations the analysis of non-deductive explanation is to be coveredby the other disjunct of the analysis of scientific explanation It is not anamazing truth that the explanandum sentence is entailed by the conjunctionof the explanans sentences in a (full) D-N explanation

The history of (4) discussed by Lyon is rather strange22 In the earlyarticle and again in the later article the notion of a potential explanationis introduced similar to the one I used in my discussion of Aristotle wholike them required that the premisses of an explanatory argument be trueBut in a postscript to the early article added in 1964 Hempel says that thefourth requirement characterizes what might be called a correct or trueexplanation In an analysis of the logical structure of explanatoryarguments he says (4) may be disregarded (Hempel 1965249)

For Hempel what sorts of entities are explanantia and explananda inthe non-sentence sense What is the ontology of explanation Mill as wesaw was less than clear about this Hempel speaks of lsquoexplaining aphenomenonrsquo (Hempel 1965246) or lsquowhy the phenomenon occurredrsquo(p246 and p337) but elsewhere he tells us that an explanation explainslsquoin virtue of certain explanatory factsrsquo which facts include lsquosingular factsrsquoand uniformities (p336) But then he also says that lsquothe objectofhellipexplanation in every branch of empirical science is always theoccurrence of an event of a certain kindhellipat a given place and timersquo (p233)

Hempel gives an extended discussion of this question in his later articleand it is this which I take to be his considered view

helliponly when understood in this sense as fully describable by meansof sentences can particular facts or events be amenable to scientificexplanationhellipBut the notion of an individual or particular event is often construed inquite a different manner An event in this second sense is specifiednot by means of a sentence describing it but by means of a noun phrasesuch as an individual name or definite description as for examplelsquothe first solar eclipse of the twentieth centuryrsquo lsquothe eruption of MtVesuvius in AD 79rsquo lsquothe assassination of Leon Trotskyrsquo lsquothe stockmarket crash of 1929rsquo For want of a better terminology individualevents thus understood will be referred to as concrete events and factsand events in the first sense here considered will be called sententially

140


characterizable or briefly sentential facts and eventshellipIn sum a request for an explanation can be significantly made onlyconcerning what we have called sentential facts and events only withrespect to them can we raise a question of the form lsquowhy is it the casethat prsquo As for concrete events let us note that what we have calledtheir aspects or characteristics are all of them describable by means ofsentences each of these aspects then is a sentential fact or event (thatthe eruption of Mt Vesuvius in AD 79 lasted for so many hourshellip)It would be incorrect to summarize this point by saying that the objectof explanation is always a kind of event rather than an individualeventhellipWhat might in fact be explained is rather the occurrence of aparticular instance of a given kind of eventhellipAnd what is thus explainedis definitely an individual event indeed it is one that is unique andunrepeatable in view of the temporal location assigned to it But it isan individual sentential event of coursehellip


So in sum I think we should read Hempelrsquos previous more randomremarks in the light of this long quotation Hempelrsquos theory ofexplanation is only a theory of explanation for lsquosentential factsand eventsrsquo never for singular or concrete events To simplify Ishall attribute to Hempel an ontology of explanation that utilizesfacts but I shall not discuss the rationale for so doing until thenext chapter

Mill as we saw generally down-graded uniformities of coexistenceHempel in the early article asserted that D-N explanation was lsquocausalexplanationrsquo (Hempel 1965250) and he does make it clear especiallyin the later article (pp347ndash52) that he takes a causal law to be a law ofthe succession of phenomena Hence (although he does not say soexplicitly) it is a notion available to an empiricist philosophy ofexplanation But although he like Mill believes that causation is safefor an empiricist account of explanation he is clear on what Mill seemedto waver about there are for Hempel non-causal explanations ofparticular events On his theory non-causal laws of coexistence alsohave an explanatory role to play

In a footnote to the 1964 postscript to the early article he reminds usthat causal explanation is but lsquoone variety of the deductive type ofexplanationrsquo The matter is more fully discussed in the later article Therehe qualifies the claim that explanation of the D-N type is causalexplanation in two ways (Hempel 1965352ndash3) First Hempel reminds

141


us that we explain general laws by subsumption under more generallaws and such explanation is not explanation by causes Hempel echoesMill in this

The laws thus explained or resolved are sometimes said to beaccounted forhellipthere is often a confused notion that the general lawsare the causes of the partial oneshellip But to assert this would be a misuseof the word cause terrestrial gravity is not an effect of generalgravitation but a case of ithellip

(Mill 1970311)

Second Hempel l ike Mill distinguishes between laws ofcoexistence and laws of succession For Hempel as for Mill ahallmark of causation is succession in time if a law is aboutcoexistent phenomena the law cannot be a causal law But Hempelholds that we can sometimes explain a particular occurrence byadducing a law of coexistence and in so far as we do this ourexplanations of particular events cannot all be causal explanationsThe examples he offers are the explanation of the period of apendulum by its length and explanations which utilize the Boyle-Charles gas laws and Ohmrsquos law Hempel holds that we can forexample explain the period of a particular pendulum at time t byits length at t Since this is a case of explanation by laws ofcoexistence Hempel allows non-causal explanations ie onesmaking use of non-causal laws in the explanation of both singularfacts and laws

Hempelrsquos methodology

Many who read Hempelrsquos writings on explanation for the first timeare struck by the apparent arbitrariness of the conditions he lays downin his analysis of explanation On the view we have just describedone type of full explanation deductive-nomological explanation hasthe form of a deductively valid argument one of whose premissesbeing a true lawlike generalization But there seems to be very littleargument in Hempel for this and most other requirements that hementions How could we show that this is a reasonable requirementfor explanation of any type Why should we accept it In terms ofthe terminology of chapter I does Hempel use the language usersrsquo

142


approach or the technical approach in justifying his analysis ofexplanation

With the exception of some remarks on the I-S model of explanation(Hempel 1965391) and an argument against treating laws as rules ofinference (p356) Hempel nowhere appeals even to a vague and ambiguouslanguage use as a support for any of his requirements for explanationHempel explicitly denies that he is writing a dictionary entry lsquoExplicatingthe concept of scientific explanation is not the same as writing an entryon the word ldquoexplainrdquo for the Oxford English Dictionaryrsquo (pp412ndash13)The point of this remark may not be entirely clear since it is hard to thinkof any philosopher who thought of philosophical explication as just thesame thing as dictionary definition But whatever precisely the remarkmeans its thrust seems to be a rejection of the language usersrsquo approach

Hempel asks that his explication be judged by the following constraints

Like any other explication the construal here put forward has to bejustified by appropriate arguments In our case these have to show thatthe proposed construal does justice to such accounts as are generallyagreed to be instances of scientific explanation and that it affords abasis for systematically fruitful logical and methodological analysis ofthe explanatory procedures used in empirical science It is hoped thatthe arguments presented in this essay have achieved that objective


There seem to be two constraints on scientific explanation mentionedin the above quote (1) doing justice to generally agreed instances ofscientific explanation (2) affording a basis for systematically fruitfullogical and methodological analysis of the explanatory proceduresof science As for (1) it is not clear what doing lsquojustice torsquo suchaccounts means since he asserts elsewhere that

these models are not meant to describe how working scientists actuallyformulate their explanatory accounts Their purpose is rather to indicatein reasonably precise terms the logical structure and the rationale ofvarious ways in which empirical science answers explanation-seeking-why-questions

(Hempel 1965412)

So doing justice to agreed instances of scientific explanation doesnrsquotmean doing justice to them as they actually occur but rather doing

143


justice to something that can in some way be elicited from themlsquotheir logical structure and rationalersquo But Hempel gives no clue as tohow we are meant to determine what the logical structure or rationaleof an agreed instance of scientific explanation is independently ofhis own account

As for (2) Hempel never tells us what is meant by lsquosystematicallyfruitful logical and methodological analysis of the explanatory proceduresused in empirical sciencersquo What makes one explication or analysis of aconcept like that of scientific explanation more fruitful than another

I find Hempelrsquos explicit remarks on the methodology of what he isdoing infuriatingly vague and difficult to pin down But his frequent useof terms like lsquoidealrsquo lsquoabstractionrsquo lsquoschematizationrsquo (see Hempel 1965412for example and my remarks on this topic in chapter I) offers the realclue to what he is doing His methodology like his distinction betweenscientific and ordinary explanation that I discussed in chapter I dependsessentially on the distinction between complete and partial explanationand is logically dependent on it It does not matter whether anything wecall an explanation (even those which lsquoare generally agreed to be instancesof scientific explanationrsquo) is a complete explanation or whether anygenuinely complete explanation has ever been given

However explanations when ordinarily given and generally agreed tobe such must have a certain relation to complete explanations as these arespecified in Hempelrsquos models The relevant relation is the relation of beingpartially like (in ways described by Hempel) actual explanations inscience and in ordinary affairs are any kind of explanation at all onlybecause they are partially like the ideal ones Hempel describes Realordinary explanations just about every where and always turn out to beincomplete in some way mere explanatory sketches partial explanationselliptical formulations enthymemes or whatever which fall short of thecriteria for adequate explanation that Hempel lays down23

We have found then that the explanatory accounts actually formulatedin science and in everyday contexts vary greatly in the explicitnesscompleteness and precision with which they specify the explanansand explanandum accordingly they diverge more or less markedlyfrom the idealized and schematized covering-law models But grantingthis I think that all adequate scientific explanations and their everydaycounterparts claim or presuppose at least implicitly the deductive orinductive subsumability of whatever is to be explainedhellip


144


His position is I think to be placed somewhere between a purelytechnical and a purely language usersrsquo approach24

Would this be a helpful analogy Just as ideal English grammar isimplicit in spoken English grammar and the rules of deductive andinductive logic are implicit in the deductive and inductive inferences wedo actually make so too Hempel supposes that complete explanations(as specified by his models) are implicit in the ones we actually give

This purported analogy would be misleading There must be awidespread overlap between how English is actually spoken and how wedo infer on the one hand and how it should ideally be spoken and how weshould ideally infer on the other Methodologically the idea of idealpractice and the idea of actual practice must intersect

And this is an overlap or intersection that Hempel need not assume inthe case of actual explanations and ideally complete ones The purportedanalogy as applied to explanation would give us a language usersrsquoapproach (with room for lsquotidying up the discoursersquo) rather than an approachto the analysis of explanation somewhere between the two approachesHempelrsquos method assumes unlike the analogy that the set of actualexplanations and the set of ideal explanations could be (and indeedprobably are close to being) wholly disjoint non-overlapping As I saidabove it is logically possible that no one has ever actually given an idealcomplete explanation

(In truth surely some actual explanations have been complete idealexplanations in Hempelrsquos sense without any relevant information beingomitted But it must not be forgotten just how hard it is to give a completeexplanation Since its premisses must all be true an ideal explanationmust make use of strict exceptionless laws and we are able to stateprecious few of them except at the highest and most abstract level ofscientific theory Additionally in singular explanations at any rate it isnecessary to assume that one is dealing with a closed system and theseclosure assumptions although they have great heuristic value are rarelytrue25 And so on)

But how do we know that actual explanations are only partial Howdo we know that the ideals Hempel proposes in the light of which actualexplanations are seen to be only partial are appropriate for judging themIn order to answer these sorts of questions I must be in a position todecide what belongs in a complete explanation and what is merelypresupposed by a complete explanation or is the support or ground forthe complete explanation How do I decide whether actual explanationsare partial because they lack whatever an ideal explanation would have

145


or whether actual explanations are complete the extra information in theso-called ideal being merely presupposed by complete explanations butno proper part of them

Hempelrsquos view (and the same could be said for any argument theoryof explanation like Millrsquos) builds what might be thought of as the supportor grounds for the explanation into the complete explanation itselfConsider the analogous question for prediction Suppose I predict that ewill occur on the basis of the occurrence of c and the law that whenevera C an E What is my complete prediction There are two views onemight hold with the content of the predictions put into parenthesis tomake the views unambiguously clear and distinct (1) my completeprediction is that (e will occur) and the basis on which I make theprediction although no part of the complete prediction itself is that c hasoccurred and whenever a C an E (2) my complete prediction is that (ewill occur since c has and whenever a C an E)

Which view would be more plausible In general one cannot includein the content of onersquos warnings criticisms predictions and so on all ofthe bases on which they are made How can we decide what to includeand what not In the case of prediction it seems to me that a completeprediction has the form specified by (1) rather than (2) If completepredictions were like this and if complete explanations were validarguments we would have sufficient reason to deny Hempelrsquos thesis ofthe structural symmetry between predictions and explanations But perhapsexplanations are not typically arguments but rather like predictions canbe singular sentences I return to this question more fully in chapter VI

Hempel on the symmetry thesis

Does Hempel like Mill subscribe to the symmetry thesis (or the thesisof the structural identity of explanation and prediction as Hempelalso calls it) On this question there is a shift in his views In theearly article he says that the difference between explanation andprediction lsquois of a pragmatic characterrsquo

If E is given ie if we know that the phenomenon described by E hasoccurred and a suitable set of statements [of laws and initialconditions]hellipis provided afterwards we speak of an explanation ofthe phenomenon in question If the latter statements are given and E isderived prior to the occurrence of the phenomenon it describes we

146


speak of a prediction It may be said therefore that an explanation ofa particular event is not fully adequate unless its explanans if takenaccount of in time could have served as a basis for predicting theevent in question

(Hempel 1965249)

Hempelrsquos position here is identical to Millrsquos save for the fact thatunlike Mill Hempel confuses the definition by including as arequirement for prediction that E be derived prior to the occurrenceit describes and not just prior to our gaining knowledge of itsoccurrence Hempel thereby excludes the case of retrodiction fromhis analysis The symmetry thesis as stated has two distinct parts (a)every successful explanation is a potential prediction (b) everysuccessful prediction is a potential explanation In the early articleHempel subscribes to both (a) and (b)

Hempel returned to the symmetry thesis in the later article His positionthere is that the first portion of the symmetry thesis (a) is sound but thatthe second portion (b) lsquois indeed open to questionrsquo (Hempel 1965367)or is lsquoan open questionrsquo (p376) The official discussion of this as it relatesto the D-N model occurs on pp 374ndash5 The case is that of Koplik spotssmall whitish spots on the mucous linings of the cheeks which are anearly symptom of measles From the law that the appearance of Koplikspots is always followed by the manifestation of measles and theinformation that a specific patient has Koplik spots one can predict thatthe patient will develop measles but one cannot explain the subsequentmeasles on the basis of the appearance of the Koplik spots

One might wonder why Hempel says only that (b) is an open questionDoesnrsquot the Koplik spots case simply refute (b) Hempel says about thiscase that it

does not constitute a decisive objection against the second subthesisFor the reluctance to regard the appearance of Koplik spots asexplanatory may well reflect doubts as to whether as a matter ofuniversal law those spots are always followed by the latermanifestations of measles Perhaps a local inoculation with a smallamount of measles virus would produce the spots without leading to afull-blown case of the measles If this were so the appearance of thespots would still afford a usually reliable basis for predicting theoccurrence of further symptoms since exceptional conditions of the

147


kind just mentioned would be extremely rare but the generalizationthat Koplik spots are always followed by later symptoms of the measleswould not express a law and thus could not properly support acorresponding D-N explanation

(Hempel 1965375)

Hempelrsquos reply to this objection to (b) seems unacceptable Firsteven if the generalization linking Koplik spots and measles fails tobe an exceptionless law Hempel should accept that Koplik spots canI-S explain measles which they most certainly do not Secondlsquoperhapsrsquo the generalization fails to be a law But perhaps it does notfail to be a law In any case surely there must be some cases in whicha symptom for a phenomenon and the phenomenon itself areconnected in an exceptionless lawlike manner And even if there werenone it is perfectly coherent to imagine a (physically and not just alogically) possible case in which this is so For example supposewhatever may in fact be the case that no local inoculation with asmall amount of the measles virus would produce spots but no measlesFinally it simply is not true that a reluctance to regard Koplik spotsas explanatory of measles reflects onersquos doubts about the exceptionlessnature of the Koplik spots-measles connection Even if I werepsychologically certain that the connection is exceptionless I wouldnot believe that the spots explain the measles

In general in any case in which there is a necessary condition n forsome phenomenon p (measles is a necessary condition for Koplik spotssyphilis is a necessary condition for paresis) one will be able to predict non the basis of p because p will be sufficient for n (one can predict measlesfrom Koplik spots predict syphilis from the presence of paresis) butoften as in these two cases p will not explain n

There are other types of counterexample to (b) the second part of thesymmetry thesis and Hempel deals with some of them in the context of adifferent discussion the discussion of non-causal D-N explanations that Imentioned above These are cases of lsquoreversibilityrsquo in which either therelevant law asserts a biconditional relationship (eg Aristotlersquos law thata planet twinkles iff it is not near) or is a functional law (an equation) thatequates the values of two variables (Both of these sorts of examples canemploy laws either of successive or of coexistent phenomena)

There are many such examples One well-known one is SylvainBrombergerrsquos example of the height of the flagpole the length of itsshadow and the angle of elevation of the sun26 We can both predict

148


and explain the length of the shadow on the basis of the other twofactors and the theory that light travels in straight lines we couldpredict the angle of elevation from the other information but hardlyexplain it

Recall the case in which we explain a pendulumrsquos period by its lengthHempel notes

The law of the simple pendulum makes it possible not only to infer theperiod of a pendulum from its length but also conversely to infer itslength from its period in either case the inference is of the form (D-N) Yet a sentence stating the length of a given pendulum in conjunctionwith the law will be much more readily regarded as explaining thependulumrsquos period than a sentence stating the period in conjunctionwith the law would be considered as explaining the pendulumrsquoslengthhellip In cases such as this the common-sense conception ofexplanation appears to provide no clear grounds on which to decidewhether a given argument that deductively subsumes an occurrenceunder laws is to qualify as an explanation

Hempelrsquos remark about the failure of the common-sense conceptionof explanation to provide grounds for deciding which of the two(length of pendulum by its period period of the pendulum by itslength) is really an explanation is wide of the mark for Hempelrsquoslsquotechnicalrsquo concept of scientific explanation does not do this eitherOne can predict the period of a pendulum on the basis of its lengthand predict its length on the basis of its period If (b) of the symmetrythesis were correct both would also be (potential) explanationsHowever although one can successfully predict length on the basisof period that prediction is not a potential explanation So one of thepredictions is not a potential explanation and hence (b) must bewrong

Ohmrsquos law asserts that the intensity of a constant electrical current ina circuit is directly proportional to the electromotive force and inverselyproportional to the resistance Boylersquos law says that the pressure of afixed mass of gas at a constant temperature is inversely proportional to itsvolume Hookrsquos law claims that the force required to produce a distortionin an elastic object is directly proportional to the amount of distortionThese and similar laws assert a numerical equivalence hence they allowprediction in both directions But in many of these examples of functionallaws we would not allow that explanation can go in both directions We

149


can explain the amount of distortion (elongation) to a steel spring by thequantity of force applied to the spring but not vice versa

So some successful cases of prediction via functional laws and lawscontaining biconditionals are not potential explanations The second partof the symmetry thesis (b) is false

Hempel on inductive-statistical explanation

Hempelrsquos thesis that some explanations have the form of an inductiveargument is as far as I am aware an important addition to theempiricist philosophy of explanation Even in the early articlelsquoStudies in the Logic of Explanationrsquo Hempel and Oppenheim hadindicated the existence of another non-deductive type of explanation(Hempel 1965251 278 and pointed out by Hempel in the 1964postscript 291) For Hempel these are two models of completeexplanation lsquounder a reasonable extension of the idea of explanatorycompleteness any explanation conforming to our statistical modelshould qualify as formally completehelliprsquo (p 418)

The I-S model is introduced thus

Explanations of particular facts or events by means of statistic-probabilistic laws thus present themselves as arguments that areinductive or probabilistic in the sense that the explanans confers uponthe explanandum a more or less high degree of inductive support or oflogical (inductive) probability they will therefore be called inductive-statistical explanations or I-S explanations


Hempelrsquos first example of an I-S explanation is this we explain whyJohn Jones recovered from a streptococcus infection on the groundsthat he had been given penicillin and that it is highly probable that aperson with such an infection who is given penicillin will recoverFor Hempel the basic form of an I-S explanation is this

The first premiss says that the probability of somethingrsquos being Ggiven that it is F is r The second premiss asserts that b is F The

150


double line under the second premiss shows that we are dealing witha non-deductive argument The conclusion that b is G is not madecertain by the premisses but only probable to degree r (less than 1)Hempel never says how high r must be in order for the premisses toexplain the conclusion He says only lsquovery highrsquo and that given thepremisses the conclusion is lsquopractically certainrsquo

Hempel is a probabilist But we saw that a deductivist can accept anon-deductive relation between explanans and an explanandum in a partialexplanation (lsquopartialrsquo here means lsquopart of a full explanationrsquo notnecessarily restricted to the more limited sense of lsquopartialrsquo that Hempeluses) In a simple case we may omit information which we assume thatthe audience is aware of or information which we do not as yet possessSo why arenrsquot all I-S explanations just incomplete D-N explanations

Perhaps this is the answer in an incomplete D-N explanation I omitlsquomention of certain laws or particular facts that are tacitly taken forgranted [or unknown] and whose explicit inclusionhellipwould yield acomplete D-N argumentrsquo (Hempel 1965415) But in a complete I-Sexplanation I omit mention neither of relevant singular facts nor of alaw rather what makes an explanation a complete I-S explanation asopposed to an incomplete D-N explanation is the presence of a statisticalor stochastic law like the law that only a high proportion of those whotake penicillin recover from a streptococcus infection An I-S explanationlsquomakes essential use of at least one law or theoretical principle ofstatistical formrsquo (p 380)

It is not clear why this reply would yield two different models ofexplanation in the way intended by Hempel There is of course thedistinction that the above paragraph draws but what is not clear is whyanything of importance hangs on it Why should it matter if the informationomitted from the explanation from ignorance or from other pragmaticconcerns is reflected in the omission of particular matters of fact or intotal omission of a law rather than in an incomplete statement of a lawWhy should that distinction be important enough to ground a distinctionbetween two types of explanation Why not these two types of explanationinstead ones that omit some relevant particular fact and those whichinclude all relevant particular facts

If we lived in a deterministic world stochastic laws would be merelyincomplete statements of deterministic laws In that case it is hard to seewhy the distinction mentioned above would matter The same ignoranceor voluntary omission of pragmatically-irrelevant features that mightmanifest itself in the omission of the law or of particular facts from an

151


explanation might just as well manifest itself in an incomplete statementof the law In a deterministic world there does not seem to me to be anyimportant difference between an incomplete D-N explanation and an I-Sexplanation for every I-S explanation is really only a partial D-Nexplanation27

But of course we do not or might not live in a deterministic worldwith strictly universal laws In that case statistical laws are not justincomplete statements of deterministic laws Hempel ultimately and Ibelieve correctly ties the distinction between incomplete D-Nexplanations and complete I-S explanations to the question ofnondeterminism

It seems inadvisable to construct an extended concept of explanatorycompleteness in such a way as to qualify all statistical explanations asincomplete For this qualification carries with it connotations ofdeficiencyhellipthe early explanatory uses of statistical laws andtheorieshellipwere often propounded in the belief that themicrophenomenahellipwere all subject to strictly universal lawshellipBut thisidea has gradually been abandonedhellip


Hempelrsquos later I-S examples of the explanation by means ofMendelian genetic principles of the distribution of red and white peaplants resulting from a cross of pure whites and pure reds of theexplanation of the radioactive decay of radon and the explanation ofcertain aspects of the Brownian movement of small particlessuspended in a liquid all offer what are perhaps more serious examplesof I-S explanation because they are or may well be explanations ofgenuinely nondeterministic phenomena

Hempel says that lsquostatistical explanation is quite independent of theassumption of strictly universal lawsrsquo (Hempel 1965418)28 I would gofurther in keeping with claims I made in chapter I Even in a deterministicworld there will be statistical explanations but they will only be a ratherspecial and interesting kind of incomplete D-N explanation But the ideaof a complete I-S explanation is not independent of the assumption ofstrictly universal laws such an idea presupposes that some laws are notstrictly universal The concept of a complete I-S explanation (or its non-argument analogues) needs a metaphysical backing and only some formof nondeterminism and hence a high or a low dependency theory ofexplanation can supply the rationale for it

152


Hempel on epistemic ambiguity

Hempel spends some time discussing the problem of the epistemicambiguity of I-S explanations (Hempel 1965394ndash405) My remarksin the preceding section bear closely on this difficulty in Hempelrsquosaccount

It is not possible to have a deductive argument such that if the premissesare all true and imply some conclusion c those same premisses with theaddition of any further true premiss imply -c But with inductive argumentthis is possible Suppose I know that John has a streptococcal infectionand has been given penicillin (F) so I conclude with a probability of r(which makes me practically certain) that John will recover (G) If laterI learn that John is an octogenarian with a weak heart (H=has streptococcalinfection and has been given penicillin and is an octogenarian with aweak heart) I may revise my probability estimate and indeed mayconclude that it is practically certain that he will not recover (-G) Theabove argument could be represented as follows

The difficulty is that two different inductive arguments both withtrue premisses can inductively support to a high degree and (if wehad no further way to rule this out) therefore explain twocontradictory conclusions

As Hempel says lsquoThe preceding considerations show that the conceptof statistical explanation for particular events is essentially relative to agiven knowledge situationhelliprsquo (Hempel 1965402) Relative to theknowledge of Johnrsquos infection and receipt of penicillin we are entitled todraw one conclusion if the knowledge situation changes and we acquireadditional relevant information about his age and heart condition it maybe that we are entitled to draw the opposite conclusion

Hempel therefore imposes what he calls the lsquorequirement of maximalspecificityrsquo29 Put informally

The general ideahellipcomes to this In formulating or appraising an I-Sexplanation we should take into account all that informationhellipwhichis of potential explanatory relevance to the explanandum event ie

153


all pertinent statistical laws and such particular facts as might beconnected by the statistical laws with the explanandum event


Hempel also repeats what he takes to be a different formulation ofthe same idea that the individual must be referred lsquoto the narrowestreference class of which according to our total information theparticular occurrence under consideration is a memberrsquo (Hempel1965398) Of course many of the reference classes to which Johncan be assigned are irrelevant he might be a member of the classoctogenarian with a weak heart with streptococcal infection givenpenicillin and wears a straw hat That class is narrower still than theclass of Hs but the additional information that accounts for its beingnarrower is statistically irrelevant to the question of Johnrsquos recoveryWhat Hempel wants is something like lsquothe reference class no partitionof which is known to be statistically relevantrsquo Hempel claims thatthis solution lsquodisposes of the problem of epistemic ambiguityrsquo sinceof two rival inductive arguments both of which confer high probabilityon their conclusion at least one must violate the requirement ofmaximum specificity

In any probability argument appropriate choice of reference classconstitutes a problem and on this there is a vast literature What issurprising in this and here I follow Coffa30 is that Hempel has chosen toput the resolution of this difficulty in the way he does The distance betweenD-N arguments and I-S arguments is made very great by this manoeuvreFor Hempel the notion of an objective I-S explanation apart from theknowledge situation in which we find ourselves makes no sense WesleySalmon and others have stressed that there is an alternative way to selectthe appropriate reference class for an I-S explanation the correct referenceclass is one that is objectively homogeneous no further statisticallyrelevant partition of the class being objectively and not just epistemicallypossible In a deterministic world explanations which use objectivelyhomogeneous reference classes will be D-N explanations explanationswhich use unhomogeneous reference classes will be I-S explanationswhich are merely the epistemically available parts of complete D-Nexplanations It is only in a nondeterministic world in which there will beI-S explanations which use objectively homogeneous reference classes

Hempelrsquos thoughts on this subject are a matter of some speculationfor he is not very forthcoming In this speculation I follow Coffa If every

154


I-S explanation must be relativized to an epistemic context then for somereason Hempel must be asserting that no reference class mentioned in anI-S explanation could be objectively homogeneous31 If such referenceclasses are not objectively homogeneous then there must be in principlesome further partition possible that is statistically relevant in explainingthe explanandum The suggestion would then be that only reference classesmentioned in D-N explanations are objectively homogeneous all I-Sexplanations employing as they do only unhomogeneous referenceclasses must be the epistemically available parts of D-N explanations I-S explanations reflect a gap in our knowledge but not the objectivegappiness of the world Although admitting I-S explanations as a modeland his remarks on pp 417ndash18 notwithstanding in which he accepts theidea of indeterminism (or nondeterminism) Hempel has not fully takenon board the thought that the world might be objectively nondeterministicand that in such a world I-S explanations will have a role that cannot beplayed by D-N explanations however much we may know and be preparedto say about what goes on in that world

Summary

I have in this chapter in the main raised questions rather thananswered them In particular I have asked questions about the formof and the metaphysical lsquobackingrsquo for explanations Hempel isclearest about these matters all complete or full explanations aredeductively sound or inductively good arguments some singularexplanations are non-causal explanations Mill was less decisive aboutboth of these matters Given his remarks on real inference it is not atall obvious why he did not introduce a category of real explanationof particulars by particulars which would not require the inclusionof laws or generalizations in the explanans Are all singularexplanations causal Are explanations always or typically argumentsDo all full or complete explanations include laws I return to thesequestions in chapters VI and VII

155

CHAPTER V

The Ontology of Explanation

Explanation and epistemology

Epistemology and metaphysics come together to give us ourconception of explanation I have tried in the previous chapters toshow how metaphysical commitments make a difference to a view ofexplanation and in the next chapter I return to these metaphysicalissues In this chapter despite its apparently lsquometaphysicalrsquo title Ibring out some of the ways in which epistemological (in the broadestsense) considerations play their part This chapter returns to a themethat arose in the discussion of Aristotlersquos distinction between the perse and the incidental cause of something Polyclitus might beexplanatory of the statue when described in one way (as the sculptor)but not when described in another (as the musical or the pale man)

Peter Strawson following Hume draws a similar distinction when hespeaks of natural and non-natural relations

hellipcausality is a natural relation that holds in the world betweenparticular events or circumstances just as the relation of temporalsuccession does or that of spatial proximity We also and rightlyassociate causality with explanation But if causality is a relation whichholds in the natural world explanation is a different matter hellipit is nota natural relation in the sense in which we perhaps think of causalityas a natural relation It is an intellectual or rational or intensionalrelation It does not hold between things in the natural world things towhich we can assign places and times It holds between facts or truths1

156


I think of natural relations as falling within the province ofmetaphysics intensional relations as falling at least in part withinthe province of epistemology2 The purpose of this chapter is to workthrough the issues raised so succinctly in the quotation from Strawson

Extensionality and the slingshot

What sorts of entities stand in the explanation relation Of coursewe sometimes explain laws and sometimes or perhaps even alwaysuse laws when we (fully) explain I ignore until chapter VI the placeof laws in explanation But what other entities stand in this relationother than laws It is people who explain explananda by explanantiabut I shall simplify by treating lsquoexplainsrsquo as expressing a dyadicrelation even if it is really a triadic one since we are here uninterestedin the ontology of persons

Recall from chapter I that I following Hempel have limited the scopeof explanation for which an analysis is being sought to cases of explanationthat None of the theories about the relata of the explanation relation thatI shall discuss have anything whatever to say about such cases asexplaining how to ride a bike explaining how the two men shook handsexplaining where the Wash is The cases to which the ontologicalalternatives I shall discuss address themselves arise most readily in casesof explaining why although as I also have made clear we cannotdistinguish the cases we want simply by that grammatical feature alone

Our discussions of Plato Aristotle Mill and Hempel have thrown upnumerous candidates as the relata of the explanation relation Formscauses per se facts phenomena concrete events sentential events anordered pair consisting of an event and a particular description of it aresome of the most obvious To this list we might add statements andpropositions (these are not the same as sentences) Nor should we assumethat only one of the contenders can win the contest perhaps all of themcan be the relata of the explanation relation The choice should not beassumed without further argument to be exclusive

In the course of this chapter I must say certain things about (token)events What are events and in particular what is the criterion forindividuating and identifying events On this topic I should here like tosay as little as possible In chapter II I have already dodged the issue ofwhether events were to be taken as wide or narrow In this chapter Iassume without argument that token event identity does not require the

157


identity of the properties used in the definite descriptions of the eventThat is I accept the view that the event e say orsquos being P at t can beidentical with event f say orsquos being Q at t even though the property P the property Q I note that this view is controversial and that what I haveto say in what follows depends on this view of event identity I call thislsquothe rough-grained criterion of event identityrsquo

In what follows questions of extensionality will arise Let me start byintroducing some terminology with which to discuss questions ofextensionality Extensionality is a single idea but the question ofextensionality arises for different expressions of a language As Haackputs it

A context is extensional if co-referential expressionsmdashsingular termswith the same denotation predicates with the same extension orsentences with the same truth-valuemdashare substitutable within it withoutchanging the truth-value of the whole lsquosalva veritatersquo ie if Leibnizrsquolaw holds for it otherwise it is intensional3

In what follows I use lsquointensionalrsquo and lsquonon-extensionalrsquosynonymously

Letrsquos call the first sort of truth-preserving substitutability of singularterms with the same denotation lsquotransparencyrsquo the second sortsubstitutability of predicates with the same extension lsquopredicateextensionalityrsquo the third sort substitutability of sentences with the sametruth-value lsquosentence extensionalityrsquo (or lsquotruth-functionalityrsquo)

John listens to his favourite nature programme and on it there is ananimal which as John is told has a heart John does not know that thatanimal is the Queenrsquos oldest corgi and being a biological ignoramusdoes not know that all and only animals with hearts have kidneys Considerthe true sentence lsquoJohn believes that the animal on his favourite natureprogramme has a heartrsquo The context following lsquoJohn believes thathelliprsquo isnot transparent because lsquothe Queenrsquos oldest corgirsquo cannot be substitutedfor lsquothe animal on his favourite nature programmersquo salva veritate Nor isthat context predicate or sentence extensional since the coextensivepredicate lsquohas a kidneyrsquo cannot be substituted salva veritate for lsquohas aheartrsquo and a sentence with the same truth-value eg lsquoJohn is ignorantabout biologyrsquo cannot be substituted salva veritate for lsquothe animal on hisfavourite nature programme has a heartrsquo

There is a fourth sort of extensionality that we shall need Considertwo predicates lsquoPrsquo and lsquoQrsquo which express or stand for the same property

158


(whatever the criterion of property identity the reader prefers no doubtweaker than synonymy but surely stronger than coextensionality) Callany pair of such predicates lsquoco-typical predicatesrsquo A context is co-typicalpredicate extensional iff there is substitutability salva veritate of predicateswhich are co-typical The idea is that lsquoPrsquo and lsquoQrsquo are co-typical if (andonly if) they stand for the same property and if there is any plausiblecriterion of property identity weaker than synonymy this will not be thesame as lsquoPrsquo and lsquoQrsquo having the same meaning or intension

The property of being a triangle=the property of being a three-sidedclosed plane figure Consider the sentence lsquoJohn believes that the figurebefore him is a trianglersquo The context following lsquoJohn believes thathelliprsquo isco-typical predicate extensional iff it follows that John believes that thefigure before him is a three-sided closed plane figure (views about whetheror not this is so may differ)

It is important to see that as far as my argument is concerned lsquoPrsquo andlsquoQrsquo may stand for the same property whether or not lsquoP=Qrsquo is necessaryor contingent a priori or a posteriori The reader is free to plug in hisfavourite views here Although the specific example I offered above isone that is a priori and necessary I do not mean to suggest that otherexamples must have the same epistemic status Perhaps if lsquoPrsquo and lsquoQrsquo areco-typical predicates they have the same extensions in all possible worldsbut perhaps not I see no reason to become involved in disputes about theepistemic status of statements of property identity I do assume later inthe chapter that there are some examples which are a posteriori butwhatever else the reader is inclined to believe about these matters is asfar as I can see consistent with what I wish to say

There is a well-known argument sometimes called lsquothe slingshotrsquowhich purports to show that if a context is transparent (and if there issubstitutability of logical equivalents salva veritate) then it is truth-functional4 Here is one version of that argument which I repeat almostverbatim from the account by John Mackie Let lsquoprsquo represent a sentenceand lsquoF(p)rsquo a sentence containing the sentence represented by lsquoprsquo Furtherwe suppose (a) that logical equivalents are interchangeable in lsquoF(hellip)rsquosalva veritate (b) and that lsquoF(hellip)rsquo is transparent if lsquoprsquo is

Now consider the class of xrsquos such that both (x=x) amp p If p is true thisclass will be the universal class if p is false it will be the empty classFurther consider this statement the class of xrsquos such that both x=x amp p isidentical with the class of xrsquos such that x=x That statement will be logicallyequivalent to lsquoprsquo because whatever lsquoprsquo may be it is true when lsquoprsquo is trueand false when lsquoprsquo is false Finally if lsquoqrsquo has the same truth-value as lsquoprsquo

159


then the class of xrsquos such that x=x amp p will be the same class as the classof xrsquos such that x=x amp q since when lsquoprsquo and lsquoqrsquo are both true each willbe the universal class and when lsquoprsquo and lsquoqrsquo are both false each will bethe empty class

The argument now goes as follows

(1) p=q (by assumption)(2) F(p) (by assumption)(3) F(the class of xrsquos such that x=x amp (by substitution of

p is identical to the class of xrsquos logical equivalents insuch that x=x) lsquoF(hellip)rsquo)

(4) F(the class of xrsquos such that x=x amp (from (3)substitutionq is identical to the class of xrsquos of co-referring terms)such that x=x)

(5) F(q) (from (4) by substitu-tion of logical equiva-lents)

(6) F(p) F(q) (from (2) amp (5) by co-nditional proof)

(7) F(q) F(p) (by a similar series ofsteps)

(8) F(p) F(q) (from (6) amp (7))(9) p q F(p) F(q) (from (1) amp (8) by co-

nditional proof)As Mackie says

hellip(9) says that the context lsquoF(hellip)rsquo is truth-functional and this hasbeen proved from the suppositions that lsquoF(hellip)rsquo is such as to allowsubstitution within it of logical equivalents and of co-referringexpressions That is if lsquoF(hellip)rsquo is both transparent and allowssubstitution of logical equivalents it is truth-functional

(Mackie 1974250ndash1)

Although there are various possible objections to this argument to befound in the literature I shall not pursue them here I shall accept theargument without discussion In particular I accept without argumentthat the principle of the substitutability salva veritate of logicalequivalents holds for the contexts under discussion although this mustbe the Achillesrsquo heel of the slingshot if there is one at all I do assume

160


though that the argument only shows that if a context is transparentwith reference to definite descriptions then it is truth-functional theargument does not go through for names5

The relata of the explanation relation

There are three serious lsquocandidatesrsquo for being the relata of theexplanation relation6 The first is events the second facts I havelittle to say about the third candidate statements (and propositionsno distinction between the two being required for the purposes ofthis discussion) Facts and events are the only two candidates whosecase I consider in some detail Sentences are not such a candidatefor they do not explain and are not explained in the conceptuallyprimary sense But statements remain as a candidate even if sentencesdo not so that I shall have to say something about them (Butremember that even if it were statements that explain statements itcould only be so in virtue of certain natural relations holding betweenthe worldly things such statements are about) Since I will argue thatevents tout court never explain or are explained I do not need to dealseparately with the mixed possibilities on which facts mightsometimes explain events or events sometimes explain facts7

If events are even sometimes the relata of the explanation relationthen at least sometimes explanation is by a particular event tout court ofa particular event tout court Or alternatively explanation is at leastsometimes of the occurrence of one particular event by the occurrence ofanother These come to the same there being no difference as far as I cansee between explaining the fire at 10 Downing Street and explaining theoccurrence of the fire at 10 Downing Street This proposal for the relataof the explanation relation can be somewhat broadened to include statesin addition to events to cover the sort of case in which we might want tosay that some unchange to borrow Ducassersquos apt expression explains oris explained For simplicity I shall continue speaking only of eventsexplaining events leaving it to the reader to understand that other worldlyparticulars or individual chunks of reality like states or unchanges aswell as events might be included

Two writers who accept that there are some explanations of events byevents are David Lewis8 and James Woodward9 David Lewisrsquos theorythat all explanation is causal explanation is restricted to the explanationof particular events in reply to an alleged counterexample he asserts that

161


lsquoI donrsquot agree that any particular event has been non-causally explainedrsquo(Lewis 1986223) So there must be at least some explanations of particularevents if Lewis is right whether or not he is right in thinking that allexplanations of particular events are causal

James Woodward in two very interesting articles argues that there isa sense of explanation lsquohellipcausally explainshelliprsquo which is transparent Inthis transparent sense the explanation relation does hold between eventsapart from how they are described lsquoIf what is explained is genuinely anindividual occurrence then the singular term occurring in the effectposition will function purely referentiallyhelliprsquo (Woodward 1986279)Woodwardrsquos view is supported by the distinction he draws between whatan explanation explains and what it presupposes but does not explainlsquohellipthere there is a clear difference between explaining why some particularevent which has a certain feature occurred and explaining why thatevent has that featurersquo To take Woodwardrsquos example if the short circuitcausally explains the fire and if the fire was purple and odd-shaped thenit follows that the short circuit explains the purple odd-shaped fire whichis the same event differently described (Notice that Woodward uses arough-grained criterion of event identity the event of the house being onfire and the event of the house being on purple and odd-shaped fire canbe identical in spite of the fact that two different properties are involvedin the two descriptions)

According to Woodward the expression lsquothe purple odd-shaped firersquofunctions in the above example purely referentially to refer to the fireThe explanation does not explain why the fire was purple and odd-shapedit merely presupposes that the fire has these characteristics and referenceto the fire happens to be fixed via this description Nor argues Woodwardshould we try to escape the transparency point by thinking of theexplanation of the fire as really an explanation of the fact that there wasone and only one event in the vicinity which was a fire rather than anexplanation of the event tout court To explain the former I would haveto explain why no other fires occurred at the same time in the vicinity Inexplaining the fire by the short circuit it is presupposed that there wasonly one such fire but this presupposition is not explained by the shortcircuit

On this view this transparent sense of lsquoexplainsrsquo stands for a relationbetween particular events occurrences things in the world apart fromhow they are described or referred to or conceptualized This relation is anatural relation like that of causation and not a non-natural relation torevert to Strawsonrsquos terminology It is a natural relation in the sense that

162


its relata are lsquonaturalrsquo rather than intensional entities It is consistentwith Woodwardrsquos view that there is or may be another explanationrelation to which lsquoexplainsrsquo in a second non-transparent sense refersand which is a non-natural relation But at least for the transparent senseof that term there is no need to import facts or statements or truths ordescriptions or anything else into the analysis of explanation In explainingthe fire by the short circuit I explain the occurrence of an event in theworld apart from any linguistic or cognitive considerations about how itis talked about or conceptualized

The difficulty I have with the Woodward-Lewis view of eventexplanation is that no explanation sentences are fully transparent in spiteof the sorts of considerations that Woodward adduces I remind the readerthat the topic under discussion is that of explanation which includes causalexplanation but is not the topic of causation itself Causation might be atransparent relation even if causal explanation is not If lsquoe causes frsquo istransparent it does not follow that the corresponding assertion ofexplanation lsquoe causally explains frsquo must be Nor of course would Ideny that we often take the simple assertion of causality as explanatorybut the success of our so doing can only be insured when lsquoe causes frsquomeets more conditions than would have to be met simply for lsquoe causes frsquoto be a literal truth

Consider the following argument

(1) The hurricane explains the loss of life(2) The hurricane=the event reported in The Times on Tuesday(3) The event reported in The Times on Tuesday explains the loss

of life

Or to pick an example adapted but slightly altered from Mackie10

(1a) Oedipusrsquo marrying his mother explains the tragedy that ensued(2a) Oedipusrsquo marrying his mother=Oedipusrsquo marrying the woman

he thought least likely to bring tragedy to Thebes(3a) Oedipusrsquo marrying the woman he thought least likely to bring

tragedy to Thebes explains the tragedy that ensued

What Mackie says is this the first description lsquohelps to explain the tragedyin a way that thehellip[latter] does not What is referentially transparent maybe for that very reason explanatorily opaquersquo

My view is that both (3) and (3a) are literal fasehoods If the conclusionsof these two arguments are false the best diagnosis of whathas gone wrong

163


in the argument is that despite appearances the first premiss of eachargument is not transparent not to be construed asstating that a relationobtains between two events as the Lewis-Woodward proposal for theontology of explanation proposes One should take the falsity of theconclusion to throw doubt on the conception that particular events (orstates or whatever) can transparently explain events Explanation is nevera natural relation It never touches things that directly and immediatelybut is always mediated through the features or characteristics which areappropriate for explanation Oedipusrsquo marrying his mother is the sameevent as Oedipusrsquo marrying the woman he thought least likely to bringtragedy to Thebes but the event is explanatory as conceptualized in thefirst way but not as conceptualized in the second way

Are the conclusions of the two arguments really false because the oneevent doesnrsquot explain the second at all or are the conclusions literallytrue but just poor(er) explanations This raises a very complicated questionabout the nature of such a distinction which I have already touched on inthe first chapter but let me say here partly by way of repetition justenough to produce what I hope will be an adequate underpinning for thethought that the conclusions of the argument are literally false

There are pragmatic considerations in giving explanations First it islegitimate to give less than the whole explanatory truth to an audiencegiven onersquos knowledge of its interests and existing knowledge Theaudience may only want to know about some part of the full cause Or itmay not want a very specific description of that cause but only a moregeneral descriptionmdashan audience of historians might want to know onlythat a plague in tenth-century China reduced the population level but notneed to know what the plague was Or there may be two levels of relevantdescription of the cause a macro-description and a micro-description forexample and one rather than the other of the descriptions might beappropriate for the audience one was addressing11

Poor explanations are ones that make these choices between relevantdescriptions or between parts of the relevant description in the wrongway (lsquorelevancersquo let us suppose being cashed out for us by the theory ofexplanation at hand I shall discuss the issue of relevance more fully inchapters VI and VII)

But none of these pragmatic considerations covers the case of theintroduction of an outright irrelevant description even though theirrelevant description is a true description of the cause The connectionbetween the event irrelevantly described and the topic at hand can onlybe seen in a derivative and parasitic sort of way by someone who happens

164


to know the irrelevant description and knows that it fits this event Toselect inappropriately from the relevant information is a poor explanationfor the purposes at hand to offer explanatorily irrelevant information isto fail to explain at all In the conclusions of the above two argumentsjust such wholly irrelevant descriptions are being introduced Althoughtrue descriptions of the cause they are entirely irrelevant descriptionsfrom the point of view of explanation

The conclusions of these arguments are then false I grant that theperson who knows the identities might feel reluctant to count theconclusions as wholly literal falsehoods since there is a sense in whichhe can see what the conclusions are trying to get at The conclusionsmention events which when differently described do explain

One might draw the distinction in this way there is a difference betweengiving an explanation of something on the one hand and asserting orimplying that there is some explanation of that thing without actuallygiving it on the other So there are two different and incompatible waysin which one might understand (3) and (3a) If they are taken as giving anexplanation as the Lewis-Woodward thesis must suppose then they aresimply false If on the other hand they are taken as asserting or implyingthat there is some explanation of the explanandum event which involvesthe first event when differently described they are truths

Someone who does not know that the hurricane was the event reportedin The Times on Tuesday may take (3) to be implying that there is someexplanation of the loss of life which involves a different description forthat very same event described as lsquothe event reported in The Times onTuesdayrsquo But he will not take (3) as giving that explanation The theoryof singular event explanation is as it stands false Explanation evencausal explanation can never just be of events by events sansqualification

When discussing explanation writers often slip naturally andsometimes unconsciously into the terminology of facts Strawson spokein the quotation at the beginning of the chapter of facts standing in thenon-natural explanation relation to other facts We have already noted thevagaries of Hempelrsquos remarks on this point and we decided to take as hisconsidered view the one he expresses in the extended discussion in hislater article12 in which he distinguishes between concrete events andsentential facts and events and argues that it is lsquosentential facts and eventsrsquowhich are explained in an explanation

Facts are the second candidate for the relata of the explanationrelation lsquothe fact that p explains the fact that qrsquo where typical

165


expressions referring to facts contain whole sentences here representedby lsquoprsquo and lsquoqrsquo Adopting facts as the relata might still yield a theory ofsingular explanation of a sort if facts like events are a kind of individualentity in the widest sense But the relation that relates facts if suchthere be will be unlike causation and in some sense yet to be exploreda non-natural relation

If facts are to come on the stage as a serious ontological candidate itwould be nice if we had a catalogue of the various kinds of facts thatthere might be I introduced various species of facts in chapter IVsingular facts existentially general facts universally general factsstochastic facts and facts of identity I do not have a complete catalogue(for instance I do not discuss the lsquoRussellianrsquo questions of whetherthere are conjunctive facts disjunctive facts and negative facts) I shallbe mostly concerned in this chapter with singular facts A singular factis a fact about a particular namely the fact that that particular o hassome property P (the fact that the fire was purple and odd-shaped) lsquoPrsquomight be a relational property and so the fact might be about two ormore particulars13 Facts of identity are singular facts even so theydeserve special mention Some of the issues about extensionality andtransparency will be irrelevant to some kinds of facts (eg there are noquestions of transparency for universally and existentially general facts)but this should be fairly obvious

Other accounts of explanation make statements or propositions therelata of the explanation relation Peter Achinstein for example holdsthat an explanation always includes as a constituent in addition to an acttype a certain kind of proposition14

Donald Davidson has an account of the logical form of explanationsentences and a proposal for their analysis15 (Proposals concerning logicalform and analysis constrain one another but I ignore these issues here)First consider his remarks about their logical form Suppose Jackrsquos fallingdown explains his breaking his crown For him this in correct logicalform is (with Davidsonrsquos numbering)

(8) The fact that Jack fell down explains the fact that Jack broke hiscrown

lsquoInhellip(8) intensionality reigns in that similar substitution [ie ofequivalent sentences or co-extensive singular terms or predicates] inor for the contained sentences is not guaranteed to save truthrsquo(Davidson 197586) Note that Davidsonrsquos remarks about (8) should

166


not be confused with his better-known remarks about (again withDavidsonrsquos numbering) (2) lsquoThe fact that there was a short circuitcaused it to be the case that there was a firersquo (Davidson 197584ndash5)Davidson does not think that (2) does finally reveal the logical formof causal sentences (his argument that this is so uses the slingshot)But there is never any hint that (8) does not adequately reveal thelogical form of (causal) explanation sentences

As for the analysis of such explanation assertions Davidson makesonly a single remark What for Davidson does the explanation relationrelate Davidson says at least in one passage that explanation is a relationthat relates statements lsquoExplanations typically relate statements noteventsrsquo (Davidson 197593) Perhaps Davidsonrsquos well-known dislike offacts is what inclines him and should incline us to statements rather thanfacts

Davidsonrsquos arguments against facts are only against the use of factsfor certain purposes In lsquoCausal Relationsrsquo the argument relying onthe slingshot runs as follows16 Suppose the logical form of a causalsentence was (2) the fact that p caused the fact that q Clearly theconnective lsquothe fact thathellipcaused the fact thathelliprsquo is not truth-functionalsince substitution of contained sentences by sentences with the sametruth-value does not preserve the truth-value of the whole Howeversubstitution of singular terms for others with the same denotation shouldnot touch their truth-value if for example Smithrsquos death was causedby a fall from the ladder and Smith was the first man to land on themoon then the fall from the ladder was the cause of the death of thefirst man on the moon But by the slingshot substitutability of singularterms with the same denotation entails truth-functionality contrary tothe original supposition So (2) cannot give the logical form for causalsentences

In another well-known paper on the correspondence theory of truthDavidson slingshoots in this way17 Suppose lsquosrsquo is some true sentence Onany correspondence theory the statement that s corresponds to the factthat s These correspondence contexts must be transparent if lsquoSmith fellfrom a ladderrsquo corresponds to the fact that Smith fell from a ladder and ifSmith was the first man on the moon then lsquoSmith fell from a ladderrsquocorresponds to the fact that the first man on the moon fell from a ladderSo by the slingshot since lsquothe statement that s corresponds to the factthat srsquo is transparent then for any true sentence lsquotrsquo the statement that scorresponds to the fact that t In general lsquothe statement thathellip correspondsto the fact thathelliprsquo if transparent is truth-functional If a statement

167


corresponds to any fact it corresponds to them all There is on this accountjust One Great Fact if there is any

Both applications of the slingshot suppose that the context underdiscussion is transparent (and that logical equivalents are substitutablesalva veritate) and argue that if it were transparent it would be truth-functional And if facts were to do what causation and correspondencewould require of them then expressions referring to facts must betransparent causation and correspondence contexts should permitreplacement of singular terms by coextensive singular terms salva veritateIt isnrsquot facts per se that Davidson doesnrsquot like but rather his point is thatif there are facts they arenrsquot any good for causation or correspondence

Pari passu if one fact explains another wonrsquot it explain all facts Itdepends on what facts are like Our intuitions may tell us that lsquothe factthat p explains the fact that qrsquo is not even transparent (we shall have tosee) if explanation contexts are not even transparent the slingshot cannotget started towards the conclusion that if one fact explains another itexplains them all

This may help disarm an argument against facts Is there any reason tochoose facts over (true) statements as the relata of the explanation relationAs the discerning reader will note by the end of this chapter this is apressing problem for me because the facts required for explanation aremore like true statements or propositions than facts might even ordinarilybe thought to be

But there is still an important metaphysical difference betweenstatements and even these unordinary facts required for explanation Facts(empirical ones at any rate) include worldly particulars (like persons andphysical objects for example) as constituents18 Facts may include morethan just these worldly particulars but they do include at least themStatements even true ones are not so composed Having rejected theWoodward-Lewis idea that worldly events can explain worldly eventstout court my strategy is to keep explanation as worldly as possible foras long as possible I start with particulars and their properties and addconceptualization or description to them In the end I have to admit agreat deal of this additional conceptualization and description

There are then two competing theories that survive the eliminationof event explanation (a) the theory that makes facts the relata of theexplanation relation and (b) the theory that casts true statements in thisrole but adds that statements can explain statements only in virtue ofthe natural relations obtaining between worldly things If the readerwere to insist that (a) and (b) are equivalent the difference between

168


them being merely stylistic I have no strong reason to resist the claimother than the metaphysical thoughts above NLWilson19 for examplesays that lsquofactrsquo and lsquotrue propositionrsquo are lsquosynonymsrsquo (Wilson 1974305)Even if Wilson and others are right though my discussion still has apoint Many of those who write on explanation use the idea of factsexplaining facts quite uncritically What follows is an attempt to spellout what is involved in the idea If it turns out that it is equivalent to theidea of statements explaining statements so be it But to repeat I donot think that the concept of a fact and the concept of a true statementor proposition are the same there are I believe metaphysical differencesbetween them

Explaining facts

Letrsquos return to Woodwardrsquos distinction Suppose I want to explainthe housersquos burning with a purple odd-shaped flame and not just toexplain the housersquos burning which burning happened to be a burningwith a purple odd-shaped flame That is I want to direct theexplaining onto the purpleness and odd-shapeness of the flame Theexplanation of the housersquos burning with a purple odd-shaped flamedemands something that is not equally demanded by the explanationof just the housersquos burning

If (counterfactually) we had agreed to the idea of event explanationand if we were also prepared to adopt a very fine-grained criterion ofevent (and state) identity according to which if two event descriptionsuse different properties it follows that the descriptions describedifferent events then we could see this second explanation as an eventexplanation too On this fine-grained criterion the housersquos burningand the housersquos burning with a purple and odd-shaped flame wouldcount as two different token events and it would be no surprise thattheir explanations were different We could then refuse to accept thatan explanation of the burning is an explanation of the burning with apurple and odd-shaped flame We could say that the short circuit isthe explanation of the housersquos burning the presence of a certainimpurity in the combustible material (say titanium) is the explanationof the housersquos burning with a purple and odd-shaped flame and thatboth explanations are event explanations being explanations of (twodifferent) particular or token events

169


But I have already rejected this fine-grained criterion of event andstate identity and the view that explanation can be of events (Even ifwe adopted the fine-grained analysis at this point in my argument itwould not survive because of additional problems I raise later) Weneed some other way to get at the difference between the two explanandathe housersquos burning and the housersquos burning with a purple and odd-shaped flame The explanation of the housersquos burning with a purpleodd-shaped flame is best understood according to me as an explanationof the fact that an event the housersquos burning has a specific feature vizbeing a burning with a purple and odd-shaped flame Any concreteparticular has an indefinitely large number of characteristics or featureswhat we want to explain may be not just the particularrsquos occurrenceand certainly not its having all the features that it does have whichwould be an impossible task

What is needed is what I would call lsquoa feature-introducing operatorrsquowhich introduces a (usually short) list of features of the particular forwhich an explanation may be required In lsquothe fact that the fire is purpleand odd-shapedrsquo at least two features of the fire seem to be introducedthat the fire is purple and that it is odd-shaped Not every explanationexplains all the features introduced by the fact locution for context mightmake it appropriate to explain some features and to ignore others Selectionof introduced features for explanation is a different and pragmatic context-dependent matter we often indicate such selection by stress emphasisand so on

But features cannot even be selected for explanation which are notintroduced by the fact locution at all The fire was no doubt a fire thatoccurred at some time t but the fact that the fire occurred at t is a differentfact from the fact that it was purple and odd-shaped and its occurring att is not a feature introduced by the latter fact Even Woodwardrsquos allegedexample of explaining the occurrence of the fire sans phrase can beconstrued as explaining the fact that the fire occurred where this factlocution unlike the fact that the fire had certain features introduces theoccurrence but nothing else as the feature of the fire which matters forthe explanation

Aristotlersquos terminology was as we saw designed for a similar purposeit isnrsquot Polyclitus qua the pale or the musical man who explains the statuebut Polyclitus qua sculptor or qua sculpting Aristotle was alive to thenon-extensionality of lsquohellipexplainshelliprsquo and my discussion here is intendedto build on his insight lsquoquarsquo is Aristotlersquos feature-introducing operatorlsquothe fact thathelliprsquo is mine

170


Since the point of a feature-introducing operator is to introduce featuresfor explanation all of the features introduced by lsquothe fact that helliprsquo areassumed to be explanatorily relevant When this assumption is not metthe whole explanation claim is false For example to return to an examplementioned briefly in chapter I it is a truth that the fact that Jones is a manexplains the fact that Jones did not become pregnant But it is false thatthe fact that Jones is a man who regularly takes birth control pills explainsthe fact that Jones did not become pregnant In that falsehood Jonesrsquosregularly taking birth control pills is an introduced feature Apresupposition of that explanation claim is that the feature regularly takingbirth control pills is explanatorily relevant to Jonesrsquos failure to becomepregnant Since the presupposition is false the explanation claim whichpresupposes it is false too

If I were asked to explain the fact that the fire was purple and odd-shaped there are not just two but three features which fall within thescope of the explanation I mentioned above the two obvious featuresthe purpleness and the odd-shapeness of the fire I agree with Woodwardthat an explanation of why the fire was purple and odd-shaped presupposesthat but does not explain why there was a fire Butmdashand this can bebrought out by appropriate stress or emphasismdashI may want to know whyit was a fire to which the purpleness and odd-shapeness attachedthemselves as it were Here the lsquorather-thanrsquo locutions are helpfulknowing that something or other was so oddly coloured and shaped whywas it a fire and not something else (perhaps a cloud)

So being a fire in spite of the fact that it finds its place in theexplanandum not by means of a predicate but rather by means of a definitedescription is also a feature introduced for the purposes of explanationIn lsquothe fact that the D is Prsquo the feature being a D is also introduced eventhough the logical function of lsquothe Drsquo is to fix reference The same appliesmutatis mutandis when lsquothe fact that the D is Prsquo is the explanansExplanatory weight can be carried by the referring expression when it isa definite description as well as by the predicates

If the fact that the hurricane measured force 10 explains the fact thatthere was a subsequent tidal wave and even though the hurricane was theevent reported in The Times on Tuesday the fact that the event reported inThe Times on Tuesday measured force 10 does not explain the fact thatthere was a subsequent tidal wave (Intuitions to the contrary may ariseonly because given that the property in the explanans is the property ofmeasuring force 10 it is easy to work out what the relevant definitedescription of the event referred to as lsquothe event reported in The Times on

171


Tuesdayrsquo must in fact be viz the hurricane) The event which figures inthe explaining fact must be relevantly referred to as well as relevantproperties or features attributed to it to make explanation work

This confirms the idea that the slingshot argument has no toe-hold inexplanation in contrast to correspondence and causation The expressionlsquothe fact thathellipexplains the fact thathelliprsquo is not transparent at least not fordefinite descriptions unlike causation and correspondence contexts thereis no substitutability of co-referring definite descriptions in explanationcontexts salva veritate Two different definite descriptions of a particularmay utilize two different properties and when this is so since fact locutionsmake properties matter definite descriptions are not replaceable salvaveritate in lsquothe fact thathellipexplains the fact thathelliprsquo

The non-extensionality of facts

I now propose to speak freely of facts as well as fact locutions Iassume that if I specify truth conditions for the fact locution lsquothefact that p=the fact that qrsquo I am entitled to speak with serious onticintent of identity conditions for facts I shall always mean by lsquoa factlocutionrsquo a locution with the form lsquothe fact that p=the fact that qrsquoIf in what follows I say that facts or their identity conditions areor are not transparent predicate extensional etc the reader willunderstand how to translate such remarks from the material to theformal mode if necessary My strategy is first to establish the truthconditions for fact locutions and then to look at lsquothe fact that pexplains the fact that qrsquo in the light of that discussion

Are the identity conditions for facts predicate extensional They arenot as the invalidity of the following argument makes clear

(1) The fact that x is P=the fact that x is P(2) (x) (Px=Qx)(3) The fact that x is P=the fact that x is Q

Since if P Q (3) will be false (1) does not permit substitution ofcoextensive predicates salva veritate An argument similar to the oneabove shows that the identity conditions for facts are not sentenceextensional (truth-functional) either

That the identity conditions for facts are not predicate extensional suitsthem for their feature-introducing role in explanation The fact that x iscordate is a different fact from the fact that x is renate even though the

172


properties of being cordate and renate are coextensive Happily so becausean explanation of the fact that x is renate will not do as an explanation ofwhy x is cordate A feature introducer introduces a feature without therebyintroducing all other features coextensive with the first feature Factlocutions offer a means for introducing the property and making theproperty matter in a way that events and state of affairs do not (or do soonly at best controversially)

So far my facts are similar to those of NLWilson

For consider a domain consisting solely of sugar cubes one of whichis a For such a domain the class of white things is identical with theclass of cubical things Nevertheless the fact that a is white is notidentical with the fact that a is cubical And that is because theproperties whiteness and cubicalness though coextensive are notidentical

(Wilson 1974306)

I discussed the non-transparency of lsquothe fact thathellipexplains the factthathelliprsquo for definite descriptions in the preceding section It is alsotrue (this is a different but related claim) that lsquothe fact thathellipis identicalto the fact thathelliprsquo is non-transparent for definite descriptions Thefact that Cicero was the greatest Roman orator is identical to the factthat Cicero was the greatest Roman orator Cicero=the man whodenounced Cataline But the fact that Cicero was the greatest Romanorator is not the same fact as the fact that the man who denouncedCataline was the greatest Roman orator Failure of fact locutions tobe transparent (with reference to definite descriptions) like failureof predicate and sentence extensionality makes them suitable (thusfar at any rate) to be the relata of the explanation relation

Facts worldly or wordy

Some have argued that facts are a special kind of combined linguisticand non-linguistic item JLAustin in his debate with Strawson ontruth claimed that lsquoand so speaking about ldquothe fact thatrdquo is acompendious way of speaking about a situation involving both wordsand worldrsquo20 My denial of sentence and predicate extensionality anddefinite description transparency for fact locutions does not suggestthe involvement of words or conceptualization in the constitution of

173


facts My remarks on the feature-introducing character of facts bringproperties or features or characteristics onto the scene but do not bringwords or concepts to the fore In what way are facts on the side of thewords (or concepts) at all as Austin says

The fact that the fire is purple introduces the purpleness of the firewhether for explanation or whatever It isnrsquot the word lsquopurplersquo that isrelevant or even the concept of purpleness but the real colour of the realfire however described or conceptualized In order to see the lsquowordsrsquocharacter of facts I think one needs to look at the questions of the co-typical predicate extensionality and the name transparency of facts UnlikeAustin I prefer to speak where possible of the conceptual componentrather than the words component of facts It will turn out that what often(although not always) matters in explanation is how we conceptualize orthink about something not what words we use in the thought orconceptualization

The co-typical predicate extensionality of facts

For each of name transparency and co-typical predicate extensionalitythere are the following two separate questions (1) Are fact locutionsname transparent (co-typical predicate extensional) (2) Given theanswer to (1) can facts so conceived be the relata of the explanationrelation

So far given the failure of fact locutions to be definite descriptiontransparent sentence extensional (truth-functional) and predicateextensional facts were suited to be the relata of the explanation relationSuppose the answers to the questions about the name transparency andco-typical predicate extensionality of fact locutions make facts unsuitablefor explanation Rather than impose a theory of explanation onto facts tobegin with to ensure their suitability for explanation I try to answer (1)on the basis of my (and hopefully your) intuitions My claim will be thatfacts if the deliverances of intuition are as I think they are wonrsquot do thewhole job required by explanation I refer to such ultimately inadequate(for the purposes of explanation) facts as lsquoordinary factsrsquo They pass thetest for explanation by failing the test for definite description transparencypredicate and sentence extensionality because facts can make propertiesmatter But in the case of co-typical predicate extensionality and nametransparency we move beyond the mattering of just properties here factsfail the test for explanation by passing it for name transparency and co-

174


typical predicate extensionality To this extent they wonrsquot fully do whatexplanation requires of its relata

Let me take the co-typical predicate extensionality question firstSuppose property P=property Q Are the identity conditions for factsco-typical predicate extensional NLWilson thinks that they are forthe following reason lsquohellipred red is the color of ripe strawberries Fromwhich it follows that thehellip[fact] that Socrates is red is identical withthehellip[fact] that Socrates is the color of ripe strawberriesrsquo (Wilson1974305) Wilson argues in this way lsquotwo facts will be identical ifthey have the same constituents in the same orderrsquo (p306) On Wilsonrsquosview since the fact that Socrates is red and the fact that Socrates is thecolour of ripe strawberries lsquocontainrsquo the same individuals and the sameproperties they must be the same fact Neither fact introduces anythingnot introduced by the other How could the facts be different There isno property or object or any other item which is a lsquoconstituentrsquo in onethat is not in the other

I agree with Wilsonrsquos general view (although I have doubts aboutWilsonrsquos specific example) Facts are co-typical predicate extensionalSuppose the property P=the property Q (Let lsquoPrsquo and lsquoQrsquo both be namesof properties rather than descriptions of them Recall that it makes nodifference to my argument whether the identity statement is necessaryor contingent let readers suppose whatever they wish about theepistemic status of statements of property identity which use namesof properties) The fact that x is P and the fact that x is Q in such acase differ neither in the individuals they are about nor in the featuresof the individuals that are introduced The first fact introduces thefeature P the second the feature Q And ex hypothesi these are thesame feature

There are examples other than Wilsonrsquos which lead me to think thatthe identity criteria for facts on any theory of facts which is true to howfacts are ordinarily thought about are co-typical predicate extensionalThese are examples such as the fact that a gas has temperature t=thefact that its constituent molecules have mean kinetic energy m the factthat ice is water=the fact that ice is composed of H2O molecules I dotake the relevant property identities on which these fact identities restto be a posteriori for otherwise the examples would be of little interestfor scientific explanation As I said above it is controversial whetherthese identities are contingent or necessary but a decision on this isirrelevant for what I have to say here (There is also a question about theidentity conditions for facts in those cases in which the property identities

175


are a priori eg there is the question whether to count the fact thatHarry is an eye doctor as the same fact as the fact that Harry is an oculistIf facts are the same when the relevant property identity is a posteriorithey surely would be the same when the relevant property identity is apriori)

But now we must turn to the second question that I mentioned aboutthe co-typical predicate extensionality of facts If facts as ordinarilyunderstood are co-typical predicate extensional can such facts do whatexplanation requires I assume that the following is an explanation (of amacro-state by a micro-state) an explanation for the fact that a sample ofideal gas b has temperature t is the fact that brsquos constituent moleculeshave mean kinetic energy m Such an explanation relies on the propertyidentity temperature=mean kinetic energy

There are other explanations at other levels for this fact Forexample another explanation of the fact that b has temperature t isthat I stuck b in the oven at gas mark 4 But an explanation for itshaving temperature t and the one most appropriate in certain scientificsituations is in terms of the mean kinetic energy of its molecules Inany case if the reader is inclined to dispute that this is a bona fideexample of an explanation let him take as an example whateverexample he wishes of an explanation of a macro-state by an underlyingmicro-state such that the property of being in that macro-state isreductively identified with the property of being in that micro-stateReduction of one science to another has often been taken asparadigmatic of explanation whether paradigmatic or not suchreductions must yield some examples of property identities which areexplanatory

Consider the following four claims

(1) No (empirical) fact explains or even partly explains itself (theexplanation relation is irreflexive)

(2) Having temperature t=being composed of molecules with mean kineticenergy m (a statement of property identity whether i t bemetaphysically necessary or contingent)

(3) The fact that b has temperature t is explained (or partly explained)by the fact that brsquos molecules have mean kinetic energy m

(4) Facts are co-typical predicate extensional

But (1)ndash(4) are inconsistent because (2)ndash(4) jointly imply that the(empirical) fact that b has temperature t explains or partly explainsitself which (1) says is never so

176


I take this argument as demonstrating that explanation is not just arelation between facts as constituted by worldly particulars and theirproperties apart from how they are conceptualized If P=Q the fact thatx is P and the fact that x is Q introduce the same feature What matters inexplanation isnrsquot only property introduction but the way in which weconceptualize the property viz whether the property P is introduced asproperty P or as property Q

If facts are thought of in the ordinary way Austin was wrong Ordinaryfacts as co-typical property extensional entities are not combinations ofwords (or concepts) and the world Even ordinary facts are not whollyextensional it is true since properties matter to them in a way for whichpredicate and sentence extensionality cannot account But their co-typicalpredicate extensionality makes ordinary facts unsuitable as the relata ofthe explanation relation We need relata for that relation for which notonly properties or features matter but the way in which we conceptualizeor cognize them matters too

I therefore introduce lsquospecialrsquo facts which are constructed to do justwhat explanation requires of its relata21 If readers think that thedeliverances of my intuition about facts are in error and that what belowI call lsquospecial factsrsquo are what they think ordinary facts are like I have nogreat objection The philosophical point about what we need to do thework that explanation requires remains unaltered

I continue to call such things lsquofactsrsquo but lsquoin the special orepistemicized sensersquo Such a special fact might also be thought of as anordered pair of an ordinary fact and a complete conceptualization ofthat fact (Alternatively instead of thinking of explanation as a relationbetween such ordered pairs it might be thought of as a four-placerelation whose terms are an ordinary fact a complete conceptualizationof that fact a second ordinary fact and a complete conceptualization ofthe second fact)

What I call lsquoconceptualizationsrsquo are I think very much like whatNathan Salmon if I understand him rightly calls that lsquoby means ofwhich a proposition is graspedrsquo22 Just as he thinks that a propositioncan be grasped in different ways so that belief has to be thought of as atriadic relation between a believer a proposition and a way of graspingthat proposition (lsquoa mode of presentationrsquo) so similarly I say thatexplanation is a relation between ordinary facts plus the ways in whichthose facts are grasped or their modes of presentation (see Salmon 1986117ndash20) Both in the case of his modes of presentation and for myconceptualizations there is a connection not just with the semantic (in

177


his case) and not just with properties (in mine) but also with theepistemological (see p120)

I also continue to speak of facts as the relata of the explanation relationbut one must remember that I intend lsquofactrsquo in this special lsquoepistemicizedrsquosense which may not accord with the way in which lsquofactrsquo is normallyunderstood Whether a fact as normally understood explains or is explaineddepends at least in part on the way in which the properties involved areconceptualized relative to the conceptualization of a property in one waythe fact may be explanatory relative to a different conceptualization ofthe same property the fact is not explanatory To this extent it must beadmitted that explanations are not fully independent of how we thinkabout things23

In the argument sketched above (4) is true of facts in the ordinarynon-epistemicized sense (3) is true only for the special facts which includea conceptual component So the argument is invalid since it turns on anambiguity in the meaning of lsquofactrsquo (1) is true on my view The relata ofthe explanation relation are two different lsquoepistemicizedrsquo facts the factthat it has temperature t and the fact that its constituent molecules havemke m These are two different epistemicized empirical facts and so noepistemicized empirical fact explains itself

The name transparency of facts

We still have to deal with the question of the name (as opposed tothe definite description) transparency of fact locutions Zeno Vendlerargues that lsquofacts are referentially transparent propositions eventrue ones are opaquersquo24 And Wilson again lsquoit followshellipthat the[fact] that Socrates is red is identical with the [fact] that the teacherof Plato is redrsquo On their view if o=i the fact that i is P=the factthat o is P

We can see a certain ambiguity in Vendlerrsquos and Wilsonrsquos assertionsIn my view they are both wrong for the cases they mention I have alreadycovered these sorts of cases before in my discussion of definite descriptiontransparency My intuitions tell me that the fact that Socrates is red is notthe same fact as the fact that the teacher of Plato is red I can account forthe difference in terms of property introduction Even though lsquothe teacherof Platorsquo is functioning as a referring term as such it has descriptivecontent Thus it brings additional or different properties into the factAlthough the first fact is constituted by (they lsquoare the joint full inventory

178


constituents ofrsquo the fact in Wilsonrsquos parlance) Socrates and the propertyred the second fact has more constituents to wit Socrates Plato theproperty red and the relational property being the teacher of

(An exception to the above will have to be made for the following sortof case the fact that the water in the glass is warm and the fact that thestuff in the glass composed of H2O molecules is warm If beingwater=being composed of H2O molecules the two definite descriptionsemploy the same property So this sort of example will have to be handledlike the case of proper names below)

However Wilson and Vendler are right for names ordinary facts arename transparent Consider the fact that Cicero died in 43 BCE25 SinceCicero and Tully are the same person the fact that Cicero died in 43 BCEand the fact that Tully died in 43 BCE are one and the same fact Torepeat the argument used above in the case of property identity howcould the facts be different They involve the same individual or particularand all the same properties or features of that individual Nothing is aconstituent of one that is not a constituent of the other Fact locutions aretransparent for proper names of individuals or particulars

The identity conditions for ordinary facts are sensitive to the rigidityor otherwise of contained singular terms The fact that the greatest Romanorator died in 43 BCE the fact that the greatest Roman Stoic philosopherdied in 43 BCE and the fact that Cicero died in 43 BCE are three differentfacts and only the last mentioned fact is the same fact as the fact thatTully died in 43 BCE

Given these identity conditions how do facts so identified fare for thepurposes of explanation I have already argued that the definite descriptionnon-transparency of facts suits them for explanation Someone who doesnot know the identity the event reported in The Times on Tuesday=thehurricane can do nothing with lsquobecause of the fact that the event reportedin The Times on Tuesday occurred at midnightrsquo as an explanation for thefact the tidal wave occurred in the early hours of the next morning If theperson does not know that the referent of lsquothe event reported in The Timeson Tuesdayrsquo and lsquothe hurricanersquo are the same the reply far from being anexplanation is simply mysterious We didnrsquot need to tamper with theidentity conditions for ordinary facts in order to take on board this pointabout their role in explanation

On the other hand letrsquos consider the case of proper names We haveagreed that the fact that Tully died in 43 BCE and the fact that Cicerodied in 43 BCE are the same (ordinary) fact They differ with regard to noconstituent neither a particular nor a property nor feature So unlike the

179


case of definite descriptions the criteria for ordinary fact identity are notalready lsquoepistemicizedrsquo in the case of proper names of individuals anymore than they were in the case of co-typical predicates

We have no slingshot problems with the concession that facts are name-transparent for the slingshot argument only goes through for definitedescriptions not names But if differences in how one names an individualmake no difference to the identity criteria for facts such facts will not beadequate for the requirements of explanation Suppose the question israised why Cicerorsquos speeches stop in 43 BCE We can make the samegeneral point that we did before about the hurricane and the event reportedin The Times on Tuesday Someone who does not know that Cicero isTully can do nothing with lsquobecause Tully died in 43 BCErsquo as an explanationfor the fact that Cicerorsquos speeches stop at that date Without knowledgeof the identity this retort is also simply mysterious

So if we want entities suitable to be the relata of the explanation relationour special facts should be ordered sets of ordinary facts andconceptualizations andor names both of the properties and of theindividuals who are the constituents of the facts Cicero named as lsquoCicerorsquomight explain whereas Cicero named as lsquoTullyrsquo may fail to explain Hereit seems more appropriate to speak in terms of names rather thanconceptualizations It makes perfectly good sense I think to speak ofmean kinetic energy conceptualized as mean kinetic energy and meankinetic energy conceptualized as temperature these are two different waysin which to conceptualize one and the same property But there are noconcepts of Cicero and of Tully which might be variously applied to oneand the same person lsquoCicerorsquo and lsquoTullyrsquo are rather different nameswhich might be variously applied to him

Are the doubly epistemicized facts required for explanation to borrowan apt term from Stephen Schiffer lsquopleonasticrsquo or lsquonon-pleonasticrsquo26 Sofar it might seem that they are to be taken as non-pleonastic They arenon-pleonastic because I take the expression lsquothe fact that p explains thefact that qrsquo even when lsquofactrsquo is employed in the epistemicized sense tohave a relational analysis and I take lsquothe fact that prsquo and lsquothe fact that qrsquoas singular terms which refer to facts (or ordered pairs of facts and lsquomodesof presentationrsquo of such) (If one thinks of the fact that p and the fact thatq as two intensional objects and calls them lsquoarsquo and lsquobrsquo then the sentencelsquoa explains brsquo is transparent since any singular term that designates thesame fact (understood in my special way) is substitutable salva veritate)On the other hand Schiffer connects up the idea of ontologicalcommitment with that of existence which is lsquolanguage-independentrsquo

180


(Schiffer 1987145) That seems to me a conflation The special factsrequired as the relata of the explanation relation may not be language-independent (and not conceptualization-independent) any more than theexistence of words is language-independent or the existence of conceptsis concept-independent So this commitment to facts does not entail full-blooded realism about them in one sense of that word To this extentsince the facts I require for explanation turn out to be entities dependenton human conceptualization and thought any realist ontologicalcommitment to them would have to be so qualified

My argument shows that there is an epistemic requirement inexplanation facts explain facts only when the features and the individualsthe facts are about are appropriately conceptualized or named The conceptof explanation is partly lsquoepistemologicalrsquo But this is by itself no concessionto a pragmatic theory of explanation for the explanatory relevance of theway in which things are conceptualized may not be audience-relative Inone important sense of the term an epistemological conception ofexplanation can be objective Knowledge is uncontroversially anepistemic concept and no one argues from that fact alone that objectiveknowledge is impossible

In the remainder of the book I assume that such special or epistemicizedfacts are the relata of the explanation relation even when looseness orease of expression may have me speaking of events explaining eventsThis can always be translated as the fact that such an event occurredWhat counts as an appropriate conceptualization This introduces a newtopic to which I also return in the next and the final chapter

181

CHAPTER VI

Arguments Laws and Explanation

Although I use this chapter and the next to pronounce on a number ofthe claims about explanation that I have described in the historicalsection of the book there are two issues to which I want especially toattend The first which I tackle in this chapter is the thesis commonto Aristotle Mill and Hempel that full explanations are validarguments the second which I treat in chapter VII is Hempelrsquos viewthat some singular explanations are non-causal explanations On theway to making these two points something like a general view ofexplanation will emerge That general view is put tentatively andwith some hesitation I regard it more as a research project than as afinished theory that is able as it stands to meet all difficulties

The first issue for discussion then concerns the claim that explanationsare arguments On Aristotlersquos Millrsquos and Hempelrsquos accounts ofexplanation explanations are arguments although sometimes ellipticallyor enthymetically presented Probabilists and deductivists although theydisagree about whether there are any explanatory arguments with a non-deductive form both hold an argument theory of explanation If anexplanation is an argument then (on any plausible account of what sortof argument this will be) it will have to include at least one lawlike premiss

Since argument theorists include laws as premisses in their account ofexplanation the first issue also involves the question of the relationshipbetween explanation and laws What is a law of nature There are widelydifferent responses to this question in the literature In what follows andindeed throughout the book I assume(d) that the lsquoorthodoxrsquo answer is

182


correct a necessary condition for a sentencersquos stating a deterministic lawof nature is that it be a true universally quantified generalization On theorthodox view sentences which state deterministic laws of nature typicallyhave or entail something with this form Although theuniversally quantified conditional might also be more complicated thanthis (eg the consequent might also be existentially quantified) it willmake no difference to the argument if we only consider sentences withthis simple conditional form I consider in these remarks only the case ofdeterministic laws and neglect stochastic laws

No orthodox theorist would consider this condition by itself sufficientAccidental generalizations have this form too Further universallyquantified material conditionals are true when their antecedent terms aretrue of nothing So if this condition were by itself sufficient forlawlikeness and if nothing in the universe was an F then both of thefollowing would state laws of nature and

There are various proposals for adding further conditions to the oneabove Some are proposals for strengthening the generalization by addinga necessity-operator laws of nature are stated by nomically necessaryuniversally quantified generalizations1 Others ascribe to the universallyquantified generalization an additional special epistemic status or a specialplace in science or impose further syntactic requirements2 My argumentis neutral between all of these variants of the orthodox proposal

On the other hand suppose that the orthodox view does not provideeven a necessary condition let alone a sufficient one for somethingrsquosbeing a law of nature If so my argument would have to proceed somewhatdifferently I am sympathetic to some of these non-orthodox views but Ido not deal with any of them here nor with how their acceptance wouldalter my argument3

I start by way of outlining some of the standard counterexamples toHempelrsquos account of full explanation which will be useful for thediscussion in this and in the next chapter Whether or not they arecounterexamples to Millrsquos or Aristotlersquos accounts as well will depend onthe efficacy of the proposed cure in Hempelrsquos case My view will be thatsome are counterexamples to the accounts of all three thinkers but Ipropose to begin the discussion by taking them to be allegedcounterexamples only to Hempel

These counterexamples cluster around two difficulties (a) irrelevanceand (b) symmetry I do not say that Hempelrsquos account has no resourcesfor replying adequately to any of these standard counterexamplesalthough I do think that this is true in some cases I indicate where I

183


believe that this is so The counterexamples purport to show thatHempelrsquos account of explanation even if necessary could not besufficient However I argue that in thinking through an adequateresponse to the counterexamples we will see that Hempelrsquos requirementsare not even necessary for (full) explanation Those requirements weredescribed in chapter IV pp 138ndash9

The standard counterexamples irrelevance

The first reason (A) for holding that Hempelrsquos conditions forexplanation could not be sufficient turns on the fact that there can bederivations that meet all of Hempelrsquos requirements for D-N (or I-S)explanation but whose premisses are obviously irrelevant to theexplanation of the conclusions of those derivations In the main Ishall only be concerned in this chapter with the explanation of singularfacts (as I have so restricted myself throughout the book) but wemight note some counterexamples which concern the explanation oflaws as well Here is one taken from Ardon Lyon which concernsthe explanation of empirical laws by deductive subsumption4

(1) All metals conduct electricity (2) Whatever conducts electricity is subject to gravitational

attraction(3) All metals are subject to gravitational attraction

As Lyon points out no one would regard the conjunction of (1) and(2) as explaining (3) in spite of the fact that the latter does followfrom the former because (1) and (2) are irrelevant to the truth of (3)lsquoMetals are not subject to gravitational attraction because they conductelectricity non-conductors are subject to gravitational attraction tojust the same degreersquo (Lyon 1974247) Lyonrsquos counterexample isdirected against Hempelrsquos account of the explanation of laws but itis easy to construct a parallel counterexample to Hempelrsquos accountof the explanation of singular facts The explanandum in questionwould be that this bit of metal is subject to gravitational attractionand the explanans will include the fact that this bit of metal conductselectricity

Another alleged counterexample to Hempelrsquos analysis of theexplanation of laws is offered by Baruch Brody5

184


(1) Sodium normally combines with bromine in a ratio of one-to-one

(2) Everything that normally combines with bromine in a ratio of one-to-one normally combines with chlorine in a ratio of one-to-one

(3) Sodium normally combines with chlorine in a ratio of one-to-one

Brody claims that this derivation has no explanatory power whateverand I agree with him But even if the reader were to insist that it hassome such power it doesnrsquot have much and Hempelrsquos analysis doesnot offer us the materials for saying why that should be so AlthoughBrody does not say so one could say that the problem here too is oneof explanatory irrelevance The ratio in which bromine and chlorinecombine is surely irrelevant for explaining (but not necessarilyirrelevant in other ways) the ratio in which sodium and chlorinecombine even though the two ratios are related in a lawlike mannerAs with Lyonrsquos counterexample it is simple to convert Brodyrsquoscounterexample to one concerning the explanation of a singular factthe fact that this bit of sodium combined with this bit of chlorine in aone-one ratio

Two further counterexamples which I wish to mention are specificallydirected to irrelevance in the case of the explanation of singular factsThe first example is adapted from Peter Achinstein6 Suppose that poorJones (he is so often ill) eats at least a pound of arsenic and dies withintwenty-four hours and that eating at least a pound of arsenic inevitablyleads to death within twenty-four hours Does it follow that the argumentbelow is an explanation of Jonesrsquos death

(1) Jones ate at least a pound of arsenic at time t (2) (x) (x eats at least 1 lb arsenic at tx dies within 24 hours after t)

(3) Jones dies within 24 hours of t

Suppose consistently with the above suppositions that Jones wasrun over by a bus and died soon after ingesting the arsenic In thiscase the deduction will not be explanatory since Jones although hewould have died from the arsenic had he not been run over by a bussoon after eating the poison was actually killed by the bus It is thebus and not the arsenic which explains his death in spite of theargument given above meeting all of Hempelrsquos conditions

One can generalize Achinsteinrsquos example to any case in which thereis causal pre-emption Suppose some event e has two potential causes c

185


and d in the sense that c occurs and causes e and that d also occurs anddoes not cause e but would have caused e if c had not occurred d is apotential alternative cause of e but is pre-empted by the actual cause c7

In any such case there will be an Achinstein-style counterexample to theD-N account of the explanation of singular facts since there will be aderivation (with all true premisses etc) to the explanandum via a premissset which includes a premiss about the pre-empted cause but not oneabout the actual cause and hence no explanation of the explanandum soderived The pre-empted cause is explanatorily irrelevant to theexplanandum thus derived

I do take the lesson of this counterexample to be important so it willbe worth dwelling on it Is there a way of meeting this allegedcounterexample from the existing resources of Hempelrsquos theory Onemight think that it can be met by the introduction of a ceteris paribusclause in the statement of the law (2) and the addition of a further premiss(which will in this case be false) that says that other conditions are in factequal8 So the lsquoirrelevant explanationrsquo since it includes a false premisswill fail to be an explanation on Hempelrsquos own account After all therejoinder goes no one can die who is already dead the arsenic will bewhat kills Jones only if he hasnrsquot already died from some other causeThe arsenic ingestion is relevant only if the ceteris paribus clause in thelaw is met and the clause will exclude the case in which an alternativecause operates

I fail to see how the ceteris paribus clause response will meet thedifficulty at hand A ceteris paribus clause is inserted in a law as a meansof saving an apparently falsified law from real falsification other thingsare not equal so the law is true after all However in the counterexampleJonesrsquos being run over by a bus does not even apparently falsify the lawthat whoever eats at least a pound of arsenic dies within twenty-four hoursAfter all after eating the arsenic Jones did die within the required timeperiod So how could Jonesrsquos bus-related death present any kind of evenapparent difficulty for the law about what happens to people after theyingest at least a pound of arsenic Any difficulty for that law must involvesomeonersquos failure to die in some circumstances or other and poor deadJones is no example of that

In general when c (the bus hitting Jones) causes e (the death of Jones)there is no argument from this fact to the falsity of the law that whenevera D (an ingestion of at least a pound of arsenic) then an E (a death) Inparticular one does not need to rephrase the law as Whenever a D thenan E unless there is some alternative cause that operates to bring about

186


an E It is true that whoever eats a pound of arsenic at t dies within twenty-four hours even when sometimes death of arsenic ingestors is actuallybrought about by buses or something else

One further reply to this counterexample might dispute that (2)correctly expresses the intended law Suppose we interpret the law asitself including a causal claim eating a pound of arsenic causes deathwithin twenty-four hours If laws are universally quantifiedgeneralizations (remember that we are assuming throughout that thisis so) how should we represent lsquoeating 1 lb arsenic at t causes deathwithin 24 hoursrsquo in such a way that it would retain an explicit causalclaim Perhaps in this way (x) (x eats at least 1 lb arsenic at eating atleast 1 lb arsenic at t causes xrsquos death within 24 hours) There may besome other way in which to capture the causal claim in an explicitway within the universally quantified generalization but I cannot seewhat it might be

This generalization is falsified by the case in which Jones eats thearsenic but the bus causes his death so a ceteris paribus clause wouldhave to be inserted into it after all If this is the law it surely intends toassert that onersquos eating that much arsenic will cause death unlesssomething else causes it instead The qualification lsquounless somethingelse causes it insteadrsquo would be included in the ceteris paribus clauseThe law should therefore be expressed as (2) (x) (x eats at least 1 lbarsenic at t amp ceteris paribusxrsquos eating at least 1 lb arsenic at t causes xrsquosdeath within 24 hours) The explanatory argument which uses (2) wouldhave to include an additional premiss (2) Other things are equal If thebus and not the arsenic kills Jones (2) would be false and so the argumentwould fail to be explanatory on Hempelrsquos own account Can we concludethen that on this view of what the law is the ceteris paribus strategycould handle the arsenic-and-bus counterexample to Hempelrsquos accountafter all

I think not for two reasons First this strategy is simply not availableto Hempel No supporter (like Hempel) of the orthodox view of lawswould accept (2) as giving the correct form for a causal law Secondthere are additional problems about what the explanandum would be which(2) would help to explain the explanandum certainly would not be asgiven by (3) lsquoJones dies within 24 hours of trsquo The explanandum explainedby (2) could only be (3) lsquoeating at least 1 lb arsenic caused Jonesrsquosdeath within 24 hours of trsquo One might wrongly suppose that this willpresent no difficulty for Hempel since (3) follows from (3) one explains(3) and (3) follows from (3) hasnrsquot one explained (3) as well

187


As Peter Lipton has pointed out9 this assumption is not availableto Hempel Hempelrsquos D-N model of explanation is itself not closedunder logical entailment Suppose conclusion c is derived from andexplained by law L and initial conditions i The disjunction i or clogically follows from c But the explanation of c by the conjunctionof L and i cannot on Hempelrsquos account be an explanation of i or cbecause L is not essential to the derivation of i or c from the conjunctionof L and i

I have no doubt there is some way to handle the arsenic-and-bus casebut the introduction of a ceteris paribus clause into the law is simply notit Nor do I think that there are any resources available in Hempelrsquos accountas it stands for satisfactorily dealing with it

The arsenic-and-bus counterexample is interesting for anotherreason It provides an additional example of the asymmetry betweenexplanation and prediction Someone who produces the aboveargument (1)ndash(3) cannot be said to have explained Jonesrsquos death buthe certainly will have been able to predict it successfully He predictsthat Jones will die and his prediction is correct Moreover he hasoffered excellent grounds for his prediction Given that Jones drankthe arsenic the predictor could be certain that Jones would die Onecan predict via a pre-empted cause even though one cannot explainvia one Any rejoinder which wishes to claim that the above argumentyields neither a successful prediction nor a successful explanation willowe us a fuller account of successful prediction than has been thusfar at any rate provided

A second example of explanatory irrelevance which constitutes acounterexample to Hempelrsquos analysis of explanation of singular facts isone taken from Wesley Salmon10

(1) Every man who regularly takes birth control pills avoids pregnancy (2) John Jones has taken his wifersquos birth control pills regularly

(3) John Jones avoided becoming pregnant in the past year

The same sort of case can be made out for someone lsquowho explainsthe dissolving of a piece of sugar by citing the fact that the liquid inwhich it dissolved is holy waterrsquo A sentence which states the factthat the sugar dissolved in that liquid can be derived from but hardlyexplained by sentences stating the fact that the liquid is holy waterand the relevant law connecting water and the dissolution of sugarThe fact that the water is holy water is not relevant to the explanation

188


of the dissolution If (1) above is rephrased as a stochastic rather thanas a deterministic law it will serve as an irrelevance counterexampleto the Hempelian account of I-S explanation

A determined advocate of Hempelrsquos models of explanation might tryto insist that the inclusion of explanatorily irrelevant material in theexplanans in Salmonrsquos counterexamples might make the explanationspoor(er) but that they are still explanations none the less In chapter I Idistinguished between cases in which an explanation is bad and cases inwhich there is no explanation at all and in chapter V I applied thatdistinction specifically to the example of the inclusion of explanatorilyirrelevant information in the explanans I argued the following case inchapter V that someone is a man who takes birth control pills entails thatthe person is a man and the personrsquos being a man explains why thatperson does not become pregnant but the personrsquos being a man whotakes birth control pills does not explain in the least why the person doesnot become pregnant

I agree with Salmon about this and my discussion of the inclusionof irrelevant properties within fact locutions in chapter V was intendedto support his view The richer information has the explanatorilyrelevant information buried in it its being water is included in its beingholy water the personrsquos being a man is included in the personrsquos beinga man who takes birth control pills But the richer information doesnot explain some explanandum just in virtue of the fact that the weakerinformation which it includes and hence entails does explain it Theadditional information which makes it richer but which is explanatorilyirrelevant overrides and kills the explanatory power of the weakerinformation when it is added to it As Salmon said irrelevance is fatalto explanation

The examples which I group under (A) all teach the same lesson Therecan be derivations which meet all of Hempelrsquos conditions for theexplanation of a singular fact but whereas they are wonderful derivationsthey offer no explanation of what is derived This is because the premissesare explanatorily irrelevant to the conclusion or contain misleadingexplanatorily irrelevant additional information even though they do implythe conclusion

One might have hoped to explicate this concept of explanatoryrelevance as statistical relevance (as Salmon once did) but this seems ahopeless task The thought might be that a manrsquos taking birth controlpills is statistically irrelevant to his becoming pregnant since if oneconsiders the set of men who do take these pills and the set of men who

189


do not the incidence of pregnancy is the same to wit nil So the regularingestion of birth control pills by a man fails the test of statisticalrelevance and hence might thereby be thought to fail the test forexplanatory relevance

However the imposition of statistical relevance has a number ofunacceptable consequences Consider this argument which is due to JohnMeixner11 Assume that the following argument is an explanation at somelevel although admittedly not a very powerful or deep one of the factthat this sample of material dissolved in water (If the argument is not anexplanation of its conclusion it certainly does not fail to be one as aconsequence of the statistical irrelevance of the premisses)

(1) All salt dissolves in water (2) This sample is salt

(3) This sample dissolves in water

But if statistical relevance were added as an additional necessarycondition for explanation the above argument would not beexplanatory If this sample is salt then it has a physical probability nof dissolving in water If this sample had been baking soda orpotassium chloride it would have had the same probability n ofdissolving It is statistically irrelevant to dissolution whether thesample is salt or potassium chloride or baking soda Moreover to saythat the sample is salt is more informative than to say only that it iseither salt or baking soda or potassium chloride

If explanatory relevance were just statistical relevance it wouldtherefore not be possible to explain why this sample dissolved in wateron the grounds that it was salt since the fact that the sample was saltincludes additional statistically irrelevant information just as we cannotexplain why this lump of sugar dissolved on the grounds that it was placedin holy water since the fact that the water is holy water is statisticallyirrelevant additional information

If we take statistical relevance seriously then the only grounds thatwill do in the explanation of why this sample dissolved in water will beeither (a) that it was [salt v potassium chloride v baking sodahellip] or (b)that it was a substance which had a certain molecular structure m all andonly samples of which (including salt baking soda potassium chlorideetc) dissolve in water

The first horn of the dilemma (a) seems unacceptable I can explainthe dissolution of the material by its being salt without having to include

190


all of the other disjuncts which have the same probability of dissolutionThe fact that this sample was salt surely does explain its dissolving inspite of the fact that when compared to a samplersquos being one of theother substances it is statistically irrelevant that it is salt The secondhorn (b) is equally unacceptable Of course science strives for depthin explanation no doubt it is true that a deeper explanation in terms ofmolecular structure is a better explanation than the shallow explanationin terms of the materialrsquos being salt But we can explain the dissolvingof this substance in water on the grounds that it is salt when we do notknow what the relevant molecular structure is Moreover when we cometo know what molecular structure m is the materialrsquos having m may bea better explanation than the one in terms of the materialrsquos being saltBut a less good explanation is still an explanation When we possess thebetter explanation it does not follow even then that its being salt is nolonger an explanation at all on grounds of statistical irrelevance Thelesson of Meixnerrsquos discussion is this we can sometimes explain withinformation some part of which is statistically irrelevant to what weare explaining so explanatory relevance cannot be understood asstatistical relevance12

If applied to Achinsteinrsquos arsenic-and-bus case Meixnerrsquos argumentwould have an even more telling point to it Surely we can explainJonesrsquos death by the ingestion of arsenic when that is what kills himand by a large and fast-moving bus when it is that which does thedirty work The probability of Jones dying after ingesting a pound ofarsenic is let us say only 098 Suppose also that coincidentally theprobability of Jones dying after getting hit by a large fast-moving busis 098 Whether it is a bus or the arsenic is therefore statisticallyirrelevant to Jonesrsquos dying so each disjunct would be ruled out ashaving explanatory power on its own on the statistical relevance theoryOnly the disjunction itself which includes all the disjuncts which givethe same probability of death will be explanatory or if not thedisjunction then some very vague formulation as lsquosomethinghappening which gives a 098 probability of dyingrsquo Yet this seemswrong Each of ingesting arsenic and being run over by a bus canexplain death when appropriately cited on its own whether or notdying has the same statistical probability on both

191


The standard counterexamples symmetry

The second reason (B) for holding that Hempelrsquos conditions couldnot be sufficient for singular explanation has to do with lsquoexplanatoryrsquosymmetries Hempelrsquos account of singular explanation in terms ofderivability from true empirical premisses permits intuitivelyobjectionable cases in which (part of) the explanans can be explainedby the explanandum as well as explain it How can we amend theaccount so that such symmetries of lsquoexplanationrsquo will not arise

Both James Woodward and Peter Achinstein have argued (or implied)that the explanation relation is not an asymmetric relation as is usuallysupposed and that there are or can be bona fide cases of acceptablesymmetrical explanation explanatory mutual dependence between twosingular facts13 However both would of course concede that there aresome cases in which symmetrical explanation must be ruled out (ie inthe case of causal explanation) The explanation relation even if notasymmetric is surely not symmetric If not asymmetric it must be non-symmetric This is enough for my argument here All the examples I shallconsider in this part of the chapter are cases in which symmetricalexplanations are intuitively unacceptable I do not need to retain thestronger claim that the explanation relation itself is asymmetric At theend of chapter VII I offer (following Achinstein) an example of what Ithink is a bona fide case of symmetric explanation and show why andhow symmetric explanations may sometimes be acceptable

There are a number of these lsquosymmetryrsquo counterexamples whichchallenge Hempelrsquos account of singular explanation many of which derivefrom Sylvain Bromberger and Michael Scriven14 We have already touchedon some of these examples in the discussion of Hempel There are reallytwo kinds of cases that generate these unacceptable symmetries Firstthere are equations which show that the numerical value assumed by someproperty of a system at time t is a function of the values assumed by otherproperties of a system at time t or an earlier time t- (Ohmrsquos law Hookrsquoslaw the Boyle-Charles laws for ideal gases the length and period of apendulum)

Second there are laws with biconditionals which can include casesboth of laws of coexistence and of laws of succession A barometer fallsiff a storm is approaching the light received from the galaxies exhibitsa shift towards the red end of the spectrum iff the galaxies are recedingfrom us and (Aristotlersquos case) a planet twinkles iff it is not near To

192


this we can add Salmonrsquos confused rooster who explains the rising ofthe sun on the grounds of his regular crowing15 These equations orbiconditionals will allow the derivation of the height of the flagpolefrom the length of the shadow and the length of the shadow from theheight of the flagpole the length of the pendulum from its period andits period from its length the approaching storm from the fall in thebarometer as well as the fall of the barometer from the approachingstorm the receding of the galaxies from the red shift as well as the redshift from the recession of the galaxies the rising of the sun from thecrowing of the cock as well as the crowing of the cock from the risingof the sun

But in each of these pairs the first derivation would be non-explanatory the second explanatory Equations and biconditionals permitsymmetric derivations but since at least these examples do not providesymmetric explanations there must be more to singular explanation thanwhat Hempelrsquos theory thus far allows

Hempel as we saw lsquodealtrsquo with this by suggesting that there may notreally be true biconditionals in such cases (he supposed it will be recalledthat there might be cases of Koplik spots without measles) But what wehave to establish is how given that there may really be true biconditionalsor equations of this kind which allow derivations lsquoin both directionsrsquo weare able to distinguish the explanations from the derivations which fail toexplain

A proposed cure and its problems the causal condition

It is not a novel thought that the cure for the problems of irrelevanceand symmetry (A) and (B) that Hempelrsquos analysis of D-N explanationfaces (at least for the explanation of singular facts explanation oflaws would be quite a different matter) is to be found by stipulatingthat the premisses include something about the cause of the event tobe explained This was Aristotlersquos suggestion in chapter III for theexamples of the twinkling planets and the deciduous vines Millrsquosofficial theory which requires that the premisses include the statementof a causal law has similar resources for dealing with thecounterexamples At least some explanations are on such an accountdeductively valid arguments with true premisses which have empiricalcontent one of which is a lawlike generalization (thus far AristotleMill and Hempel can agree) but also one of which mentions or

193


specifies in some way the cause of the explanandum event (the finalrequirement would have to be added to the Hempelian account but isalready explicit in the accounts of the other two)

How would the causal requirement help with the problem of symmetryGiven the angle of the sunrsquos elevation it is the height of the flagpole thatcauses the length of the shadow and not vice versa the change inatmospheric pressure that causes the rise or fall of the barometer and notvice versa the receding of the galaxies that causes the red shift and notvice versa The causal requirement will also help with irrelevance It wasthe bus but not the arsenic his being a man but not his taking birth controlpills the substancersquos being water but not its being holy water which iscausally relevant to the death of Jones the pregnancy failure and thedissolution of the sugar So causation seems a way both to rule outsymmetric lsquoexplanationsrsquo (anyway where these are unwelcome) andirrelevant lsquoexplanationsrsquo

One might doubt whether causation will in fact help with irrelevanceSuppose we have a jar in which there is some sugar We add to the sugarsome water appropriately blessed by the local priest What caused thedissolution of the sugar In part its being immersed in the water Butthe sample of water just is a sample of holy water so if the immersionin the water caused the dissolution then the immersion in the holy watercaused it If the immersion in the water not only caused but also explainsthe sugarrsquos dissolution doesnrsquot the immersion in the holy water explainit too

No for we have distinguished in chapter V between causation andcausal explanation It is true that the immersion in the water and hence inthe holy water causes the dissolution of the sugar But it is the fact that itwas immersed in the water in the jar that causally explains the fact that itdissolved and even though the water is holy water the fact that it wasimmersed in the holy water in the jar is a different fact from the fact thatit was immersed in the water in the jar The fact that it was immersed inthe holy water in the jar introduces a feature that the other fact does notintroduce And that additional feature the waterrsquos being holy is causallyirrelevant to the dissolution and hence irrelevant to the explanation of thedissolution A similar diagnosis will be available in the other cases ofexplanatory irrelevance we have looked at The purported explaining factsintroduce features which are causally irrelevant to what is being explained

Many contemporary writers have converged on the necessity ofincluding such a causal requirement Thus Salmon reversing his earlierattempts to explicate explanation on the basis of statistical relations and

194


without mention of causation says that lsquoThe explanatory significance ofstatistical relations is indirect Their fundamental import lies in thefacthellipthat they constitute evidence for causal relationsrsquo and lsquoThe timehas comehellipto to put the ldquocauserdquo back into ldquobecauserdquorsquo16 Or lsquoTo givescientific explanations is to show how events and statistical regularitiesfit into the causal network of the worldrsquo (Salmon 1977162) Othersincluding Baruch Brody have hit upon the same idea of supplementingHempelrsquos account with some sort of causal information17

The difficulty with this otherwise extremely attractive view has beenpointed out by Timothy McCarthy18 It is easy to construct examples ofderivations which meet all of Hempelrsquos conditions plus the conditionthat there be a premiss which mentions the actual cause of the event to beexplained but which still fail to be explanatory McCarthy has given severalsuch examples

His first example (slightly amended) is this Let e be any event letlsquoD(e)rsquo represent any sentence describing e and let lsquoC(e)rsquo be a sentencewhich describes c ersquos actual (and not its pre-empted potential) cause (c isdescribed under its causally relevant description) Let represent any lawutterly irrelevant to the occurrence of e (It wonrsquot matter if you want tostrengthen the requirement and make the law a causal law) Finally let obe any object such that Ao Consider the following derivation

(1) (2) C(e) amp Ao (3) ~B(o) v ~C(e) v D(e)

(4) D(e)

This derivation of lsquoD(e)rsquo from premisses (1)ndash(3) meets all of Hempelrsquosconditions + the suggested causal supplement ersquos cause is describedby lsquoC(e)rsquo in premiss (2) Moreover lsquoC(e)rsquo is essential to the derivation(as is the law) Yet no one would say that we have here an explanationof e because even though c ersquos cause is described in or mentionedby a premiss it is not made causally and hence explanatorily relevantto ersquos occurrence There is still a notion of lsquoexplanatory relevancersquothat lsquoderivation + mention of cause of what is to be explainedrsquo simplyisnrsquot getting at As McCarthy says

One might suppose that the idea is to mirror the causal dependence ofe on its cause by the deductive dependence in d [the derivation] of adescription of e upon a description of ersquos cause That is an interesting

195


idea immediately however we may begin to suspect a gap in theargument The basic worry may be put in this way why should it followmerely because a D-N derivation of a sentence describing e ineliminablyinvolves in some way or other a description of ersquos cause that thisdescription functions in the derivation to show (causally) why e occursNo obvious reason exists why a D-N derivation of a sentence describinge could not depend on a description of ersquos cause in some way quiteunrelated to the causal dependence of e on that cause

(McCarthy 1977161)

McCarthy shows that various attempts to outmanoeuvre this objectionwill fail In particular his argument can be sustained even if anadditional condition due to Kim is imposed19 That condition is thislet all the singular sentences in the premisses be put in completeconjunctive normal form Then the condition requires that none ofthose singular sentences is a logical consequence of the explanandumitself However the following derivation meets all of Hempelrsquosrequirements + the causal requirement + Kimrsquos conjunctive normalform condition In the derivation below lsquoC(o)rsquo describes the cause oforsquos turning black which let us suppose is orsquos being immersed in abucket of black paint

(1) All crows are black (2) (x) (y) (x turns the colour of y amp y is black x turns black) (3) C(o) amp Henry is a crow (4) ~C(o) v o turns the colour of Henry

(5) o turns black

Even though (1)ndash(4) meet all of Hempelrsquos requirements + the causalsupplement + Kimrsquos condition no explanation of (5) has been given

There is surely something right in the demand that lsquocausersquo be put backinto lsquobecausersquo But what has gone wrong in the above examples Tosimplify in both derivations call the cause lsquocrsquo and the effect to beexplained lsquoersquo Although it is true that one of the premisses in both of theabove derivations says that c occurs and although it is true that this premissis essential to the derivation no premiss asserts of c that it is the cause ofe The derivation gets us as it were to ersquos occurrence from crsquos occurrencenot via the fact that c causes e but rather via a law irrelevant to crsquos causinge There is no connection between c and e other than that of logicalderivability of the latterrsquos description from the formerrsquos (plus an irrelevant

196


law) and that type of connection simply isnrsquot enough to ensure explanationof the conclusion by the premisses As McCarthy puts it

The reason is precisely that the logical dependence of lsquoD(e)rsquo on lsquoC(e)rsquohas nothing at all to do with the causal dependence [and hence theexplanatory dependence] of e on the event described by lsquoC(e)rsquo becausethe law mediating the deductive relation between lsquoC(e)rsquo and lsquoD(e)rsquo iscausally irrelevant to the occurrence of e20

In the note to the preceding sentence I argue that various furtherattempts to strengthen the causal requirement which require that thelaw not be irrelevant to the occurrence of the effect will still leave uswith non-explanatory derivations

There is a very simple way to bring the cause and the explanandumevent together in the right and relevant way in order to ensure explanationnot by including as a premiss a singular statement which merely describesor mentions the cause of the explanandum event e but rather by includingas a premiss a singular statement which asserts of that cause that it is thecause of e The relevant premiss in McCarthyrsquos arguments would say forexample not only that c occurs but also that c is the cause of e21 If thiswere added it seems that the derivation would become explanatory Andsurely it is this that is lacking in McCarthyrsquos examples which accountsfor the fact that they are not explanatory This simple and expedient methodavoids all the difficulties we have found in trying to capture explanatorydependence or relevance by logical dependence of conclusion onpremisses Explanatory dependence at least in this example is capturedby an explicit statement of the causal dependence of the effect on thecause Why just mention the cause in one of the premisses Why shouldnrsquota premiss actually assert the causal dependence of explanandum event onexplanans event

I do not think that every such additional premiss must use the wordlsquocausersquo The premiss might assert that e occurs because c occurs or thatthe reason for e is c or some such22 In so far as we are here restrictingourselves to singular causal explanation all of these will be ways of sayingroughly the same thing The point is this the premiss under considerationwill have to itself assert the dependence of effect on cause and thisdependence cannot be captured by logical dependence The occurrenceof the expression lsquohellipis is the cause ofhelliprsquo although frequently the way inwhich this is done is hardly essential (remember that throughout I assumethat the descriptions in the causal claim are the ones relevant for

197


explanation) other alternative expressions like ones which use lsquobecausersquoor lsquois the reason forrsquo and which also capture this sense of non-logicaldependence will do equally well

However there are at least two important consequences of this lastsuggestion that we must note First Hempelrsquos (and Millrsquos) requirementthat there be a lawlike generalization in the premisses which is essentialfor the derivation is rendered unnecessary On the suggestion beingcanvassed we have in the argument a premiss that explicitly says thecause of the explanandum event is such-and-such and that premiss byitself will entail the statement that the explanandum event occurred withoutthe addition of any further premisses at all In particular no premiss statinga universal general fact no law will be required for the derivation of theexplanandum So the first consequence is the redundancy of laws in (atleast some) explanations

There is a second important consequence of this suggestion Why thinkof explanations as arguments at all True we could think of the explanationas an argument with a single premiss

(1) c is the cause of e(2) e

But the derivation of lsquoersquo from lsquoc is the cause of ersquo is trivial It issimpler and nothing is lost if we think of this explanation ascomposed of a singular sentence lsquoc is the cause of ersquo (or lsquoe becauseof crsquo etc) Since in fact all of the premisses save this one will beredundant the explanation really just consists in the one remainingsentence that says that the cause of the event to be explained wassuch-and-such

Deductivism and probabilism agreed that all full explanations arearguments if McCarthyrsquos argument and my elaboration of it above aresound then at least sometimes full explanations are not arguments butsentences McCarthyrsquos argument in conjunction with my suggestion forremedying the defect to which it points does not show that fullexplanations are never arguments that conclusion would be too strongBut I would go further typically full explanations are not arguments butsingular sentences or conjunctions thereof

Is construing a specific bit of discourse as a sentence rather than anargument simply a matter of personal aesthetic preference on my partMcCarthyrsquos argument and my subsequent remarks were intended tomotivate the choice of sentence over argument The explanation must

198


explicitly include some word like lsquobecausersquo lsquoreasonrsquo lsquocausesrsquo etc andit is just this that the idea of an explanatory argument was meant to avoidby attempting to capture the dependence which such expressions get atby the idea of deductive or inductive logical dependence of a conclusionon premisses We have seen how this strategy fails and have seen thatonly explicit assertions or statements of the relevant dependence will doHence such explanations typically consist on my view of sentences ratherthan arguments

Let me mention one not very promising line of reply to this Is thereany real difference between an argument theory and a non-argument (orspecifically a sentence) theory Isnrsquot the difference between an argumentand a sentence theory somewhat superficial There is indeed a way totrivialize the distinction between an argument and a sentence Anyargument can be rewritten as a conditional sentence with the premissesas the antecedent and the conclusion as consequent Such a conditionalsentence if true is necessarily true The explanatory sentences envisagedby a non-argument theory if true are contingently true Explanations aretypically contingently true sentences or conjunctions thereof The sentencelsquoo is G because o is F and all F are Grsquo is if true contingently true eventhough the corresponding assertion of entailment lsquoif all F are G and o isF then o is Grsquo is a necessary truth

Moreover any attempt to minimize the difference between an argumenttheory and a non-argument sentence theory works more to my advantagethan to Hempelrsquos It is a doctrine central to Aristotlersquos Millrsquos andHempelrsquos accounts that explanations are arguments In so far as thedistinction between an argument and a sentence is minimized it is a centraldoctrine of theirs whose importance is being reduced

We have at a sweep a convincing reason for dismissing anyargument theory of explanation whether deductivist or probabilist(We still have the choice between certainty high and low epistemicprobability theories of explanation the first two being the non-argument analogues of deductivism and probabilism) In particularthis criticism strikes at the very heart of the Mill-Hempel theory andthe Aristotelian theory of scientific explanation for all three thinkersheld that all full explanations were deductive or inductive argumentsThese accounts of explanation not only fail to offer sufficientconditions for full explanation but more importantly they fail even toprovide necessary ones The criticism is not that explanations are notjust arguments but rather arguments plus something moreexplanations are typically not arguments at all

199


McCarthyrsquos reason although the only one I here discuss is not theonly one advanced by non-argument theorists for not taking explanationsto be arguments Achinstein offers two reasons against construingexplanations as arguments the illocutionary force view (which I havealready discussed and dismissed in chapter I) and the problem ofaccounting for emphasis23 Salmonrsquos reason among others is ratherdifferent an argument theory of explanation cannot deal with explanationswhich confer low probabilities There is no such thing as an argumentwhose conclusion has a low probability on the premisses so if there arelow epistemic probability explanations at least they cannot be arguments24

I remain uncommitted concerning Achinsteinrsquos second and Salmonrsquosreasons for adopting a non-argument theory of explanation

If explanations are typically not arguments what place do laws havein explanation Can we argue that since explanations typically are notarguments therefore explanations typically do not include laws AlthoughI do believe that many full explanations do not include laws I do notthink that the absence of laws from even some explanations at all followsfrom the fact that some explanations are not arguments

The requirement that explanations always include at least one lawlikegeneralization has been closely bound up with argument theories ofexplanation That is to say if all explanations were deductively valid orinductively good arguments they would (given the addition of some furtheruncontroversial assumptions) have to include a lawlike generalization asa premiss25 But the inverse is not true it does not follow from the factthat not all explanations are arguments that a law is not a part of everyfull explanation It only follows that if laws are a part of full explanationswhich are not arguments the idea of their parthood in such cases is not tobe cashed out as that of a premiss in an argument For example suppose(S) is an explanation of why e happened (S) lsquoe occurred because of thefact that c occurred and that whenever a C an Ersquo (S) is a sentence not anargument and yet it includes the statement of a law

However McCarthyrsquos example in conjunction with my additionalremarks about the solution for the difficulty he detects and Scrivenrsquosexample below also convincingly show that laws are not part of everyfull explanation in any sense of parthood The idea that full explanationsdo not always include laws (and therefore are not always arguments) isnot a novel one In different ways and from different points of view RyleScriven Salmon and Achinstein (and others too the list is not intendedto be exhaustive) have said this or similar things about the role of lawsin explanation26 For example in numerous papers Michael Scriven said

200


things similar to what I would wish to maintain about the role of laws orgeneralizations in explanation (although I do not need to agree with anyof his specific examples) In lsquoTruisms as the Grounds for HistoricalExplanationsrsquo he defended the view that the following was a perfectlycomplete or full explanation as it stood the full explanation of why (a)William the Conqueror never invaded Scotland is (b) that lsquohe had nodesire for the lands of the Scottish nobles and he secured his northernborders by defeating Malcolm King of Scotland in battle and exactinghomagersquo (Scriven 1959444) The explanation (b) is a conjunctivestatement formed from two singular statements and contains no lawsExplanations which lack laws are lsquonot incomplete in any sense in whichthey should be complete but certainly not including the grounds whichwe should give if pressed to support themrsquo (p446) Notice that Scrivencan be taken as making a weaker and a stronger point (a)rsquos full explanationwhatever it is includes or may include no law (b)mdashwhich includes nolawmdashis (a)rsquos full explanation I agree with the weaker of Scrivenrsquos pointsthere are some full explanations which do not include laws and (a)rsquos fullexplanation is likely to be such an example I do not necessarily agreethat (b) is (a)rsquos full explanation I return to this distinction below

Scrivenrsquos example above is an explanation of a human action It issometimes argued in the case of human actions that they are explicablebut anomic The thought here is rather different Human actions mightbe perhaps must be nomic law-governed The first of Scrivenrsquos claimsis that although or even if human actions are always nomic sometimesthe laws or lsquotruismsrsquo which lsquocoverrsquo them form no part of their fullexplanation27

Scriven makes it clear that he intends the point as a point aboutexplanation generally not just as a point about the explanation of humanaction

hellipabandoning the need for lawshellipsuch laws are not available even inthe physical sciences and if they were would not provide explanationsof much interesthellip When scientists were asked to explain the variationsin apparent brightness of the orbiting second-stage rocket that launchedthe first of our artificial satellites they replied that it was due to itsaxial rotation and its asymmetry This explanationhellipcontains no laws

(Scriven 1959445)

I have been arguing that some full explanations do not include lawsBut laws are still important even to those cases of explanation which

201


do not include them in other ways Indeed the argument view byinsisting that laws are a part of every full explanation has tended toneglect the other ways in which laws are important to explanationLet me add some remarks about how laws are still important for theexplanation of the world about us all consistent with my above claimthe remarks will also permit me to sharpen my view somewhat on therole of laws and generalizations in explanation

First to repeat what I mentioned above I have argued that there aresome full explanations of which laws form no part in any sense Butmany full explanations do include laws and this seems to be especiallyso in the special sciences Indeed this is one way in which actualexplanations whether lsquoidealrsquo or not in science and ordinary affairstypically differ Explanations in science typically include relevant lawsalthough even when this is so their inclusion in the explanation will notnecessarily be as a major premiss of an argument lsquoo is G because o is Fand all F are Grsquo is a (contingently true) sentence which includes a law butis not an argument

Second laws are important for the resolution of many types ofpuzzlement Clearly citation of an appropriate regularity can show thatthe phenomenon about which I may be perplexed or puzzled is in anycase not atypical or extraordinary or irregular in any way Given Millrsquosview of the epistemic circularity of deduction it was not easy to seewhy he thought explanations had to be deductive arguments with atleast one lawlike premiss One line of response I proposed on his behalfwas that what a covering law lsquoexplanationrsquo of for example the Duke ofWellingtonrsquos mortality could do was to show how the good Dukersquosmortality fits into a pattern of nature the deductive lsquoexplanationrsquo placeshis mortality within the context of a wider generalization and hencewithin the context of a uniformity of nature I believe that Mill wasthinking along such lines as these since explanation for him was alwaysthe fitting of facts into ever more general patterns of regularity But theanswer that I gave on his behalf invites the following observationexplaining the Dukersquos mortality is one thing fitting his mortality into amore general pattern however worthy that may be is something elseTo learn that something is not irregular is not the same thing as to explainit Not all resolutions of puzzlement or perplexity are ipso factoexplanations

There is a third way in which laws can be important Does the explanansreally fully explain the explanandum Perhaps it is not adequate to explainit fully something may be missing How can I justify my claim that the

202


explanans fully does the job it is meant to do On Scrivenrsquos (1959446)view suppose I claim that the full explanation of e is c If I am challengedabout the adequacy or completeness of my explanation I can justify myclaim to completeness and thereby rebuff the challenge by citing a law(or truism) eg that all C are E (c being a C e being an E) This is whatScriven calls the lsquorole-justifying groundsrsquo that laws provide in supportof a claim that one has given a full explanation The law or truism canjustify my assertion that c is the full and adequate explanation of e withoutbeing part of that explanation Although Scriven does not say so therecan be no objection to offering the full explanation and the justificationfor its fullness in a single assertion but if this is done we should be clearthat what we have is a full explanation and something else and not just afull explanation

It is for this reason that I distinguished Scrivenrsquos weaker and strongerclaims above I agreed that a full explanation for (a) included no lawsbut I did not necessarily agree that (b)mdashwhich included no laws-was(a)rsquos full explanation The full explanation of orsquos being G is the fact thato is F only if it is a law that all F are G sans exception Suppose the lawin question is a more complex law which says (x) (Fx amp Kx amp Hx amp Jx

Gx) A full explanation of why o is G would be the fact that o isFampKampHampJ In this way my view of full explanation is in at least oneway close to Hempelrsquos in spite of my rejection of his or any argumenttheory of explanation Full explanations on my view as on his maywell be close to ideal things if almost no one ever gives one that tellsus a lot about the practical circumstances of explanation-giving butprovides no argument whatsoever against such an account of fullexplanation

There may be perfectly good pragmatic reasons why we are entitled togive a partial explanation of orsquos G-ness it may be that orsquos being KampHampJis so obvious that one never needs to say anything more than that o is FBut the law (or lsquotruismrsquo) provides the criterion for what a complete or fullexplanation is I do not want to commit myself about the lsquofullnessrsquo ofScrivenrsquos explanation for Williamrsquos non-invasion of Scotland since thisraises issues about whether there are any laws which lsquocoverrsquo human actionsand which are also expressible in the vocabulary of human action itselfas Aristotle seemed to believe This would also involve a discussion ofhow lsquotruismsrsquo in Scrivenrsquos parlance differ from laws and I avoid thisissue here28

But to turn to his second example I am sure that the explanation ofthe variations in apparent brightness of the orbiting second-stage rocket

203


that launched Americarsquos first artificial satellite in terms only of its axialrotation and asymmetry cannot be its full explanation I agree that its fullexplanation whatever it is need not include a law but since the explanationScriven offers fails to contain any particular information about forinstance the source of light that was present it could not be a fullexplanation Scrivenrsquos own remarks about the role-justifying grounds thatlaws provide helps make this very point The particular explanation Scrivenoffers as full can be seen to be only incomplete not because it does notinclude a law but because the law provides the test for fullness whichScrivenrsquos explanation fails

Fourth on my view there is still a connection between singularexplanation and generality but not through the presence of a law Supposeit is argued that the following is a full explanation29 (F) object o is Gbecause o is F It seems to me that someone who insists that this cannot bea full explanation because of the absence of a law has to motivate thethought that (F) could not really be a full explanation by showing what itis that (F) omits which is not omitted once a law is added to theexplanation (Recall that we are already assuming that argument theoriesof explanation have been rejected so he canrsquot fault the absence of the lawon the grounds of non-derivability of explanandum from explanans withoutit) He must I think say this the real full explanation is only (FL) objecto is G because o is F and

But can we pinpoint what it is that the law is meant to add to (F) Whathas (FL) got that (F) lacks Return to the thought developed at length inchapter V that what matters to explanation are properties30 When orsquosbeing F fully explains orsquos being G it isnrsquot (to put it crudely) that orsquos beingF explains orsquos being G there is nothing special about o in any of thisRather it is orsquos being F that explains orsquos being G Explanatory impact iscarried by properties and there is generality built into the singularexplanation by the properties themselves without the inclusion of a lawThis implicit generality surely implies that other relevantly similar Gswhich are F will get the same full explanation that o got

Of course there is one obvious sense in which an explanation of orsquosbeing G in terms of orsquos being F could be incomplete The explanationmight fail to specify or cite all of the explanatorily relevant properties orcharacteristics of o But all of the relevant properties of o can be citedwithout inclusion of any law generalization

Suppose for the sake of argument that it is an exceptionless law ofnature that In this case the only property of o relevant forexplaining why o is G is orsquos F-ness In such a case it seems that orsquos being

204


G can be fully explained by orsquos being F What could the inclusion of thelaw or generalization add to the explanation that o is G because o is F

In lsquo rsquo the only information that could be relevant to theexplanation of orsquos being G is already given by the property linkage betweenbeing F and being G which is already expressed by (F) That part of theinformation in the generalization which is about (actual or possible) Fsother than o which are also G is simply irrelevant to the explanation oforsquos being G In short everything relevant to the explanation of orsquos beingG is already contained in (F) since that claim already makes the requisiteproperty connection between being F and being G Assuming that thegeneralization can connect properties at all (it is unclear that ageneralization can do this even when strengthened by a necessityoperator) what (FL) does that is not done by (F) is to extend the connectionto cases other than o And this canrsquot have any additional explanatoryrelevance to orsquos case The case of temporally and spatially distant F-objectswhich are G is surely not relevant to o One might ask about explanationthe question Hume asked himself (but believed he could answer) abouthis constant conjunction theory of causation lsquoIt may be thought thatwhat we learn not from one object we can never learn from a hundredwhich are all of the same kind and are perfectly resembling in everycircumstancersquo31

My view is even more radical than the suggestion that emerged inchapter IV that Mill could have considered a type of real explanationparallel to his account of the fundamental kind of real non-deductiveinference Such Millian considerations would certainly dispense with thegeneralization that all F are G in the explanation of o which is F beingG If the manhood of individual persons does not explain their mortalityhow could putting all the cases together as it were into a generalizationhelp get explanation off the ground How could a generalization havesome supervenient explanatory power that each instance of thegeneralization lacks

Although such a view dispenses with generalizations it does notdispense with the relevance to orsquos case of other Fs which are also Gs ThisMillian inspired view of explanation would retain as relevant to theexplanation of orsquos being G the F-ness and G-ness of other particulars ae i u etc Mill thought that we could (really) infer (and let us supposeexplain) the Dukersquos mortality not from a generalization but from hisresemblance to other individual men who were mortal Yet it is hard tosee how if the Dukersquos manhood cannot explain his mortality introducingthe manhood and mortality of people other than the Duke (whether by a

205


generalization or by the enumeration of other particular instances) couldexplain it What is the relevance to the good Dukersquos mortality of themortality of men spatially and temporally far distant from him

On my more radical view neither the generalization that all Fs are Gsnor the F-ness and G-ness of other particulars is required to be any partof the full explanation of orsquos being G In the case being supposed theonly fact required for the full explanation of orsquos being G is orsquos being Feven though the generalization and the explanation of other particularsrsquoG-ness by their F-ness and so on are implied or presupposed by the fullexplanation of orsquos being G by orsquos being F The question Hume askedquoted above if it has any bite at all bites not only against a constantconjunction theory of causation (which brings a generalization intoprominence) but even against a weaker theory of causation which makespart of the analysis of an instance of a causal relation information aboutany other individual instances of that causal relation

Generalizations get their revenge

The above remarks attempt to spell out a number of ways in whichlaws and generalizations are important for explanation withoutnecessarily being part of them There is yet another way closelyconnected with the third and fourth ways mentioned above It issufficiently important to separate it from the others The point assumesthat things are explanatory only as described and hence builds onthe discussion of facts in the last chapter

Aristotle it will be recalled thought that laws provided the criteria forthe selection of the descriptions under which the explanans explains theexplanandum Why did the match light I struck it and my striking of thematch was let us suppose the penultimate thing that ever happened tothe match Or my striking of the match was the event that caused thematch to light Why then can I explain the fact that the match lit by thefact that the match was struck and not by the different facts that thepenultimate thing that ever happened to the match occurred or that thecause of its lighting occurred even though these three singular facts (thefact that the match was struck the fact that the cause of the matchrsquos lightingoccurred the fact that the penultimate thing that ever happened to thematch occurred) are all facts about the same causal event but differentlydescribed In virtue of which of the features of a cause is the cause fullyexplanatory of the effect

206


Aristotlersquos reply would be that the explanatory features are the oneslinked in a law (whether deterministic or stochastic) To be sure thatstrikings of matches are followed by lightings of matches is itself nolaw nor any part of a law of nature We must therefore extendAristotlersquos point to include not only features linked in a law but alsofeatures nomically connected in the appropriate way in virtue ofunderlying laws (more on appropriateness in chapter VII) In virtue ofthe underlying laws of physics and chemistry striking and lightingbut not for example being a penultimate occurrence and a lightingare nomically related It is not that the laws need be any part of theexplanation rather the laws provide the criteria for determining underwhich descriptions one particular explains another (which singularfact explains another) Laws permit selection of the vocabularyappropriate for singular explanation

The above allows me to make a closely related point about the role oftheories in explanation Scientists often cite theories in explaining aphenomenon For example the theory of gravity explains why the mooncauses the earthrsquos tides the law of inertia explains why a projectilecontinues in motion for some time after being thrown subatomic particletheory explains why specific paths appear in a Wilson cloud chamberAnd theories consist (perhaps inter alia) of generalizations But (a) itdoes not follow that theories are explanatory in virtue of their generality(b) nor does it follow that the way in which they are explanatory is in allcases by being part of the explanation I have already argued for (b) ButI now wish to argue for (a) Theories help to explain singular facts invirtue of supplying a vocabulary for identifying or redescribing theparticular phenomena or mechanisms at work which are what explainthe explanandum facts

The examples of lsquosyllogistic explanationrsquo that I used in my discussionof Mill might have struck the reader as exceedingly artificial whoeverwould have thought the reply might go that the Duke of Wellingtonrsquosmortality could be explained by his manhood and the generalization thatall men are mortal And in admitting that some explanations do includelaws (especially in the sciences) I gave this example lsquoo is G because o isF and all F are Grsquo These generalizations are lsquoflatrsquo in the sense that theyare simple generalizations that use the same vocabulary as do the singularexplanans and explanandum descriptions Flat generalizations do notcontribute at all to singular explanation

However from the fact that flat generalizations are explanatorilyuseless it hardly follows that all are What is needed so the reply might

207


continue are generalizations which employ a theoretical vocabularywith greater depth than lsquomanrsquo and lsquomortalrsquo Perhaps the vocabularyshould be in deeper terms that refer to the fragility of hydrocarbon-based life forms To explain why o is G in terms of orsquos being F if a lawis to be included typically a scientific explanation will cite a law with avocabulary which is different from and deeper than the vocabulary ofwhich lsquoFrsquo and lsquoGrsquo are part Only as such could the generalizations beexplanatory

And such a reply is correct But it confirms rather than disconfirmsmy view If generalizations or laws were always per se explanatory thenflat ones ought to help explain (perhaps not as well as deep ones but theyshould explain to some extent none the less) The fact that only ones thatare deep relative to the vocabulary of the explanans and explanandumsingular sentences (in general theories) will help explain at all is anindication that they are explanatory in virtue of offering a deepervocabulary in which to identify or redescribe mechanisms but not just invirtue of being generalizations And even so to return for a moment to(b) the generalizations that make up the wider or deeper theory may helpto explain by offering that alternative vocabulary and without being partof the explanation itself

I argued before that often full explanations do not include laws butthat they sometimes do especially in the special sciences When lawsare included within an explanation as they sometimes are the purposeof the inclusion is to introduce a vocabulary different from the one usedin the explicit descriptions of the particular explanans and explanandumevents On the one hand if the less deep vocabulary used to describethe particular phenomena were wholly expendable the theoreticvocabulary could be explicitly used to describe them and any mentionof the law would be redundant If on the other hand no deeper vocabularywere available there would be no purpose for a law to serve Laws findtheir honest employment in singular explanation in situations betweenthe two extremes when the less deep vocabulary used to describesingular explanans and explanandum is to be retained at that level but adeeper vocabulary is available and needs introduction

One important role that theories play in science is to unify superficiallydiverse phenomena32 In virtue of a unifying theory what seemed likedifferent phenomena can be brought under one set of deep structural laws

By assuming that gases are composed of tiny molecules subject to thelaws of Newtonian mechanics we can explain the Boyle-Charles law

208


for a perfect gas But this is only a small fraction of our total gainFirst we can explain numerous other laws governing the behavior ofgaseshellip Second and even more important we can integrate thebehavior of gases with the behavior of numerous other kinds ofobjectshellip In the absence of the theoretical structure supplied by ourmolecular model the behavior of gases simply has no connection atall with these other phenomena Our picture of the world is much lessunified

(Friedman 19817)

On my view there is a difference between unification and explanationUnification of a phenomenon with other superficially differentphenomena however worthwhile a goal that may be is no part of theexplanation of that phenomenon If other menrsquos mortality couldnrsquotexplain why the good Duke is mortal when his own manhood doesnrsquotthen the fragility of other hydrocarbon-based life forms couldnrsquotexplain the Dukersquos fragility or mortality when his own hydrocarbonconstitution doesnrsquot It doesnrsquot matter from the point of view ofexplanation whether there are any other phenomena which getexplained by the deeper vocabulary the point is that the vocabularygives a new and more profound insight into the phenomenon at handwhether or not the vocabulary unifies it with other phenomena

209

CHAPTER VII

A Realist Theory of Explanation

On Millrsquos official account of explanation all explanations of singularfacts seemed to require laws of efficient causality (although we notedthat there was some evidence that Mill himself was prepared toconsider the matter differently) Hempel on the other handspecifically allows for non-causal explanations of singular facts Platoand Aristotle used lsquocausersquo so widely that even though all explanationsinvoke a lsquocausalrsquo factor much more is included than Mill wouldcertainly have allowed Whose claim is (more nearly) correct

The question I wish to deal with in this chapter is the question of non-causal explanation We discussed in chapter VI symmetry and irrelevancedifficulties faced by Hempelrsquos account of explanation Causal asymmetrywill ensure explanatory asymmetry in those cases in which the asymmetryof explanation is thought to be desirable Causal relevance will also provideus with a way to ensure explanatory relevance So causation seems a goodbet for explaining explanation

But are all singular explanations causal explanations In my remarksthroughout the book I have moved rather freely between lsquoexplanationrsquoand lsquocausal explanationrsquo Indeed when I introduced various distinctionsbetween kinds of theories of explanation in chapter I I did so by adoptingan interim assumption all explanation of particular or singular facts iscausal explanation (this excludes of course the case of explanation oflaws by more general laws) It is now time to look at this question in amore sustained way

What hangs on this question I think that a great deal does I agreewith much of what Kim writes in lsquoNoncausal Connectionsrsquo and I applythe lesson it teaches to theories of explanation1 He there argues that

210


Events in this world are interrelated in a variety of ways Among themthe ones we have called dependency or determination relations are ofgreat importance Broadly speaking it is these relations along withtemporal and spatial ones that give a significant structure to the worldof events The chief aim of the present paper has been to show thatcausation though important and in many ways fundamental is not theonly such relation and that there are other such determinative relationsthat deserve recognition and careful scrutiny

(Kim 197452)

There appear to be dependency relations between events that are notcausal and as I shall argue universal determinism may be true even ifnot every event has a cause These non-causal dependency relationsare pervasively present in the web of events and it is important tounderstand their nature their interrelations and their relation to thecausal relation if we are to have a clear and complete picture of theways in which events hang together in this world

(Kim 197441)

Metaphysically Kimrsquos point is that the world is structured by variousdeterminative or dependency relations of which causal relations areonly a proper subset Not all metaphysical relations structure theworld in the relevant sense accidental correlation relations betweentypes of objects or properties are lsquorealrsquo metaphysical relations butthey result from the worldrsquos structure rather than help to structure itNotice that Kim unlike me does not distinguish between the conceptsof determination and dependency

On my view it is the presence of these lsquostructuralrsquo determinative (anddependency) relations that makes explanation possible They are not allthat is required for as I have stressed these are metaphysical relationsand explanation is an epistemological idea Conceptualization must beconsidered in any complete account of explanation as I have tried to doin chapter V Whether the explanation relation relates those real objectsor events directly or only relates statements or facts about them the basisfor explanation is in metaphysics Objects or events in the world mustreally stand in some appropriate lsquostructuralrsquo relation before explanationis possible Explanations work when they do only in virtue of underlyingdeterminative or dependency structural relations in the world

If the causal relation were the only type of determinative relation thereis then one might expect all singular explanations to be causal But if

211


Kim is right if there are other types of determinative relations they mightprovide the basis for non-causal singular explanations I would have calledthe theory of explanation I advance lsquoa determinative theoryrsquo to capturethis idea that explanation rests on appropriate metaphysical relations butI do not wish to beg the question of whether causation is a deterministicor nondeterministic idea To allow for the possibility of nondeterministiccausal explanation ie high or low dependency explanations (and indeedfor the possibility that there may be nondeterministic relations other thancausation that underpin explanations) I prefer to call the theory ofexplanation lsquorealistrsquo The idea behind this realist theory of explanation isthat explanation rests on real metaphysical relations whether they bedeterministic or nondeterministic ones I remain neutral in the disputebetween determinative high and low dependency theories of explanation

Are all singular explanations causal explanations

The literature seems divided in its answer to the above question VanFraassen for example argues that all explanation is by way of fittingthings into the causal net but lsquothe causal netrsquo is defined by him aslsquowhatever structure of relations science describesrsquo2 Van Fraassenseems quite uninterested in the details of causation trivially whateverscience reveals is causal in the only sense in which he appearsinterested Similarly John Forge attempts to salvage Salmonrsquos causaltheory of explanation by saying that lsquoa causal process is one governedby scientific laws (theories)rsquo3 If one adopted a concept of causationthat was this wide it would indeed be an easy task to show that allexplanation of singular facts was causal But the victory would bepyrrhic relying as it would on an unmotivated and ad hocunderstanding of causation

Salmon on the other hand defends the thesis that all scientificexplanation (that is singular scientific explanation I shall not alwaysrepeat this qualification in what follows) is causal explanation using forthe purpose a narrower and more plausible account of causation lsquoCausalprocesses causal interactions and causal laws provide the mechanismsby which the world works to understand why certain things happen weneed to see how they are produced by these mechanismsrsquo4 In a similarvein Richard Miller claims lsquoAn explanation is an adequate descriptionof underlying causes helping to bring about the phenomenon to beexplainedrsquo5 Although Millerrsquos account of the concept of causation is

212


unusually free of aprioristic restrictions the concept is based on a core ofcases and is extendible to further cases by rational procedures for suchextension So Millerrsquos conception of causation although malleable andadaptable is definite enough not to be amorphous and able to coveranything one could wish For Salmon and Miller unlike for van Fraassenall explanations are causal in a specific enough sense of lsquocausersquo so thatlsquoall (scientific) singular explanations are causal explanationsrsquo is somethingmore than just a definition or a tautology

Many perhaps even most other writers have disagreed with the claimthat ties explanation so intimately with causal explanation and haveproduced lists of apparent counterexamples to the thesis We shall belooking in some detail at a few of those counterexamples below Let megive a fuller flavour of this widespread disagreement by repeating a randomselection of the lists of these allegedly non-causal explanations

Philip Kitcherrsquos non-causal cases are the explanation of why neon ischemically inert by quantum chemistry and various explanations in formallinguistics6 Nancy Cartwright mentions generally explanations invokinglaws of association as non-causal lsquothe equations of physics hellip[for instance]whenever the force on a classical particle of mass m is f the acceleration isfmrsquo and the laws of Mendelian genetics7 Clark Glymour argues that there

remains however a considerable bit of science that sounds very muchlike explaining and which perhaps has causal implications but whichdoes not seem to derive its point its force or its interest from the factthat it has something to do with causal relations (or their absence)8

Glymourrsquos examples are all concerned with explaining gravitationand electro-dynamics on the basis of some variational principle andhe gives three examples of this Peter Railton says that lsquosome particularfacts may be explained non-causally eg by subsumption understructural laws such as the Pauli exclusion principlersquo9 John Forgereminds us that

helliplaws of co-existen ce are not ca usal lawshelliplaws of co-existence doin fact appear in scientific explanations Some of these explanationsare of considerable significance such as those involving applicationsof classical thermodynamics in chemistry10

What sort of argument should we accept as decisively defeating acausal theory of explanation How do we decide which of the above

213


cited examples are bona fide examples of non-causal explanationLet me mention four such lines of attack which I do not think will dothe job First Peter Achinstein provides a number of allegedcounterexamples to a causal theory of explanation some of whichare examples of the explanation of an instance of a law eg the factthat since c occurred e occurred by means of the law of which it isan instance (the law that Es when Cs) Letrsquos call these lsquoinstanceexplanationsrsquo Instance explanations in this sense are not argumentsbut sentences which assert that some singular relational or conditionalfact11 is an instance of a lawlike regularity Suppose there are theseinstance explanations as Achinstein asserts12 There is also the caseof the explanation of laws by more general laws (discussion of whichI have forsworn) Mill himself pointed out in a passage I earlierquoted that the relation of a generalization to one of its instances isnot the relation of a cause to its effect

But I do not think that we should accept any of these cases as a seriouscounterexample to a causal theory of explanation An upholder of a causaltheory of explanation like Salmon would rightly not be very impressedwith this the causal theory of singular explanation should be expressedin such a way that will allow for these types of explanation

Suppose the explanandum which figures in one of Achinsteinrsquos instanceexplanations is fRg (lsquoRrsquo stands for some relation we know not what asyet which I have thus far indicated by the rather anodyne lsquosincersquo andlsquowhenrsquo) Suppose the explanation is fRg because all Fs stand in relationR to Gs (as I mentioned in chapter VI the law will typically be expressedin a different vocabulary than is the singular claim one dependent onsome theory) Is that a non-causal explanation Surely we cannot telluntil we know for what relation lsquoRrsquo stands If it is a causal relation thenthe explanation is causal in an appropriately widened sense if not thennot The point of importance is this the fact that the explanation is aninstance explanation which cites a law as explanatory is simply irrelevantto the question of whether it is a causal explanation It is of course truethat a generalization or law never causes its instances but explaining aparticular relational causal fact as an instance of a causal generalizationcannot be a serious counterexample to a judiciously stated causal theoryof explanation

The second way in which I avoid a too easy victory over a causaltheory of explanation is this Many writers (Hempel Cartwright and Forgewere examples) dismiss the claim that all explanations are causalexplanations on the grounds that some explanations involve laws of

214


coexistence13 rather than laws of succession This dismissal assumes thatno cause can be simultaneous with its effect This seems an unwiseassumption to make about causation since it has often been questionedIt is certainly open to the defender of the causal theory of explanation toinsist that an effect can be simultaneous with its cause and such a rejoinderdoes not seem especially ad hoc or unmotivated

Any alleged example is bound to be controversial but two examplesof the simultaneity of cause and effect which are sometimes offered arethese First consider a rigidly connected locomotive and caboose Thelocomotive begins to move and the caboose begins to movesimultaneously Second when I force my fist into a pillow the impact ofmy fist creates a hollow pocket in the pillow and the impact of my fist onthe pillow and the creation of the hollow in it are simultaneous Or anywayso it might be argued There are of course alleged lsquomicrorsquo replies tothese examples14 Perhaps some of the replies are successful perhaps noneis I wish to avoid all of this controversy by eschewing this line of attackon the causal theory of explanation Of course if it is possible for a causeand effect to be simultaneous causal asymmetry cannot itself be explicatedas temporal asymmetry The view that allows simultaneous causes andeffects will have to find some other way in which to capture causalasymmetry

Third there are some cases of explanation which depend on laws whichseem to be non-causal laws of succession Explanations in such caseswill be explanations by an earlier singular fact of a later succeedingsingular fact where the relationships involved do not appear to be causalExamples are ones that utilize laws governing self-maintaining processeslike the law of inertia conservation laws and in general laws governingthe motions of objects

In order to handle such cases John Mackie distinguishes betweenimmanent and transeunt causation When a process is hindered orinterrupted lsquofrom the outsidersquo the external event is a transeunt cause ofthe later altered stages of that process On the other hand when a processcontinues uninterrupted the earlier stages of the process itself are theimmanent causes of the later stages

Mackie argues plausibly to my mind that laws like the inertia andconservation laws are causal laws namely laws of immanent causationTherefore we can say that explanations which presuppose laws like thelaw of inertia are immanent causal explanations15 I throw a projectileand it moves during the time interval tndashtrsquo at a certain velocity v Supposeno force acts on the projectile after its release It will travel at the same

215


velocity v in the interval tacutendashtacuteacute What causes it to travel with velocity vduring tacutendashtacuteacute Can we say that its travelling at v in the first interval causedit to travel at v in the second one

Mackie says that if a force acts on the projectile to slow it down thereis a transeunt cause lsquofrom the outsidersquo which in the circumstances of theprojectile moving at velocity v causes the slowing down of the projectileto less than v But if we take the absence of an external force as part ofthe circumstances then it would seem perfectly reasonable to say that thecause of the projectile moving at velocity v in the latter time period wasits moving with velocity v in the earlier time period The earlier stage ofthe process is the immanent cause of the latter stage

It is true says Mackie that we lsquoordinarily look for and recognize acause of a change in a process rather than for the mere continuance ofthe processrsquo

However while it seems strange to call this earlier phase a cause andwhile our reluctance to do so reveals something about our actual conceptof causing there are analogies which would justify our extending theexisting concept to cover thishellip The earlier phase of a self-maintainingprocess surely brings about or helps to bring about the later phase Ifthe concept of cause and effect does not yet cover them it should wecan recognize immanent as well as transeunt causation


We can speak of causation in the case in which there is a continuationrather than a change on this view Of what is it a continuance ratherthan a change The reply is motion or whatever other state of theobject is conserved through the relevant time period One thing thatMackiersquos view entails is that a cause can be an event like the movingat velocity v at an earlier time which is said to be the cause of itscontinuing to move with velocity v at a later time

Some have objected to this view on the grounds that pure motioncannot be a cause I can see no reason a priori to conclude that movingwith velocity v cannot be a cause Our ordinary conception of cause surelypermits causes of this kind It may be extraordinary to speak of causationwhere there is continuation rather than change but there is nothingextraordinary about movings being causes in the case in which the movingbrings about a change A standard lsquoscientificrsquo view of causal explanationmakes causal explanation lsquoappropriate when there is transference of energyand momentum in accordance with conservation lawsrsquo16 on such a view

216


Mackiersquos immanent causation is certainly a type of causation andexplanations which rely on these laws of succession are causalexplanations

We need not settle the question of whether the idea of immanentcausation generally or its application to these cases will ultimatelywithstand scrutiny The only thing we need to conclude is that the matteris sufficiently unclear for it to be unwise to rest a rejection of the causaltheory of explanation on such cases

Fourth and finally I will not rest my argument on examples takenfrom quantum mechanics These cases arise in discussions of quantummechanics and in particular of the so-called Einstein-Podolsky-Rosenparadox and the contributions to that problem by JSBell17 Put verysuccinctly assume two half-particles travelling in opposite directionsalong the x-axis from a singlet spin state According to quantummechanics if the measurement of the component of spin in one directionis +1 the measurement for the second must yield -1 and vice versaParadoxically the two particles can be separated by any distance andthe choice of which of the two particles on which the measurement isfirst to be made can be taken after the particles leave their singlet spinstate and the result will still be the same Moreover Bell showed in aseries of papers that an assumption of some hidden variable to accountfor these results is inconsistent with quantum mechanics (and relevantexperiments) Does the E-P-R paradox provide us with a case of non-causal explanation

There are two reasons why I do not pursue the question of whetherwe have in quantum mechanics a type of non-causal explanation Firstit is a matter of some controversy whether the idea of causation ismalleable enough to be employed in the description of the E-P-Rcorrelations Can there be non-local causation Can there be causationat a distance18 Second and more to the point it is unclear what lessonsthere are for explanation in this Suppose we reject the idea thatmeasuring the component of spin on one half particle can causallyinfluence the measurement of the component of spin we obtain on theother half-particle But perhaps explanation and causation still gotogether Can we even in the absence of one half-particle influencingthe other really explain one measurement by the other The correlationsof values obtained in the measurements of the components of spin ofthe two half-particles are certainly nomic but in spite of being nomicthey may fail to be explanatory in the absence of a causal mechanism19

All of this is highly contentious and justifies my neglect of quantum

217


mechanics in my discussion of non-causal types of explanation Butthere is nothing in my final thesis about non-causal explanation whichfollows in the next section that could not be amended to include thesecases if the reader insists that they do provide genuine cases of non-causal singular explanation

What would make an explanation non-causal

What sorts of cases if any should convince us that there are non-causal singular explanations if not these In particular what is theconcept of causation that is being used in either the assertion or thedenial of the causal theory of singular explanation I indicated earlierthat the van Fraassen-Forge concept of causation was too wide Inorder to meet the objections to a causal theory presented by laws ofcoexistence and (apparently) non-causal laws of succession I havehad to widen the idea of cause or any way argue that the concept ofcausation is wider than the opponent of the theory seemed willing toallow If one is allowed to widen the concept at will there could neverbe any definitive refutation of a causal theory of explanation Whereshall the limits of permissible widening be set

I assume without argument two features of (ordinary empirical)causation that are uncontroversial20 If we can argue against a causal theoryof explanation on the basis of them I believe that we will have produceda definitive argument against the view that all explanation of singularfacts is causal explanation The two features are these (1) nothing cancause itself (2) the causal relation is contingent I do not claim that (1)and (2) are logically independent (1) of course has been denied for thecase of allegedly necessary beings such as God or Nature-As-A-Wholeand what we might call lsquometaphysical explanationrsquo It is uncontroversialin its application to contingent beings and empirical explanation scientificand ordinary which is what is under discussion here

In (2) I intend lsquothe contingency of the causal relationrsquo in the sensethat if c causes e there is a series of metaphysically possible worlds vizone in which c occurs and does not cause e but causes something elseone in which c occurs and causes nothing one in which e occurs causedby something other than c and one in which e occurs caused by nothingThe requirement is Humeian in inspiration and I accept it There are twogrades of contingency that should be distinguished in what I have saidweak contingency says that it is possible that the cause have a different

218


effect and the effect have a different cause strong contingency says thatit is possible that the cause have no effect and the effect have no cause

The contingency is a metaphysical contingency and has nothing to dowith the descriptions one happens to use to refer to the cause and effectIt is sometimes said that the contingency or otherwise of the causal relationdepends on which descriptions of cause and effect are selected so thatfor example even if lsquoc causes ersquo is contingent lsquothe cause of e causes ersquo isnecessary This last claim is false for the relevant scope reading of thatassertion The claim lsquoConcerning the cause of e it caused ersquo ismetaphysically contingent since the event which was the cause of e mightnot have been What is necessary is merely this sentence

This necessity is not metaphysical butanalytic necessity

I now turn to some cases of explanation which I regard as successfulrefutations of a causal theory of singular explanation

Identity and explanation

No one as far as I know has ever disputed the claim that no (contingent)thing or event causes itself (1) above21 Causation in such cases mustbe a relation between two distinct existences Since there are cases ofempirical explanation in which there are not two distinct (or evendifferent) existences that figure in the explanans and the explanandumit follows that there are some cases of non-causal explanation22 Thesecases provide to my mind the least controversial examples of non-causal explanation Identity explanations presuppose that some lsquolevelrsquoof reality in some sense explains itself How this can avoid the evil ofself-explanation and what it commits us to as far as symmetricexplanation is concerned are issues which we shall have to discuss

Peter Achinstein has discussed cases of this sort and I owe muchof what follows to him23 Achinsteinrsquos examples of this type ofexplanation include explaining why the pH value of some solution ischanging on the grounds that the concentration of hydrogen ions whichthat solution contains is changing explaining why ice is water on thegrounds that it is H2O explaining why some gas sample hastemperature t on the grounds that its constituent molecules have a meankinetic energy m

In its simplest form we can sometimes explain why some particulara has property P by identifying P with a property Q which a also has In

219


a somewhat less simple form we can sometimes explain why a is P byidentifying a with the sum of its parts [bampcampd] and identifying P withsome property of the sum Q or sometimes with a property Q hadindividually by each member of the sum Achinstein argues that identityexplanations cannot be a species of causal explanation since the havingor acquiring of property P canrsquot cause the having or acquiring of propertyQ if P=Q It makes no difference to my argument whether these identitiesare metaphysically necessary or contingent

Temperature=mean kinetic energy (for some temperature t and somemke m having temperature t=having constituent molecules with mkem) I can explain a gasrsquos having a certain temperature t by its constituentmolecules having mean kinetic energy m and I can explain a change ina gasrsquos temperature by a change in the mean kinetic energy of itsconstituent molecules We explain in these cases not just by laws of thecoexistence of two types of phenomena but by property or type-typeidentities This kind of explanation relying as it does on identities cannotbe assimilated to causal explanation Identity is another of thedeterminative relations that structure the world and make for thepossibility of explanation

Just as not all statements of causal relation are explanatory (itdepends on how the cause and effect are described) so too not allidentity statements are explanatory Temperature t=mean kinetic energym temperature t=temperature t The second identity is not explanatoryThe explanatoriness of an identity like that of a causal relation alsodepends on how the things identified are described The apparatusdeveloped in chapter V permits us to avoid self-explanation In viewof the ontology of explanation for which I there argued self-explanation would have to mean explanation of a fact f by itself Interms of the identity conditions for particular changes since t=mkethe gasrsquos acquiring temperature t and its acquiring mke m is one change(or anyway letrsquos take this as uncontroversial to make the case forapparent self-explanation stronger) But in terms of the special orepistemicized facts that we have agreed that we need as the relata forthe explanation relation the fact that it has that temperature and thefact that its molecules have mke m are two distinct facts because evenif there is only one property involved it is apprehended orconceptualized in two different ways So no self-explanation isinvolved A particularrsquos having a property described or conceptualizedin one way can explain the same particularrsquos having the same propertydescribed or conceptualised in another way Explanation is an

220


irreflexive relation and a fortiori identity explanation is irreflexiveeven though identity is itself a reflexive relation

There can be explanations of the fact that a is P in terms of the fact thata is Q where P=Q even where Q and P are not related as micro-propertyto macro-property (this example is also due to Achinstein) For exampleI can explain the fact that a cow is a ruminant by the fact that the cowchews its cud Such cases have to do with the place of a thing or type ofthing within a system of classification Some may think to dismiss thissort of example by arguing that what is explained in such a case is nothingbut why the cow is called or classified as a ruminant not why it is aruminant This is not so If the explanandum were the cowrsquos beingclassified as a ruminant the explanans would have to include informationabout the classificatory scheme itself how such a scheme was adoptedand so on Someone who explains why the cow is a ruminant uses thatclassificatory scheme but does not in the explanans offer any informationabout it

Although I agree with Achinstein that this is a genuine sort ofexplanation the scientifically more interesting cases rely on micro-macro (or more generally whole-part) identities and it is hardlysurprising that this should be so It has long been the goal of scientificexplanation to explain by depth by identifying things with theirlsquounderlyingrsquo counterparts I have in mind here the sort of strategysketched in bold and optimistic strokes by Oppenheim and Putnam inlsquoUnity of Science as a Working Hypothesisrsquo24 On their view the unityof science is advanced by micro-reductions the ideal is to micro-reducethe science of social groups to the science for multicellular livingthings the latter to that for cells thence to molecular science andfinally to the science of atoms and elementary particles Such micro-reductions require the identification (or replacement) of the (non-observational) properties that figure in the reduced science by theproperties that figure in the reducing science and the lsquodecompositionrsquoof the entities of the reduced science into proper parts which are theentities of the reducing science

So understood two kinds of relations are required for micro-reductionproperty identities (unless replacement is the strategy to be adopted) andthe identification of the whole with the sum of its parts I have elsewhereexpressed my reservations about the possibility of the success of thisstrategy in the case of the social sciences25 One might be equally scepticalabout the adoption of this strategy for the putative explanation of themental by the physical However one need not sympathize with

221


Oppenheim and Putnamrsquos over-optimistic global enthusiasm for thisstrategy in order to see that the strategy of micro-reduction offers apowerful tool for explanation where it is appropriate

Letrsquos call explanations which make use of this micro-reductiveidentification strategy lsquomereological explanationsrsquo (lsquomereologicalrsquo coversnot only the whole-part relation between the entities but by a naturalextension of the idea of mereology also the micro-reductive identityrelation between the properties themselves) Mereological explanationsare the most important type of identity explanations The tradition thattakes this kind of explanation seriously has a long history I am thinkingof Hobbes for example with his stress on the resolutive-compositivemethod of science the idea that to understand something is to take itapart conceptually and then to put it back together again conceptuallyThis methodology of mereological explanation reaches back beforeHobbes to lsquoPaduan methodologyrsquo26 and before that to Aristotlersquos materialexplanation and to the pre-Socratics who wished to explain the nature ofthings in terms of some or all of the elements earth air fire and waterTo understand something is to understand its parts or components lsquoHowit isrsquo with the parts or components doesnrsquot cause lsquohow it isrsquo with thewhole which is the sum of those parts or components even though theformer can explain the latter

I think these mereological explanations are common both in scienceand in ordinary life It is important to see that explanations of thewhole by its parts are not confined to the special sciences their use inscience is a refinement of a very common and ordinary idea We takea complex and break it into its parts Like the whole the parts aresubject to changes and are in states We can then explain the states orchanges of the whole in terms of the states or changes in the partsConsider for instance an example originally due to UT Place lsquoHerhat is a bundle of straw tied together with stringrsquo27 I can explain whyher hat will not hold its shape on the basis of the floppy pieces ofstraw which make it up

By a quirk of intellectual fate what I am calling mereologicalexplanation embraces both Aristotlersquos material and formal explanationsHe of course thought of these as different but we do not The material isthe stuff out of which something is made The form is the essence of thething what makes it a such rather than a particular this But certainly bythe time of Lockersquos An Essay Concerning the Human Understanding thereal essence of gold for example was lsquothe constitution of the insensibleparts of that body on whichhellipall the other properties of gold dependrsquo28

222


Locke compares knowing the real essence of something were this possibleto knowing lsquoall the springs and wheels and other contrivances within ofthe famous clock at Strasburgrsquo So to know the essence becomes knowingthe inner constitution of a thing and this knowledge is inseparable fromknowing the parts or material (lsquothe contrivancesrsquo) from which it iscomposed

Unlike causal explanation identity explanations cannot guaranteeasymmetry Identity is itself of course a symmetrical relation As I stressedin my discussion of the irreflexivity of explanation it is only somethingas conceptualized in one way that explains the same thing conceptualizedin a different way But the irreflexivity of explanation will not help us toensure the asymmetry of explanation because sometimes an event orstate conceptualized in one way can explain itself conceptualized inanother and vice versa These symmetric explanations typically work invirtue of there being a theory (or classificatory scheme) in which an identityclaim employing both of the descriptions or conceptualizations isembedded

Consider the mereological identity between being water and beingcomposed of H2O molecules (this example is also due to Achinstein)If one assumes as background the theory which identifies variousordinary substances with chemically precise compounds and mixturesthen in the appropriate circumstances the fact that ice is water can befully explained by the fact that ice is H2O In other circumstances thefact that ice is H2O can be fully explained by the fact that it is water Itdepends on what is known and what needs explanation In virtue ofthe theory and the identities it contains a (full) explainer can move ineither of two explanatory directions The same theory permitssymmetrical full explanations in appropriately different epistemiccircumstances In this case unlike that of partial explanation epistemicand pragmatic considerations do not lead us to offer less than a fullexplanation but rather allow us to select the direction in which togive the full explanation

Are there other non-causal singular explanations

When an austere theorist surveys the relations in which objects orevents stand in the world he is happy with causation and identity butis sceptical about almost everything else The florid theorist thinksthat there are other determinative relations that lie somewhere between

223


causation and identity they are not as strict or tightly binding asidentity but not as loose or contingent as causation Cambridgedependency supervenience the by-relation (that relates actions) therelation between a disposition and its structural basis are furthersuggestions advanced by various florid theorists There is a great dealof controversy about each such alleged case In what follows I remainneutral between the two antagonists The purpose of the remainderof the chapter is to argue conditionally if there are any of these otherputative relations some may provide the basis for additional non-causal singular explanations But I do not mean to assertunconditionally that there are any additional examples of non-causalsingular explanation

So whether or not there are other cases of non-causal singularexplanation will depend I think on whether or not there are determinative(or dependency) metaphysical relations between objects events or statesother than causation and identity Kim certainly a florid theorist mentionsthese three as examples of non-causal determinative relations Cambridgedependency one action being done by means of another and eventcomposition The third event composition is similar to the ordinarymereological relation of a part to a whole but is defined for events ratherthan objects and therefore where the parthood in question is temporalrather than spatial Examples of the first two kinds rest on highlycontentious (but not obviously false) theses about event identity

An example of Cambridge dependency is this Xantippe became awidow in virtue of as a consequence of Socratesrsquo death An example ofan action being done by doing another is I open the window by turningthe knob If either of these has any consequences for a theory ofexplanation it will be the Cambridge dependency case Examples ofactions done by means of other actions lend themselves to explaining-how rather than explaining-why But the Cambridge dependency caseseems to have a clear relevance for explaining-why Socratesrsquo dyingexplains why Xantippe became a widow

Kim argues that the relation between the pair of actions related by thelsquobyrsquo relation and the relation between an event and the lsquoCambridgersquo eventwhich depends on it are neither causal nor relations of identity Letrsquosconcentrate on the Cambridge dependency case On Kimrsquos view Socratesrsquodying and Xantippersquos becoming a widow cannot be the same event onthe grounds that different properties are involved in the two descriptions(This argument rests on the fine-grained analysis of event identity whichI eschewed in chapter V) But even apart from this consideration there is

224


the problem of spatial location the first event occurred in the prison inwhich Socrates was being kept the second happened wherever Xantippewas when her husband died Since Socratesrsquo dying and Xantippersquosbecoming a widow occurred at different places by the indiscernibility ofidenticals they cannot be identical cannot be one and the same event

Nor he argues can the former be the cause of the latter They occurredsimultaneously and even if we accept the possibility of a simultaneouscause and effect since they happen at different spatial locations we wouldalso have to accept simultaneous causal action at a distance Moreover

it is difficult to think of any sort of contingent empirical law to supporta causal relation between the two events In fact the relation strikes usas more intimate than one that is mediated by contingent causal lawsGiven that Socrates is the husband of Xantippe his death is sufficientlogically for the widowing of Xantippehellip As As we might say in allpossible worlds in which Socrates is the husband of Xantippe at a timet and in which Socrates dies at t Xantippe becomes a widow at t

(Kim 197442ndash3)

So if Socratesrsquo dying and Xantippersquos becoming a widow are bothevents29 but are not the same event and if there is no causal relationbetween them and if the former explains the latter then they providean additional case of non-causal singular explanation

Another possibility for non-causal explanation centres on thesupervenience relation Kim has also written extensively about this Ifthere is such a metaphysical relation as supervenience distinct fromidentity (and causation) then it may provide some additional examplesof non-causal explanation Kim lists these as candidate cases ofsupervenience the mental on the physical epistemic features of beliefson their non-epistemic features counterfactuals on indicative facts thecausal on the non-causal relational on non-relational propertiesvaluational or moral properties on natural properties to which we canadd the social on the non-social or individual30 If the general idea ofsupervenience is to add anything lsquoextrarsquo for scientific and ordinaryexplanation it would be nice if examples of it had an a posteriori characterThe thought is this only those examples of supervenience which areknowable a posteriori could underpin any interesting empiricalexplanations31

However even if there is such a distinct metaphysical relation assupervenience in the list of alleged examples above the most obviously

225


a posteriori examples mental states on physical states and the socialproperties of something on its non-social properties or features are alsothe most controversial The idea of supervenience was first introducedwith regard to aesthetic and moral properties and these least controversialexamples are a priori in character32 Even the a priori cases would providesome sort of explanation but not the same kind as we have consideredhitherto Why was St Francis a good man Because he was benevolentWhy is that painting beautiful Because of its colour composition

I think that there are good grounds for doubting whether supervenienceis distinct from identity I am sympathetic to the view of John Bacon

Supervenience in most of its guises entails necessary coextension Thustheoretical supervenience entails nomically necessary coextensionhellipI suspect that many supervenience enthusiasts would cool at necessarycoextension they didnrsquot mean to be saying anything quite so strongFurthermore nomically necessary coextension can be a good reasonfor property identification leading to reducibility in principle Thisagain is more than many supervenience theorists bargained for Theywanted supervenience without reducibility reducibilityhellip33

The suspicion is that the whole metaphysical truth aboutsupervenience (eg of the mental on the physical) is that asupervenient property may not be identical with some single baseproperty but rather identical with a possibly infinite disjunction ofpossibly infinite conjunctions of such base properties If reducibilityis an epistemic idea reduction in such cases will be in principleimpossible But metaphysically supervenience would just be aspecially complicated case of identity

For the purposes of this chapter I need not decide whether the abovesuspicion is well-grounded or not My claim is conditional ifsupervenience is a metaphysical relation distinct from identity (andcausation) as a florid theorist would have it and if some cases ofsupervenience are explanatory then supervenience explanation would beanother type of singular non-causal explanation

Disposition explanations

A pane of glass is fragile a lump of salt is water-soluble In virtue ofthose properties each does or might do certain things The first breaks

226


when struck sufficiently hard the second dissolves when immersedin water Both have structural features which are the bases for thesedispositional features In the two examples of the glass and the saltthe relevant structures are microstructures In general there are threethings that might be considered in such explanations the structurewhich is the basis for the dispositional feature the dispositional featureitself and actual behaviour in which the dispositional feature manifestsitself

One might mean either of two things by lsquodisposition explanationrsquo Wecan explain actual behaviour by dispositional features and dispositionalfeatures by (micro)structure I shall concentrate on the second sort ofexplanation the explanation of why an object has a dispositional propertyin terms of its structural features It is only this type of explanation that Ishall mean by lsquodisposition explanationrsquo

Even Hugh Mellor who doubts that there is any philosophicallysignificant contrast between the dispositional and non-dispositionalproperties of things would agree that we sometimes explain propertieslike the property of being water-soluble in terms of properties like theproperty of having some specific micro-structure lsquoNo doubt there arevirtues in explaining properties of things in terms of other propertiesespecially in terms of those of their spatial partsrsquo34 His doubts concernthe traditional characterization of the distinction between thedispositional and the non-dispositional lsquoMy strategy will be to showthe offending features of dispositions to be either mythical or commonto other properties of thingshelliprsquo (Mellor 1974157) Others have defendedthat traditional distinction between dispositions and non-dispositionalproperties in terms of which properties of the first kind but not of thesecond logically entail subjunctive conditionals35 We can agree thatthere are some such explanations without committing ourselvesconcerning the nature of the distinction between dispositional and non-dispositional properties

(These disposition or structural explanations may simply be a typeof supervenience explanation and if the latter were a type of identityexplanation then disposition explanation raises no issue distinct fromthe ones already discussed in the section on identity explanation Oranyway so the austere theorist would have it Sugar is a molecularcompound salt an ionic one Both are water-soluble but in virtueof different microstructures Since two objects can have the samedispositional feature like water-solubility in virtue of two differentmicrostructural bases the identity if that is what it is would have to

227


be between the dispositional feature and the disjunction of thestructural ones)

Suppose that dispositions supervene on some structural basis and thatthis base-disposition relation isnrsquot just a special case of identity The floridtheorist would add even though these disposition explanations are notidentity or mereological explanations they cannot be causal explanationseither even if cause and effect can be coexistent On the florid theoristrsquosview why canrsquot the relation between (micro)structural base anddispositional property be causal

Letrsquos take as our example the explanation of the dispositional propertyof salt its water-solubility in terms of its microstructure The answer tothe above question has to do with the contingency of the causal relationRecall (2) above There were two grades of contingency to the causalrelation weak and strong The dispositional-structural property relationunlike causation fails strong contingency Dispositional properties likewater-solubility as a matter of metaphysical necessity have some structuralbasis there is no possible world in which an object can have a dispositionalfeature and no structural basis whatever for that feature36 There is nopossible world in which a lump of salt is just water-soluble and there beno structural properties of the lump of salt in virtue of which it is water-soluble The florid theorist says that there must be as a matter ofmetaphysical necessity some structural water-solubility-making propertiesof the salt

It is even more controversial whether the relation also fails weakcontingency (the florid theorist need not have a view about this in orderto distinguish causation from the structure-disposition relation) Is therea metaphysically possible world in which salt has the samemicrostructure as it does in this world but in virtue of that structurehas different dispositional properties Could it for example be water-insoluble in that possible world in virtue of the same microstructure asit has in this world That this is nomically impossible is not in disputethe question is whether it is metaphysically possible and this isdisputable

Still the fact that the structure-disposition relation fails to be stronglycontingent is by itself enough for the purposes of the florid theorist todistinguish it from the causal relation On the florid theoristrsquos view thedisposition-structure relation is neither the same relation as the identityrelation nor the same as the causal relation and this distinctivemetaphysical relation licenses further examples of non-causal singularexplanation

228


What kind of fact is the fact that salt has a certain dispositional featurelike water-solubility I believe that this fact is a singular fact but DavidLewis disagrees David Lewisrsquos view is that all singular explanation iscausal explanation He would agree with the florid theorist that theexplanation for why salt is water-soluble is not a causal explanation37

However he argues that the explanation of why some object has adispositional property is not an explanation of any singular fact at all(Lewis argues that the explanation is not an explanation of a singularevent but I have translated his thesis about singular events into theterminology of singular facts the point of his thesis is unaffected by thetranslation) Thus he claims that disposition explanation is nocounterexample to the thesis that all explanations of singular facts arecausal explanations

Disposition explanation on his view then is not singular explanationat all Rather it has this structure lsquoWhy is it that something is F BecauseA is F An existential quantification is explained by providing an instancersquo(Lewis 1986223) Lewisrsquos view is that in explaining for example whysalt is water-soluble (Lewisrsquos example is why Walt has smallpox-immunity) I explain (what I have called) an existentially general fact(and not a singular fact) Despite appearances according to Lewis if Iexplain the fact that salt is water-soluble I do not explain something withthe form lsquoFarsquo The explanandum has this form the fact that The explanans in order to count as an instance of the existentialquantification must therefore have the form Fa On Lewisrsquos view in acase in which I am explaining an existential quantification by providingan instance the property F whatever it is must appear both in theexplanans and the explanandum

How would this work for the case of the water-solubility of salt Sincethe saltrsquos micro-structure must somehow figure in the explanans Lewisrsquoslsquoarsquo must refer to that microstructure (Letrsquos call that microstructure lsquomrsquo)Since the explanans is lsquoFarsquo for what property of the microstructure doeslsquoFrsquo stand There are two possibilities to be considered the micro-propertyof making salt water-soluble or the micro-property of making salt dissolvein water

Clearly the second possibility is not available to Lewis If lsquoFrsquo standsfor the micro-property of making salt dissolve in water the existentiallygeneral explanandum must be the fact that there is something which makessalt dissolve in water This explanation is not a disposition explanation atall because the fact being explained is not dispositional on any view ofwhat disposition explanation is Built into the idea of a disposition is the

229


possibility that the behaviour in which it is manifested may never occurA disposition explanation explains why something would behave in acertain way if the appropriate conditions were ever realized thatexplanation may work even if there is no actual behaviour to explain Ifsalt never does dissolve in water despite its being water-soluble there isno possible explanation for why there is something in virtue of which saltdissolves in water because it doesnrsquot The property F must be adispositional property if the explanation is to be a disposition explanationof any sort

The first possibility was that lsquoFrsquo stood for the microproperty ofmaking salt water-soluble On this first alternative Lewisrsquos lsquoFrsquo standsfor the dispositional property of the microstructure makes salt water-soluble or perhaps for a dispositional relational property makes water-soluble (true for example of the ordered pair microstructure m andsalt) The explanandum would then be an existentially generaldispositional fact the fact that there is something which makes saltwater-soluble To explain why salt is water-soluble is really just toexplain why there is something which makes salt water-soluble Sofar so good

But on this first possibility what is the explanans The explananswould be m makes salt water-soluble That is to say microstructure mhas the property makes salt water-soluble What kind of property isthat It seems to be a dispositional property of the microstructureAccording to Lewisrsquos theory the explanans must be a singular factwith the form lsquoFarsquo But this explanans is also a dispositional fact sinceit attributes a dispositional property to something namely to themicrostructure Lewisrsquos proposal makes this singular dispositional factthe explanans for saltrsquos water-solubility microstructure m makes saltwater-soluble Since that fact attributes a (perhaps relational)dispositional property to microstructure m it must itself count as asingular dispositional fact

So Lewisrsquos view entails that there are some singular dispositionalfacts There is no inconsistency in his holding these two theses (1)there are some singular dispositional facts (2) all explainable (apparentlysingular) dispositional facts are really only existentially general factsBut the conjunction of the two implies that all genuinely singulardispositional facts are inexplicable The view seems entirely ad hoc andunmotivated On his thesis we know that there must be some genuinelysingular dispositional facts with the form lsquoFarsquo which are theexplanations for the genuinely existentially general dispositional facts

230


whatever they are But we could never know concerning some specificdispositional fact which appears to be singular whether it is genuinelysingular or only existentially general unless we know whether it is inprinciple capable of being further explained If an explanation is possibleit must be an existentially general dispositional fact after all despiteappearances only if an explanation of it is impossible can we admitthat it is a genuinely singular dispositional fact after all

To my mind this is all counterintuitive and needlessly baroque Ifwe accept that we can sometimes explain singular dispositional factslike the fact that salt is water-soluble the account is straightforwardThe explanans for this (truly and not just apparently) singulardispositional fact is a singular structural fact the fact that salt hasmicrostructure m Of course if we do accept this and if we retain Lewisrsquosadmission that this explanation is not causal we would also have toaccept that there are some non-causal explanations of singular factsand that therefore Lewisrsquos thesis that all singular explanation is causalis simply false

Again determinative high and low dependency explanations

I said in chapter I

It will be helpful in introducing this typology [of determinativehigh and low dependency theories of explanation] to assumesomething that I regard as false all explanations of singular eventsor states of affairs are causal explanations I will discuss thisassumption in chapter VII and broaden the kinds of singularexplanations that there can be It will then be easy to broaden thetypology to take account of this having already introduced it onthe narrower assumption But in the interim I will be making this(admittedly false) assumption

It is now time to make good my promise In what follows I mean therather bland word lsquothingrsquo to cover whatever the reader thinks thereis in the world apart from how we conceptualize or think objectsevents states structures properties relations and so on

In the cases of explaining singular facts so far discussed we explainedin one of at least three ways (1) we saw what makes something happen(2) we saw how what the thing is like structurally makes it have its

231


dispositional features and (3) we analysed or conceptually resolved theparticular to see what makes it what it is The lsquomakesrsquo here is ambiguousbetween lsquocausally makesrsquo lsquois the structural basis which makesrsquo38 andlsquomereologically makesrsquo All of these ideas have long traditions in thehistory of philosophy and of scientific thought Causes are events whichmake their effects occur structural features of a thing make it liable tobehave in certain ways parts and what they are like make up the wholeand make it what it is like

There is a unifying if ambiguous thought that unites all of thesecases explanations work in virtue of something determining or beingresponsible for something Explanations work only in virtue of thedeterminative relations that exist in the world The determinativerelations may be causal but they may also be whatever otherdeterminative relations there are between structure and dispositionalfeatures between an event and the Cambridge event which itdetermines between a thing or property and itself (but differentlydescribed or conceptualized)

There was an insight in the causal theory of explanation weexplain something by showing what makes it or what is responsiblefor it The fault of the causal theory of explanation was to overlookthe fact that there are more ways of making something what it is orbeing responsible for it than by causing it The general idea is theidea of determination we explain something by showing whatdetermines that thing to be as it is Causation is a particular kind ofdeterminative relation but not the only such determinative relation39

Causation was held to be a potential cure for both the ills ofirrelevance and symmetry which plagued Hempelrsquos account ofexplanation Just as the wider idea of determinative relation can curesymmetry where it is desirable to do so so too the wider idea willcure explanatory irrelevance If one thing is determined by anotherthe second is explanatorily relevant for the first on the other handif there are no determinative (or dependency see below) relationsbetween the things then they are explanatorily irrelevant to oneanother

However the above will not quite do for reasons I have given inchapter I I do not want to beg the question between determinative highand low dependency theories of explanation (and the consequentcommitment to a certainty HEP or LEP theory of explanation whichdepends on that choice) In terms of the argument of this book I wish toleave this an open question If there are nondeterministic causes and

232


one can explain in virtue of them then the explanatory idea of onething making another happen is not to be understood only in adeterministic sense

As I also said in chapter I I do not think that there are any othernondeterminative explanations other than those which would arise on thebasis of explanation by nondeterministic causes Since identity is ametaphysically necessary relation there is no room for mere dependencyin its case But if the reader can think of other candidates fornondeterministic relations that can be explanatory other thannondeterministic causation these too can be included in the view I hereadvance

When I discussed Aristotle in chapter III I said that he held (E) somethingcan be explained only by either its matter or its form or its purpose or itschange-initiator I then asked whether (E) was just an ad hoc disjunction orwhether Aristotle had some deeper reason for thinking that these four modesof explanation were exhaustive of the sorts of explanation there are I agreedwith Julius Moravcsikrsquos rationale for Aristotlersquos (E) for Aristotle a particularsubstance is a set of elements with a fixed structure that moves itself towardsself-determined goals The four elements in this definition are elementstructure motion originator and goal These correspond to and justify thefour types of explanation Since everything else that can be said to be is anaspect of substance the four types of explanation are both non-arbitraryand exhaustive (E) far from being ad hoc is the kernel of a theory ofexplanation

Kimrsquos remarks at the beginning of this chapter provide an analogousstrategy for deciding what types of singular explanation there can be forit is important as I have argued throughout the book to ground a theoryof explanation on a theory of metaphysics Metaphysically it is thisdeterminative (and possibly dependency) picture of the world that groundsexplanation of singular facts This is so even if the explanation relationitself has lsquoepistemicizedrsquo facts or statements or propositions as its relataExplanantia fully explain explananda only in virtue of how things reallyare Explanations work only because things make things happen or makethings have some feature (lsquothingsrsquo should be taken in an anodyne senseto include whatever the reader wishes to count as a denizen of reality)And the making can be taken either in a deterministic or in anondeterministic (dependency) sense

And this I think is the ultimate basis for any reply to an explanationtheorist who holds that full explanation is only and entirely a pragmaticor otherwise anthropomorphic conception On my view explanation is

233


epistemic but with a solid metaphysical basis A realist theory ofexplanation that links the determinative (or dependency) relations in theworld with explanation gets at the intuitively acceptable idea that weexplain something by showing what is responsible for it or what makes itas it is This is what in the end explains explanation

234

Notes

Chapter I Getting our Bearings

1 Karel Lambert and Gordon GBrittan Jr An Introduction to the Philosophyof Science third edition Ridgeview Publishing Company Atascadero 1987pp 14ndash17

2 Carl Hempel Aspects of Scientific Explanation Free Press New York 1965pp 335ndash6 Subsequent page numbers in my text following discussion ofHempelrsquos views throughout this book refer to this title unless otherwiseindicated

3 Michael Friedman lsquoExplanation and Scientific Understandingrsquo Journal ofPhilosophy vol LXXI 1974 pp 5ndash19 Quotation from p 5

4 Raimo Tuomela lsquoExplaining Explainingrsquo Erkenntnis vol 15 1980 pp211ndash43 Quote from p 217

5 Romane Clark and Paul Welsh Introduction to Logic Van NostrandPrinceton 1962 pp 153ndash4 The lsquodestruction at Rotterdamrsquo is their exampleFollowing Clark and Welsh I construe lsquoprocessrsquo sufficiently widely toinclude acts and activities

6 SBromberger lsquoAn Approach to Explanationrsquo in Analytical Philosophysecond series ed RJButler Blackwell Oxford 1965 pp 72ndash105 Quotationfrom p 104

7 Peter Achinstein The Nature of Explanation Oxford University Press NewYork 1983 see chapters 2 and 3 I have learned a great deal from Achinsteinrsquoswritings on explanation even on issues where I do not in the end agree withwhat he has to say Another example of an approach to explanation whichmakes explanatory acts the conceptually prior concept is to be found in RaimoTuomela op cit

8 An act of another illocutionary type to be precise For the distinction betweenillocutionary locutionary and perlocutionary acts see JLAustin How todo Things with Words second edition ed JOUrmson and Marina SbisagraveOxford University Press Oxford 1984 See especially Lectures VIII andIX pp 94ndash120

9 Illocutionary acts10 Illocutionary product11 Carl Hempel op cit p 41212 Ernest Sosa lsquoThe Analysis of ldquoKnowledge that Prdquorsquo Analysis vol 25 new

series no 103 October 1964 p 1

235

Notes

13 Edmund Gettier lsquoIs Justified True Belief Knowledgersquo Analysis vol 23June 1963 pp 121ndash3 and then by way of selected examples Michael ClarklsquoKnowledge and Groundsrsquo Analysis vol 24 no2 new series no 98December 1963 pp 46ndash8 John Turk Saunders and Narayan ChampawatlsquoMr Clarkrsquos Definition of Knowledgersquo Analysis vol 25 no 1 new seriesno 103 October 1964 pp 8ndash9 Keith Lehrer lsquoKnowledge Truth andEvidencersquo Analysis vol 25 no 5 new series no 107 April 1965 pp 168ndash75 and of course Sosa op cit

14 Michael Friedman op cit p 1315 See for example Peter Unger lsquoOn Experience and the Development of the

Understandingrsquo American Philosophical Quarterly vol 3 1966 pp 48ndash5616 Karl Popper lsquoEpistemology Without a Knowing Subjectrsquo in Karl Popper

Objective Knowledge Oxford University Press Oxford 1973 pp 106ndash52For quotes see pp 108ndash11

17 I speak in unorthodox terminology of a conceptrsquos intension (normally it iswords which have intensions) I mean by lsquointension of a conceptrsquo merely itsmodel ie the analysis of it

18 I have always liked the account of this by Stephen Toulmin Foresight andUnderstanding Harper New York 1961 and especially his sharp distinctionbetween understanding and foresight (prediction)

19 Carl Hempel op cit p 41320 Examples include Peter Achinstein op cit pp 15ndash73 Arthur Collins

lsquoExplanation and Causalityrsquo Mind vol 75 1966 pp 482ndash50021 Carl Hempel op cit p 41222 Hempelrsquos famous Deductive-Nomological and Inductive-Statistical models

are meant to provide two different sets of requirements for full scientificexplanation I discuss these models fully in chapter IV Hempel speaks of athird model the Deductive-Statistical but I ignore it here and elsewhere inthe book

23 Hilary Putnam Meaning and the Moral Sciences Routledge amp Kegan PaulLondon 1978 pp 41ndash2

24 David Lewis lsquoCausal Explanationrsquo in his Philosophical Papers vol IIOxford University Press Oxford and New York 1986 pp 214ndash40 Seeespecially pp 217ndash21 and 226ndash8

25 Hilary Putnam op cit pp 42ndash326 A full discussion of this issue would involve careful investigation of the

differences between sentences statements and propositions and of thequestion of which of the three logical relations like material implicationor strict entailment hold between But this would take us far off course letme here assume that it is sentences which entail etc other sentences

27 Almost uncontroversial since Peter Achinsteinrsquos theory of explanation mightcontrovert it See my review of his The Nature of Explanation in the BritishJournal for the Philosophy of Science vol 37 1986 pp 377ndash84

28 Carl Hempel op cit p 33629 Wesley Salmon Scientific Explanation and the Causal Structure of the World

Princeton University Press Princeton 1984 pp 15ndash1630 I draw the distinctions as I do because I think they help one to see what is

at stake in deciding between different theories of explanation Of course

236


there are many other (perhaps more illuminating for different purposes) waysin which to divide up the competing theories In particular my typologydiffers in important ways from a superficially similar one offered by Salmonin Scientific Explanation and the Causal Structure of the World pp 16ndash18

31 Carl Hempel op cit p 33732 See for example Brian Skyrms Choice and Chance Dickinson Publishing

Company Encino and Belmont California 1975 chapters I VI and VIIpp 200ndash3 JLMackie Truth Probability and Paradox Oxford UniversityPress Oxford 1973 chapter 5 David Lewis op cit 1986 Part 5 andespecially the classical source for the distinction Rudolf Carnap lsquoThe TwoConcepts of Probabilityrsquo Philosophy and Phenomenological Research volV 1945 pp 513ndash32

33 GH von Wright Explanation and Understanding Routledge London 1971p 13 Von Wright not surprisingly goes on to deny that there are any non-deductive explanations lsquoIt seems to me betterhellipnot to say that the inductive-probabilistic model explains what happens but to say only that it justifiescertain expectations and predicationsrsquo (p 14) See also Wolfgang StegmuumlllerlsquoTwo Successor Concepts to the Notion of Statistical Explanationrsquo in Logicand Philosophy ed GH von Wright Nijhoff The Hague 1980 pp 37ndash52As far as I know the best defence of probabilistic explanation is to be foundin Colin Howson lsquoOn a Recent Argument for the Impossibility of a StatisticalExplanation of Single Events and a Defence of a Modified Form of HempelrsquosTheory of Statistical Explanationrsquo Erkenntnis vol 29 1988 pp 113ndash24

34 Wesley Salmon RJeffrey and JGreeno Statistical Explanation andStatistical Relevance University of Pittsburgh Press Pittsburgh 1971 p64

35 Peter Railton lsquoA Deductive-Nomological Model of ProbabilisticExplanationrsquo Philosophy of Science vol 45 1978 pp 206ndash26 Quotationfrom p 216

36 Henry Kyburg Jr lsquoConjunctivitisrsquo in Induction Acceptance and RationalBeliefs ed MSwain Reidel Dordrecht 1970 pp 55ndash82

37 For example in Wesley Salmon Scientific Explanation and the CausalStructure of the World p 87 and in his lsquoA Third Dogma of Empiricismrsquo inBasic Problems in Methodology and Linguistics ed RButts and J HintikkaReidel Dordrecht 1977 pp 152ndash3

38 Colin Howson op cit pp 122ndash339 Wesley Salmon et al Statistical Explanation and Statistical Relevance pp

62ndash5 Scientific Explanation and the Causal Structure of the World p 4640 Salmonrsquos own exposition seems to use both epistemic and physical

probability I have set out the example trying to be clear about whichprobability is involved in the argument

41 Wesley Salmon Statistical Explanation and Statistical Relevance p 6442 Karl Popper The Logic of Scientific Discovery Hutchinson London 1972

pp 59ndash6043 This claim might be challenged in view of Peter Railtonrsquos D-N model of

probabilistic explanation I stand by my claim For Railton an explanationof why an event e lsquoimprobably took placersquo is the conjunction of a deductiveargument whose conclusion is that e had a low probability of occurrence

237

Notes

and lsquoa parenthetic addendum to the effect thatrsquo e occurred (op cit p 214)The conjunction of an argument and an addendum is not itself an argument

The conclusion of the argument on its own is not a sentence that assertsthat e occurred and so the argument by itself cannot be an explanation ofwhy e occurred Rather the conclusion of the deductive argument is only asentence assigning a probability of occurrence perhaps exceedingly smallto ersquos occurrence The argument on its own if it explains anything onlyexplains (with a conditional certainty) why e has some specific probabilityof occurrence lsquoDropping off the addendum leaves an explanation but it is aD-N explanation of the occurrence of a particular probability not aprobabilistic explanation of the occurrence of a particular decayrsquo (p 217)

44 Salmonrsquos view in lsquoA Third Dogma of Empiricismrsquo pp 149ndash66 is that lsquoanexplanation is an assemblage of factors that are statistically relevanthelliprsquo (p159)

45 Bas van Fraassenrsquos view lsquoAn explanation is not the same as a propositionor an argument or a list of propositions it is an answerrsquo Bas van FraassenThe Scientific Image Oxford University Press Oxford 1980 p 134

46 See for example his lsquoA Third Dogma of Empiricismrsquo47 David Lewis lsquoPostscripts to ldquoCausationrdquorsquo in op cit pp 175ndash84 Wesley

Salmon Scientific Explanation and the Causal Structure of the World pp184ndash205 Patrick Suppes Probabilistic Metaphysics Blackwell Oxford1984 pp 35ndash75 John Mackie The Cement of the Universe OxfordUniversity Press Oxford 1974 pp 39ndash43

48 Strong sufficiency is stronger than material sufficiency49 Wesley Salmon Scientific Explanation and the Causal Structure of the World

pp 185ndash9050 It is in this light that I understand the oft-cited case of paresis (and Salmonrsquos

example of mushroom poisoning after having ingested a certain type ofmushroom) Only a very small number of those with untreated latent syphilisdevelop paresis although the only way in which to get paresis is by havinguntreated latent syphilis However having untreated latent syphilis explainsgetting paresis although having untreated latent syphilis confers only a lowepistemic probability on that person having paresis All of this is consistentwith the certainty model if it is a partial explanation Presumably we believethat it is We believe that there is some set of conditions c perhaps unknownsuch that if one has untreated latent syphilis and is in condition c thengetting paresis is physically necessary And a full explanation of gettingparesis must refer both to untreated latent syphilis and conditions c Butthere is no reason why I cannot give a partial explanation of getting paresisjust in terms of having untreated latent syphilis

51 See Salmon Scientific Explanation and the Causal Structure of the Worldlsquohellipthe statistical relevance relations that are invokedhellipmust be explained interms of causal relations The explanationhellipis incomplete until the causalcomponentshelliphave been providedrsquo (p 22) lsquoIt now seems to me that thestatistical relationshipshellipconstitute the statistical basis for ahellip scientificexplanation but that this basis must be supplemented by certain causal factorsin order to constitute a satisfactory scientific explanationrsquo (p 34)lsquohellipstatistical statistical relevance relations are to be explained in terms of

238


causal relevance relationsrsquo (p 208) But the causation so evidenced mayitself be analysable in terms of statistical relevance relations lsquoI cannot thinkof any reason to suppose that ordinary causal talk would dissolve intononsense if Laplacian determinism turned out to be false I shall thereforeproceed on the supposition that probabilistic causality is a coherent andimportant philosophical concept In advocating the notion of probabilisticcausality neither Suppes nor I intend to deny that there are sufficientcauseshellip On our view sufficient causes constitute a limiting case ofprobabilistic causesrsquo (p 190)

52 References in the text to Salmon in what follows this note are to his lsquoAThird Dogma of Empiricismrsquo References to van Fraassen are to his TheScientific Image

53 I have discussed contrastive explanation in lsquoExplaining Contrastive FactsrsquoAnalysis vol 47 January 1987 pp 35ndash7 Peter Liptonrsquos reply (in lsquoA RealContrastrsquo Analysis vol 47 October 1987 pp 207ndash8) and Dennis Templersquosview (in lsquoThe Contrast Theory of Why-Questionsrsquo Philosophy of Sciencevol 55 1988 pp 141ndash51) are both discussed below

54 See for example Fred Dretske lsquoContrastive Factsrsquo Philosophical Reviewvol 81 1972 pp 411ndash37 Alan Garfinkel Forms of Explanation YaleUniversity Press New Haven 1981 from which the Sutton story is borrowedBas van Fraassen op cit Jon Dorling lsquoOn Explanation in Physics Sketchof an Alternative to Hempelrsquos Account of the Explanation of LawsrsquoPhilosophy of Science vol 45 1978 pp 136ndash40

55 Some (although not all) of van Fraassenrsquos alleged cases of explainingcontrastive facts can be dealt with by carefully distinguishing betweendifferent non-contrastive explananda Consider for example the differencebetween explaining why Adam ate an apple and why Adam ate the appleSee van Fraassen op cit p 127

56 David Lewis lsquoCausal Explanationrsquo in op cit pp 229ndash31 see alsolsquoCausationrsquo op cit p 177 On Lewisrsquos view a maximally true explanatoryproposition about an event is the proposition which gives the whole truthabout the entire causal history of the event (presumably stretching backto the beginning of the universe) An alternative might be to take themaximally true explanatory proposition as the one which gives the wholetruth only about the whole of the immediate cause of the explanandumevent

57 Dennis Temple op cit p 149

Chapter II Plato on Explanation

1 I am using RSBluck Platorsquos Phaedo Bobbs-Merrill Indianapolis 1955but checking that translation against the translation by Hugh Tredennick inPlato The Collected Dialogues ed Edith Hamilton and Huntington CairnsBollingen Foundation 1966

2 For this I use the Cornford translation in Hamilton and Cairns op cit3 Gregory Vlastos lsquoReasons and Causes in the Phaedorsquo Philosophical Review

vol 78 1969 pp 291ndash325

239

Notes

4 ELBurge lsquoThe Ideas as Aitiai in the Phaedorsquo Phronesis vol 16 1971pp 1ndash13

5 See for example David Melling Understanding Plato Oxford UniversityPress Oxford 1987 pp 11ndash12 for a brief discussion of this identification

6 Gregory Vlastos op cit7 I have discussed the distinction between real and so-called Cambridge change

in lsquoA Puzzle about Posthumous Predicationrsquo Philosophical Review volXCVII 1988 pp 211ndash36

8 MJCresswell lsquoPlatorsquos Theory of Causality Phaedo 95ndash106rsquo AustralasianJournal of Philosophy vol 49 1971 pp 244ndash9 Remarks relevant to thispoint on pp 246ndash7

9 Julia Annas lsquoAristotle on Inefficient Causesrsquo Philosophical Quarterly vol32 1982 pp 311ndash26

10 CCWTaylor lsquoForms as Causes in the Phaedorsquo Mind vol LXVIII 1969pp 45ndash59 His argument for this is on p 53

11 MJCresswell op cit pp 248ndash912 ELBurge op cit p 413 One obvious restriction on what can be included in the causally relevant

context and conditions is this no explicit causal information can be includedThat the token striking caused a lighting cannot be taken to be part of theconditions co-present with the token striking in other possible worlds inwhich it occurs If causal information of this sort were to be included itwould become trivially impossible to ask about other causes or effects thattoken event has in some other possible world

14 Mary Mackenzie lsquoPlatorsquos Analysis of Individuationrsquo unpublishedmanuscript

15 Hugh Mellor lsquoProbable Explanationrsquo Australasian Journal of Philosophyvol 54 1976 pp 231ndash41

16 The indicative mood counterparts of (10) and (11) which I have not botheredto list are trivial for the same sorts of reasons for which (4) and (5) weretrivial If d explains g then g has happened and ex hypothesi neither ~g norf can have happened One cannot explain what has not happened just assomething that has not happened cannot explain anything

17 There are complications here that we need not go into Since Mellor thinksthat causation is a deterministic concept (see Mellor op cit p 235) hethinks of high dependency explanation as explanation where there is nocausation at all or no causal explanation available

18 By this I mean strong sufficiency and not just material sufficiency Seechapter I Strong sufficiency requires the truth of certain counterfactuals

19 To reject (10) is certainly to reject a determinative view of explanationand hence to hold a determinative theory is to hold (10) But this doesnot follow to reject a determinative theory is to reject (10) Mellor unlikePlato holds a high dependency theory and he like Plato subscribes to(10) Indeed Mellor argues (correctly) that any low dependency theoryis inconsistent with (10) (and the addition of some uncontroversial furtherpremisses) If (10) is true then either a determinative or a highdependency theory of explanation is true a low dependency theory isfalse

240


20 In this and other of the more technical arguments in this chapter I am gratefulfor the patient help of Peter Milne

21 The idea of necessity here is strong necessity not material necessity A causeis strongly necessary for its effect iff if the cause had not in thecircumstances occurred the effect would not have occurred I discussLewisrsquos analysis of causal necessity and sufficiency in lsquoLewis and theProblem of Causal Sufficiencyrsquo Analysis vol 41 1981 pp 38ndash41 In thatarticle I did not take adequate account of the possibility of nondeterministiccausation and this is a flaw in what I wrote But I still believe that thedifficulty I claimed to find in Lewisrsquos account is still a difficulty for ananalysis of deterministic causation in the sense in which a cause is bothstrongly necessary and strongly sufficient in the circumstances for its effect

22 Much discussion of whether a cause is necessary in the circumstances forits effect centres on the possibility of causal pre-emption See for instanceWilliam Goosens lsquoCausal Chains and Counterfactualsrsquo Journal ofPhilosophy vol LXXVI 1979 pp 489ndash95

23 lsquoc occurs e has some chance x of occurring and as it happens e does occurif c had not occurred e would still have had some chance y of occurring butonly a very slight chance since y would have been very much less than x Wecannot quite say that without the cause the effect would not have occurredbut we can say that without the cause the effect would have been very muchless probable than it actually wasrsquo (David Lewis lsquoCausationrsquo PhilosophicalPapers vol II Oxford University Press Oxford and New York 1986 p176

24 John Watkins Science and Scepticism Princeton University Press Princeton1984 pp 227ndash8 I have substituted lsquodrsquo and lsquogrsquo for his lsquohrsquo and lsquoersquo

25 David Melling op cit p 136 construes the term in this way26 See Gregory Vlastos lsquoThe Third Man Argument in the Parmenidesrsquo

Philosophical Review 1954 and reprinted in Studies in Platorsquos Metaphysicsed REAllen Routledge amp Kegan Paul London 1967 pp 231ndash63

Chapter III Aristotle on Explanation

1 Richard Sorabji Necessity Cause and Blame Duckworth London 1980p 42

2 Julius Moravcsik lsquoAristotle on Adequate Explanationsrsquo Synthese vol 281974 pp 3ndash17 Quote from p 4

3 Julius Moravcsik ibid Max Hocutt lsquoAristotlersquos Four BecausesrsquoPhilosophy vol 49 1974 pp 385ndash99 Julia Annas lsquoAristotle on EfficientCausesrsquo Philosophical Quarterly vol 32 1982 pp 311ndash26

4 References to the Posterior Analytics are to the translation by JonathanBarnes Aristotlersquos Posterior Analytics Oxford University Press Oxford1975 but checked against (and occasionally taken from) The Basic Works ofAristotle ed Richard McKeon Random House New York 1966 Books Aand B of the Posterior Analytics refer to the Barnes translation Books I andII of the Posterior Analytics refer to the translation in McKeon Referencesto other of Aristotlersquos writings are to the McKeon edition

241

Notes

5 Jonathan Barnes trans op cit pp 215ndash166 The example comes from Karel Lambert and Gordon Brittan Jr An

Introduction to the Philosophy of Science Ridgeview Publishing CompanyAtascadero 1987 p 12

7 WWieland lsquoThe Problem of Teleologyrsquo reprinted in Articles on Aristotle1 Science ed Jonathan Barnes Malcolm Schofield and Richard SorabjiDuckworth London 1975 pp 141ndash60 Quote from p 147

8 See Julia Annas op cit p 3219 Peter Achinstein The Nature of Explanation Oxford University Press New

York 1983 pp 5ndash610 See for example GELOwen lsquoTithenai ta Phainomenarsquo reprinted in

Jonathan Barnes Malcolm Schofield and Richard Sorabji eds op cit pp113ndash26

11 Julius Moravcsik op cit12 That the four senses of lsquoWhyrsquo are non-overlapping is I think Wielandrsquos

view since he calls the unity provided by lsquoWhyrsquo a lsquoformal unityrsquo On theother hand Wieland also calls the question lsquoWhyrsquo lsquoa functional elementrsquowhich suggests that it is able to provide some unity more substantive than asyntactic unity for the four senses of lsquoexplanationrsquo Perhaps he thinks thatin spite of the four-way ambiguity of lsquoexplanationrsquo each of the four sensesof the term lsquoexplanationrsquo do share part of their meaning in common andthat this shared overlapping part is somehow accounted for by part of themeaning of the question lsquoWhyrsquo However Wieland nowhere develops thepossibility of overlapping meanings of the four senses and there is nothingobvious in Aristotlersquos text to support the thought

13 This I take to be the insight captured by Wesley Salmon in his ScientificExplanation and the Causal Structure of the World Princeton UniversityPress Princeton 1984 but neglected by both Peter Achinstein and CarlHempel the latter of whom concentrates almost exclusively on the epistemicrather than the metaphysical requirements of explanation

14 Why do I add the qualification lsquoif possiblehelliprsquo Perhaps it is not logicallyimpossible that there be a world that is inexplicable or in which there aresome inexplicable occurrences Perhaps things could happen that we couldnever understand This as I said in chapter I will depend on the theory ofexplanation one adopts The qualification is added in order not to beg thisopen question

15 Wesley Salmon op cit pp 240 27816 Compare his account at Physics II 5 lsquoBut secondly some events are for

the sake of something others not Again some of the former class are inaccordance with deliberate intention others not but both are in the class ofthings which are for the sake of something Hence it is clear that even amongthe things which are outside the necessary and the normal there are some inconnection with which the phrase ldquofor the sake of somethingrdquo is applicableThings of this kind then when they come to pass incidentally are said to beldquoby chancerdquorsquo

17 I am here deeply indebted to Richard Sorabji op cit pp 3ndash13 to whichwork the reader is advised to refer for detailed textual support Myinterpretation of these passages differs somewhat from his

242


18 Richard Sorabji op cit p 8 Do formal final and material aitiai alsonecessitate what they explain or is this only true of motion-originatorsAristotlersquos claim is limited to the accidentally generated and destroyed sothe necessitation might seem to be limited to the motion-originator HoweverJonathan Barnes trans op cit pp 215ndash16 argues that the matter of athing when appropriately described necessitates what it explains (Aristotlesays that the premisses are the matter or material explanation of theirconclusion and premisses necessitate their conclusion) A thingrsquos formnecessitates its being what it is three-sidedness necessitates somethingrsquosbeing a triangle Perhaps for Aristotle then all per se aitiai necessitatewhat they explain

19 There is some controversy as to whether the conclusions of such argumentsare propositions or imperatives but this does not affect my point

20 References in the text to PA are to the Posterior Analytics21 The interested reader might like to consult Jonathan Barnes trans op cit

p 184 and p 229 whom I have followed fairly closely on this issue for adiscussion of these passages and further references

22 For a discussion of this notion in Aristotle see David Hamlyn lsquoAristotelianEpagogersquo Phronesis vol XXI 1976 pp 167ndash84

23 Closer to the truth but not quite the truth since Aristotle has no account atall of the scientific explanation of particular cases

24 I deal with the difference between the non-symmetry and the asymmetry ofexplanation in chapters VI and VII

25 See Jonathan Barnes trans op cit pp 98ndash101 for a helpful discussion ofthis

26 Presumably lsquotheyrsquo refers to the premisses although this is a matter of somecontroversy

27 Baruch Brody lsquoTowards an Aristotelian Theory of Scientific ExplanationrsquoPhilosophy of Science vol 39 1972 pp 20ndash31 Discussed by TimothyMcCarthy lsquoOn an Aristotelian Model of Scientific Explanationrsquo Philosophyof Science vol 44 1977 pp 159ndash66 Nathan Stemmer lsquoBrodyrsquos Defenseof Essentialismrsquo Philosophy of Science vol 40 1973 pp 393ndash6

28 lsquoFor if an explanation requires premisses related to conclusion as cause toeffect and causes fall into four clearly recognizable types then we do havea non-circular criterion of explanationrsquo (Bas van Fraassen lsquoA Re-examinationof Aristotlersquos Philosophy of Sciencersquo Dialogue vol 19 1980 pp 20ndash45Quote from p 32)

Chapter IV Mill and Hempel on Explanation

1 Carl Hempel and POppenheim lsquoStudies in the Logic of Explanationrsquo inAspects of Scientific Explanation Free Press New York 1965 p 251 Allfurther page references to Hempel in my text are to this volume

2 John Stuart Mill A System of Logic Longman London 1970 Book IIIchapter XII section 1 p 305 References in the text to Mill are to A Systemof Logic numbers are to book chapter and section (in that order) or topage number as in the Longman edition 1970

243

Notes

3 Alan Ryan The Philosophy of John Stuart Mill Macmillan London secondedition 1971 chapter 1 pp 3ndash20

4 Also in Hempel op cit5 See Peter Urbach Francis Baconrsquos Philosophy of Science Open Court La

Salle 19876 Pierre Duhem The Aim and Structure of Physical Theory Atheneum New

York 1977 Quote from p 77 But Hempel ends his discussion of the problem of lawlikeness on an

optimistic note lsquoThough the preceding discussion has not led to a fullysatisfactory general characterization of lawlike sentences and thus of lawsit will I hope have clarified to some extent the sense in which those conceptswill be understood in the present studyrsquo (Hempel op cit p 343)

8 Or invariability of coexistence for the case of the explanation of laws Ideal with this question below

9 Also of course a particular token event of the antecedent type mentionedin the law for the explanation of a particular event

10 I have discussed these issues more fully in my lsquoCausal Scepticism or InvisibleCementrsquo Ratio vol XXIV 1982 pp 161ndash72

11 The example is Millrsquos For textual accuracy and despite my own reservationsabout the offence that it might cause I have retained it

12 Ernest Nagel The Structure of Science Harcourt Brace amp World NewYork 1961 pp 73ndash8

13 Robert Nozick in Philosophical Explanations Oxford University PressOxford 1984 pp 116ndash21 discusses the possibility of explanatory self-implication for the case of laws Suppose that the lsquoultimate lawrsquo was (P)lawlike statements with characteristic f are true lsquofrsquo might stand for somefeature like invariance or symmetry so (P) would assert that the presence ofsuch a feature was a sufficient condition for the truth of a lawlike statementIf further (P) itself has f then we can infer that (P) itself is true As Nozickstresses it is not a question of proving that (P) is true Rather assumingthat (P) is true it is a question of explaining why (P) is true by deducing it asan instance of itself Even if this sort of self-explanation of laws is logicallypossible there is I should think little possibility of finding lsquoultimatersquo lawswhich state sufficient conditions for the truth of lawlike statements in termsof features which they themselves possess

14 Robert Nozick op cit pp 116ndash1715 The interested reader might consult John Skorupski John Stuart Mill

Routledge London 1989 chapters 3 and 4 for a detailed and illuminatingaccount of Millrsquos views on these matters

16 John Skorupski ibid chapter 417 Robert Nozick op cit pp 204ndash11 and 227ndash4018 Mill distinguishes between lsquotwo parts of the process of philosophising the

inferring part and the registering parthelliprsquo (Mill op cit p 122) Mill believesthat error will arise if we ascribe to the latter some of the functions of theformer lsquoThe mistake is that of referring a person to his own notes for theorigin of his knowledgersquo For Mill uninformative deductive inference has afunction but its function is not the same as that of real inference the gainingof new knowledge of the conclusions of those inferences The function of

244


uninformative inference (which to repeat is not real inference for Mill) isto register knowledge that one already possesses lsquoAnd so in all cases thegeneral propositions whether called definitions axioms or laws of naturewhich we lay down at the beginning of our reasonings are merely abridgedstatements in a kind of shorthand of the particular facts which as occasionarises we either think we may proceed on as proved or intend to assumehellipGeneral propositions are merely registers of such inferences already madeand short formulae for making more The major premiss of a syllogismconsequently is a formula of this description and the conclusion is not aninference drawn from the formula but an inference drawn according to theformulahelliprsquo (ibid p 126)

19 Hempel op cit p 33520 Although I will work with this assumption it is not finally clear whether it

is correct so to interpret him In his The Philosophy of Natural Science(1966) Hempel is careful not to claim necessity or sufficiency for theconditions he offers in the analysis of scientific explanation

21 See Rolf Eberle David Kaplan and Richard Montague lsquoHempel andOppenheim on Explanationrsquo Philosophy of Science vol 28 1961 pp 418ndash28 David Kaplan lsquoExplanation Revisitedrsquo Philosophy of Science vol 281961 pp 429ndash36 Jaegwon Kim lsquoDiscussion On the Logical Conditionsof Deductive Explanationrsquo Philosophy of Science vol 30 1963 pp 286ndash91 Robert Ackermann lsquoDiscussions Deductive Scientific ExplanationrsquoPhilosophy of Science vol 32 1965 pp 155ndash67

22 Ardon Lyon lsquoThe Relevance of Wisdomrsquos Work for the Philosophy ofSciencersquo in Wisdom Twelve Essays ed Renford Bambrough BlackwellOxford 1974 pp 218ndash48 See especially pp 232ndash48

23 These various kinds of partiality are carefully distinguished by Hempel opcit pp 415ndash25

24 Compare RCarnap lsquoThe Two Concepts of Probabilityrsquo Philosophy andPhenomenological Research vol V 1945 pp 513ndash32 lsquoThe problem ofprobability may be regarded as the task of finding an adequate definition ofthe concept of probability that can provide a basis for a theory of probabilityThis task is not one of defining a new concept but rather of redefining anold one Thus we have here an instance of that kind of problemhellipwhere aconcept already in use is to be made more exact or rather is to be replacedby a more exact new concept Let us call these problemshellipproblems ofexplication in each case of an explication we call the old concept used ina more or less vague way either in every-day language or in an earlier stageof scientific language the explicandum the new more exact concept whichis proposed to take the place of the old one the explicatumrsquo (p 513) Carnapis adopting a version of the language usersrsquo approach on which he availshimself of the possibility of lsquotidying up the discoursersquo

25 See the discussion of this by Roy Bhaskar A Realist Theory of ScienceHarvester Brighton 1978 pp 63ndash79

26 SBromberger lsquoWhy-Questionsrsquo in Mind and Cosmos Essays inContemporary Science and Philosophy ed Robert Colodny University ofPittsburgh Press Pittsburgh 1966 pp 86ndash111 The counterexamplementioned in the text with four others can be found on pp 92ndash3

245

Notes

27 This isnrsquot quite true but the qualifications donrsquot matter here Imagine adeterministic world in which e occurs Event e might be described in termsof and explicable within two different languages L and Lrsquo Its explanationin L might be an I-S explanation and e might have only an I-S explanationin L There may be no way within the conceptual resources of L to convertthe I-S explanation into a D-N explanation To get a complete D-Nexplanation of e one might have to switch to Lrsquo The explanation of e in Lis thus not part of any D-N explanation

28 I agree with the gist of Peter Railtonrsquos remarks that genuine statisticalexplanation lsquoproperly so called is the explanation of things that happen bychancehelliprsquo (lsquoProbability Explanation and Informationrsquo Synthese vol 481981 pp 233ndash56) but unlike Railton I do not believe that this is Hempelrsquosview (as distinct from being implied by certain things he says) Indeed howcould one reconcile the quotation from Hempel in my text with the view Ifstatistical explanation were independent of the assumption of strictlyuniversal laws as Hempel says then it would be consistent with thatassumption as well See Peter Railton lsquoA Deductive-Nomological Modelof Probabilistic Explanationrsquo Philosophy of Science vol 45 1978 pp 206ndash26 In the quotation Hempel means by lsquouniversalrsquo lsquouniversally quantifiedrsquo

29 In Hempel op cit but he has refined the idea further in the light ofsubsequent criticism

30 JAlberto Coffa lsquoHempelrsquos Ambiguityrsquo Synthese vol 28 1974 pp 141ndash63

31 Hempelrsquos treatment of the epistemic ambiguity of I-S explanation must befurther evidence against Railtonrsquos attribution to Hempel of the explicit avowalof the contrary view in n 28

Chapter V The Ontology of Explanation

1 PFStrawson lsquoCausation and Explanationrsquo in BVermazen and J Hintikkaeds Essays on Davidson Oxford University Press Oxford 1985

2 As I indicated in chapter I the distinction I draw between metaphysics andepistemology is only intended to be rough and ready certainly it may bethat some things or relations belong to both provinces Facts are on myview just that sort of thing

3 Susan Haack Philosophy of Logic Cambridge University Press Cambridge1978 p 246

4 The argument is credited originally to Frege It has also been used by GoumldelQuine (lsquoThree Grades of Modal Involvementrsquo) and Church It is discussedby Robert Cummins and Dale Gottlieb lsquoOn an Argument for Truth-Functionalityrsquo American Philosophical Quarterly vol IX 1972 pp 265ndash9 John Mackie The Cement of the Universe Oxford University PressOxford 1974 Kenneth Russell Olson An Essay on Facts Center for theStudy of Language and Information Leland Stanford Junior CollegeStanford California 1987 Martin Davies Meaning Necessity andQuantification Routledge amp Kegan Paul London 1981 pp 209ndash13GEMAnscombe lsquoCausality and Extensionalityrsquo Journal of Philosophy

246


vol LXVI 1969 pp 152ndash9 My statement of the slingshot is taken fromMackie op cit

5 So Martin Davies tells me See Davies op cit6 What about phenomena I have always found it somewhat surprising that

the term lsquophenomenonrsquo occurs so frequently in the philosophy of explanationliterature Its only other frequent occurrence is in the Kantian literature Ido not know what a phenomenon is at least in the explanation literature ifit is not simply an event

7 Zeno Vendler discusses the mixed case of facts and events as the relata forthe causal relation See Zeno Vendler lsquoCausal Relationsrsquo Journal ofPhilosophy vol LXIV 1967 pp 704ndash13

8 David Lewis lsquoCausal Explanationrsquo Philosophical Papers vol II OxfordUniversity Press Oxford and New York 1986 pp 214ndash40

9 James Woodward lsquoA Theory of Singular Causal Explanationrsquo Erkenntnisvol 21 1984 pp 231ndash62 lsquoAre Singular Causal Explanations ImplicitCovering Law Explanationsrsquo Canadian Journal of Philosophy vol 16 1986pp 253ndash80 Page references in text to the last article

10 John Mackie op cit p26011 Hilary Putnam discusses a case in which there is both a geometric

lsquomacroexplanationrsquo and a lsquomicroexplanationrsquo in terms of the laws of particlephysics for the fact that a peg 1 inch square goes through a 1 inch squarehole and not through a 1 inch round hole in his Meaning and the MoralSciences Routledge amp Kegan Paul London 1978 pp 42ndash3 I referred tothis example in chapter I

12 Carl Hempel Aspects of Scientific Explanation Free Press New York 1965quote from p 423

13 How does one know what a sentence (or fact) is about See Nelson GoodmanlsquoAboutrsquo Mind vol LXX 1961 pp 1ndash24

14 Peter Achinstein The Nature of Explanation chapters 2 and 3 passim15 Donald Davidson lsquoCausal Relationsrsquo Journal of Philosophy vol LXIV

no 21 1967 pp 691ndash703 reprinted in Causation and Conditionals edErnest Sosa Oxford University Press Oxford 1975 pp 82ndash94 My pagereferences are to the Sosa collection

16 Donald Davidson op cit pp 84ndash617 Donald lsquoTrue True to the Factsrsquo Journal of Philosophy vol LXVI 1969

pp 748ndash6418 Compare Russellrsquos view of facts in his lsquoThe Philosophy of Logical Atomismrsquo

in Russellrsquos Logical Atomism ed David Pears FontanaCollins London1972 pp 51ndash72 and passim lsquoThe simplest imaginable facts are those whichconsist in the possession of a quality by some particular thingrsquo (p 53) ZenoVendler also insists on the factproposition distinction on metaphysicalgrounds lsquoPropositions belong to the people who make or entertain thembut facts are not ownedhellip The facts of the case however do not belong toanybody they are objectively ldquothererdquo to be found discovered or arrivedatrsquo (Zeno Vendler op cit p 710) The unordinary facts needed forexplanation will not be quite as objective as Vendler says but thisqualification will not erase all the metaphysical differences Vendler mentionsbetween facts and propositions

247

Notes

19 NLWilson lsquoFacts Events and Their Identity Conditionsrsquo PhilosophicalStudies vol 25 1974 pp 303ndash21 Page references in my text to his viewsare to this article On his view which identifies true propositions and factshe must say that entities things can be constituents of true propositionslsquothe notion of an entity being a constituent of a proposition may be bafflingIt is however definablersquo (p 308) Wilsonrsquos lsquodefinitionrsquo does not lessen mybafflement at the idea

20 JLAustin lsquoTruthrsquo Proceedings of the Aristotelian Society supp vol XXIV1950 reprinted in Truth ed George Pitcher Prentice-Hall Englewood CliffsNJ 1964 pp 18ndash31 Quote from p 24

21 Suppose we had agreed earlier to a fine-grained criterion of event identityto obtain a conception of an event which could cope with the ways in whichproperties matter to explanation We would now need lsquoepistemicizedrsquo eventswhich would I think take us to a conception of event unsuited to play therole for which events are introduced

22 Nathan Salmon Fregersquos Puzzle MIT Press Cambridge Mass 1986 p111 Page references in the text are to this

23 Barry Taylor lsquoStates of Affairsrsquo in Truth and Meaning Essays in Semanticsed Gareth Evans and John McDowell Oxford University Press Oxford1976 pp 263ndash84 Taylor uses intensions as the predicative element herequires in constructing facts whereas I use properties (for ordinary facts)and properties as conceptualized (for special or epistemicized facts) Clearlythere are similarities in our approaches Taylorrsquos facts are useless for a theoryof truth (see p 280) Taylor mentions other possible uses for states of affairs(and facts) but he does not mention their employment in a theory ofexplanation

24 Zeno Vendler op cit quotation in text from p 710 NLWilson op citquotation in text from p 305

25 lsquoBCErsquo and lsquoCErsquo (standing for lsquobefore the common erarsquo and lsquothe commonerarsquo) provide a good way for non-Christians to give dates non-ideologically

26 Stephen Schiffer Remnants of Meaning MIT Press Cambridge Mass 1987See especially p 51 chapter 6 (pp 139ndash78) and pp 234ndash9

Chapter VI Arguments Laws and Explanation

1 See for example William Kneale lsquoNatural Laws and Contrary-to-FactConditionalsrsquo Analysis vol 10 1950 pp 121ndash5 Karl Popper The Logicof Scientific Discovery Hutchinson London 1972 Appendix 10 pp 420ndash41 Milton Fisk lsquoAre There Necessary Connections in Naturersquo Philosophyof Science vol 37 1970 pp 385ndash404

2 Richard Braithwaite Scientific Explanation Cambridge University PressCambridge 1964 chapter IX pp 293ndash318 Ernest Nagel The Structure ofScience Harcourt Brace amp World New York 1961 chapter 4 pp 47ndash78DHMellor lsquoNecessities and Universals in Natural Lawsrsquo in DHMellored Science Belief and Behaviour Cambridge University Press Cambridge1980 pp 105ndash19

3 Fred Dretske lsquoLaws of Naturersquo Philosophy of Science vol 44 1977 pp

248


248ndash68 for a reply to Dretske see Ilkka Niiniluoto Philosophy of Sciencevol 45 1978 pp 431ndash9 David Armstrong What is a Law of NatureCambridge University Press Cambridge 1987

4 Ardon Lyon lsquoThe Relevance of Wisdomrsquos Work for the Philosophy ofScience A Study of the Concept of Scientific Explanationrsquo in WisdomTwelve Essays ed Renford Bambrough Blackwell Oxford 1974 pp218ndash48

5 Baruch Brody lsquoTowards an Aristotelian Theory of Explanationrsquo Philosophyof Science vol 39 1972 pp 20ndash31

6 Peter Achinstein The Nature of Explanation Oxford University Press NewYork 1983 Discussion of this example on pp 168 and 170ndash1

7 The case of causal pre-emption presents some difficulty for any analysis ofdeterministic andor nondeterministic causation which makes a causenecessary in the circumstances for its effect As discussed in chapters I andII David Lewis does not think that a nondeterministic cause is necessary inthe circumstances for its effect but he does think that a deterministic causeis Lewis deals (on p191) with the case of pre-emption in lsquoCausationrsquoreprinted in Ernest Sosa Causation and Conditionals Oxford UniversityPress Oxford 1975 pp 180ndash91 his treatment is discussed by WilliamGoosens lsquoCausal Chains and Counterfactualsrsquo Journal of Philosophy 1979pp 489ndash95

8 Michael Redhead suggests this reply in a paper lsquoExplanationrsquo (unpublished)lsquohellipwe we need to attend to all the relevant circumstanceshellip Again thescientific ideal assumes that all the relevant circumstances are being citedrsquo(p 5) Redheadrsquos reply to my argument is that I neglect the relevantmicrophysical circumstances linking the bus but not the arsenic with thedeath On this sort of view at best only microphysical explanation will meetHempelrsquos requirements for explanation

9 In private discussion10 Wesley Salmon lsquoA Third Dogma of Empiricismrsquo in Basic Problems in

Methodology and Linguistics ed Robert Butts and Jaako Hintikka ReidelDordrecht 1977 pp 149ndash66 Readers can learn about the nervous husbandand the religious explainer on p 150 Also section 2 of Wesley SalmonlsquoStatistical Explanationrsquo in RColodny ed The Nature and Function ofScientific Theories University of Pittsburgh Press Pittsburgh 1970 pp173ndash231 reprinted in WSalmon RJeffrey and JGreeno StatisticalExplanation and Statistical Relevance University of Pittsburgh PressPittsburgh 1971 pp 29ndash88

11 John Meixner lsquoHomogeneity and Explanatory Depthrsquo Philosophy of Sciencevol 46 1979 pp 366ndash81

12 There are two principles of explanation which might be thought to be trueThe first is a closure principle and the second has a certain similarity to aclosure principle (P1) if p explains q and q entails r then p explains r (P2)if p entails q and if q explains r then p explains r As it stands (P1) canrsquot beright since it implies that everything explains a tautology I do not knowwhether a suitably modified version of (P1) is true but I pointed out abovein the text that Hempelrsquos account of explanation cannot accept (P1) even ifsuitably modified to rule out this absurd implication about explanation of

249

Notes

tautologies According to Hempelrsquos D-N model explanation is not closedunder logical entailment

(P2) says that a statement explains everything that anything it entailsexplains (P2) is of course not available to the statistical relevance theoristto use in his own defence against Meixner On the statistical relevance theoryno statistically irrelevant information can be included in an explanans Butsince the premiss p will typically be information-richer than q p may containsome additional information statistically irrelevant to the truth of r So onthe statistical relevance theory p cannot explain r just because q does and pentails q

(P2) is unsound in any case since it falls foul of Salmonrsquos originalirrelevance objection Sometimes as Meixner says we are happy withstatistically irrelevant information (like the fact that the substance is salt)but of course sometimes we are not as in the original counterexamples Theoriginal counterexamples provide cases in which p entails q q explains rand yet p fails to explain r

13 James Woodward lsquoExplanatory Asymmetriesrsquo Philosophy of Science vol51 1984 pp 421ndash42 See also Evan Jobe lsquoA Puzzle Concerning D-NExplanationrsquo Philosophy of Science vol 43 1976 pp 542ndash9 and ClarkGlymour lsquoTwo Flagpoles are More Paradoxical Than Onersquo Philosophy ofScience vol 45 1978 pp 118ndash19 Peter Achinstein op cit p 236 lsquoIt ispossible to explain the presence of a macro-property by appeal to the presenceof an identical micro-property or vice-versarsquo Achinstein does not draw theconclusion explicitly that explanation is not asymmetric but the conclusionfollows from what he does say

14 Aristotlersquos example of vines which are deciduous because broad-leavedprovides a lsquosymmetryrsquo counterexample to Hempelrsquos account of theexplanation of laws

15 Wesley Salmon lsquoA Third Dogma of Empiricismrsquo p 15016 Wesley Salmon Scientific Explanation and the Causal Structure of the World

Princeton University Press Princeton 1984 p 192 p 9617 Baruch Brody op cit pp 23ndash418 Timothy McCarthy lsquoOn an Aristotelian Model of Scientific Explanationrsquo

Philosophy of Science vol 44 1977 pp 159ndash6619 Jaegwon Kim lsquoDiscussion On the Logical Conditions of Deductive

Explanationrsquo Philosophy of Science vol 30 1963 pp 286ndash91 Theconjunctive normal form requirement is introduced on p 288

20 McCarthy op cit pp 161ndash2 Can we strengthen the causal requirementsto rule out a McCarthy-style counterexample In the arguments so far thelaw even though it might be a causal law is lsquoirrelevantrsquo to the explanation(although it is not irrelevant for the derivation) The law may be a causallaw but it does not join the cause of the explanandum event with theexplanandum event The law premiss and the description of theexplanandumrsquos cause donrsquot mesh together In McCarthyrsquos example the law(letrsquos assume that it is an irrelevant causal law) was but the description ofthe explanandumrsquos cause was lsquoCersquo As McCarthy says lsquoLet represent anylaw irrelevant to the occurence of ersquo (p 161) In the second example thelaw relates blackness and crows but the cause of orsquos turning black has nothing

250


to do with the law the cause of orsquos turning black is having been immersedin black paint No law in the derivation related black paint immersion andturning black Perhaps a bit of tinkering is all that is needed Can we imposea further requirement and thereby escape the counterexample to Hempelrsquostheory as supplemented by the causal requirementSuppose we impose the additional requirement that not only must there bea premiss essential to the argument which describes C the particular causeof the event to be explained but that there also must be a law premissessential to the argument such that c(o) and the event to be explained inthis case orsquos turning black are covered by that law That is to say the lawitself must not be lsquoirrelevantrsquo it must bring together the event to beexplained and the cause of that event Thus the additional lsquorelevancersquoneeded can be cashed out as lsquothe law must be a covering law which coversthe token cause and effect mentioned in the explanatory argumentrsquo If thereis one law which covers both the token cause and the token effect the lawwill be a causal lawWe must not require that the explanandum event and the cause be coveredby the same law as the above suggests This would be too strong forsurely there are occasions on which we can explain an effect by its causemediately rather than immediately via two or more laws Perhaps weshould require that however many laws there are not only must thepremisses contain a description of the cause of the event to be explainedbut that both the cause and the explanandum event must be covered byrelevant laws which may relate the cause with the effect only mediatelyso that the cause and effect can each be covered by a different law Nodoubt at least one of the laws will be a causal law but it would be toostrong to require that all of the lsquointerconnectingrsquo laws be causal I canexplain the period of a metal pendulum at trsquo by citing the fact that heatwas applied to the pendulum at t the causal law that heat causes metal toexpand and the (non-causal) law that relates the length and period of apendulumEven this additional condition will not let us deal with McCarthyrsquos thirdcase which is as follows I shall first sketch the third example informallyin order to make it fully intuitive Suppose orsquos being F causes o to be GOne would imagine that the explanation of orsquos being G is orsquos being F viathe causal law (for let us suppose that it is a causal law) that whatever is Fis G But with certain other assumptions about the case we can constructan argument which meets all of the Mill-Hempel conditions evensupplemented in all the required causal ways I have suggested but whichstill fails to explainAs we have already specified orsquos being F causes o to be G What we are toimagine is a case in which the cause of an event to be explained is also thecause of the predicition of that event If a machine of type T is brought intocontact with an object which is F the objectrsquos being F causes the machine topredict that the object is G as well as causing the object to be G Moreoverthe machines are to be of type T which are lsquoinfallible predictorsrsquo if it predictsthat an object is G it follows that the object is G We can now obtain thefollowing argument noting that (2) states a causal law

251

Notes

(1) If a machine is of type T and if it predicts that an object is G itfollows that the object is G

(2) If an object is F and if a machine of type T is in the rightrelationship with the object the machine will predict that the objectis G

(3) Object o is F(4) The machine of type T is in the right relationship with object o(5) Object o is G

This argument meets all the conditions we have laid down The premissesinclude essentially a description of the cause of orsquos being G namely orsquosbeing F Further the premisses include laws which cover and connect thecause and effect and at least one of which is a causal law But still I believethe argument is not an explanation of why o is G The object o is G becauseit is F and nothing in the derivation reflects this

21 My suggestion for remedying the difficulty McCarthy points out is takenfrom or anyway inspired by Peter Achinstein op cit pp 159ndash62 188ndash92

22 This idea is close to Peter Achinsteinrsquos conception of a complete content-giving proposition I do not believe though that any purely grammaticalcharacterization of this idea is possible See Peter Achinstein ibid pp 28ndash48 and my review of his book in the British Journal for the Philosophy ofScience vol 37 1986 pp 377ndash84

23 Peter Achinstein op cit pp 78ndash8324 Wesley Salmon lsquoA Third Dogma of Empiricismrsquo pp 159ndash6225 That there must be a lawlike generalization among the premisses in an

explanatory argument does not follow simply from the assumption thatexplanations are arguments for there are sound arguments with no suchpremiss But the additional assumptions that would be needed in the case ofarguments that are explanations are straightforward and uncontroversial tothe question at hand

26 Gilber t Ryle lsquoldquoI f rdquo ldquoSordquo and ldquoBecauserdquorsquo in Max Black ed Philosophical Analysis A Collection of Essays Prentice-Hall EnglewoodCliffs NJ 1963 pp 302ndash18 Michael Scriven in a series of contributionsbut perhaps especially in lsquoTruisms as the Grounds for HistoricalExplanationsrsquo in Theories of History ed Patrick Gardiner Free PressNew York 1959 pp 443ndash75 (see p 446 page references in the text areto lsquoTruismshelliprsquo) Wesley Salmon lsquoA Third Dogma of Empiricismrsquo pp158ndash9 Peter Achinstein op cit pp 81ndash3 and also in lsquoThe Object ofExplanationrsquo in Explanation ed Stephan Korner Blackwell Oxford1975 pp 1ndash45

27 See Thomas Nickles lsquoDavidson on Explanationrsquo Philosophical Studies vol31 1977 pp 141ndash5 where the idea that lsquostrictrsquo covering laws may be lsquonon-explanatoryrsquo is developed

28 Scrivenrsquos distinction is similar to Donald Davidsonrsquos between homonomicand heteronomic generalizations See Davidson lsquoMental Eventsrsquo reprintedin his Essays on Actions and Events Oxford University Press Oxford 1980pp 207ndash27 see especially pp 218ndash20

252


29 Note that I say lsquohellipthat the following is a full explanationrsquo It is no part ofmy view that there can be at most only one full explanation for a singularfact To take just one possibility suppose one wants to explain why o is GSuppose it is a law that all D are F and a law that all F are G The fact thato is G can be fully explained both by the fact that o is F and the fact that ois D

30 I do not deny that there can be cases of explanation in which explanatoryrelevance is borne by names indeed I said as much in chapter V But I donot deal with these cases here

31 David Hume A Treatise of Human Nature ed LASelby-Bigge OxfordUniversity Press Oxford 1965 p 88 I think that much of the motivationfor the inclusion of a generalization in every full explanation stems fromthe Humeian analysis of causation

32 See Michael Friedman lsquoTheoretical Explanationrsquo in Reduction Time andReality ed Richard Healey Cambridge University Press Cambridge 1981pp 1ndash16 See also his lsquoExplanation and Scientific Understandingrsquo Journalof Philosophy vol LXXI 1974 pp 5ndash19 and the reply by Philip KitcherlsquoExplanation Conjunction and Unificationrsquo Journal of Philosophy volLXXIII 1976 pp 207ndash12

Chapter VII A Realist Theory of Explanation

1 Jaegwon Kim lsquoNoncausal Connectionsrsquo Nous vol 8 1974 pp 41ndash52 Kimrsquosown examples of non-causal determinative relations include compositionaldetermination of one event by another which is the event-analogue of what Ihave called lsquomereological determinationrsquo and two others which I do notdiscuss Cambridge determination and agency determination I have discussedCambridge determination in my lsquoA Puzzle About Posthumous PredicationrsquoPhilosophical Review vol XCVII 1988 pp 211ndash36

2 Bas van Fraassen The Scientific Image Oxford University Press Oxford1980 p 124

3 John Forge lsquoPhysical Explanation With Reference to the Theories ofScientific Explanation of Hempel and Salmonrsquo in Robert McLaughlin edWhat Where When Why Reidel Dordrecht 1982 pp 211ndash29 Quotationfrom p 228

4 Wesley Salmon Scientific Explanation and the Causal Structure of the WorldPrinceton University Press Princeton 1984 p 132 See also pp 242ndash59

5 Richard Miller Fact and Method Princeton University Press Princeton1987 p60

6 Philip Kitcher lsquoSalmon on Explanation and Causality Two Approaches toExplanationrsquo Journal of Philosophy vol LXXXII 1985 pp 632ndash9Examples are given on pp 636ndash7

7 Nancy Cartwright How the Laws of Physics Lie Oxford University PressOxford 1983 p 21

8 Clark Glymour lsquoCausal Inference and Causal Explanationrsquo in RobertMcLaughlin op cit pp 179ndash91 Quotation from p 184 His examples arefrom pp 184ndash6

253

Notes


10 John Forge lsquoThe Instance Theory of Explanationrsquo Australasian Journal ofPhilosophy vol 64 1986 p 132

11 Either a relational fact that c and e stand in some relation or a conditionalfact the fact that if c then e I do not bother to distinguish relationsfrom sentence connectives here since it makes no difference to myargument

12 Peter Achinstein The Nature of Explanation Oxford University Press NewYork 1983 pp 228ndash48 also his lsquoA Type of Non-Causal Explanationrsquo inMidwest Studies in Philosophy vol IX 1984 University of Minnesota PressMinneapolis pp 221ndash43

13 Or as I would prefer to put it some explanations of a singular fact about anevent invoke as explanans a singular fact about another eventcontemporaneous with the first

14 For example see the discussion in Tom Beauchamp and AlexanderRosenberg Hume and the Problem of Causation Oxford University PressNew York 1981 pp 236ndash40 They offer a lsquomicrorsquo reply to such cases

15 JLMackie The Cement of the Universe Oxford University PressOxford 1974 pp 154ndash9 For a view contrary to Mackiersquos see RobertCummins lsquoStates Causes and the Law of Inertiarsquo PhilosophicalStudies vol 29 1976 pp 21ndash36 The crux of Cumminsrsquos argumentseems to be lsquoa state is a condition of changelessnessrsquo and all effectsare changes A system which remains in a state of inertia during aninterval is one in which there is no change and hence one in whichthere is no effect during that interval But if there is no effect in such asystem during that interval there can be nothing which is a cause of aneffect in that system during that interval (ibid pp 22ndash4) The dubiouspremiss in the argument is that all effects are changes presumably itis this which Mackie would deny

16 James Woodward lsquoExplanatory Asymmetriesrsquo Philosophy of Science vol51 1984 pp 421ndash42 Quotation from p 436

17 Einstein Podolsky and Rosen lsquoCan Quantum-Mechanical Description ofPhysical Reality Be Considered Completersquo Physical Review vol 47 1935pp 777ndash80 JSBell lsquoOn the Einstein Podolsky Rosen Paradoxrsquo Physicsvol I 1964 pp 195ndash200 and lsquoOn the Problem of Hidden Variables inQuantum Mechanicsrsquo Review of Modern Physics vol 38 1966 pp 447ndash52 I rely on Salmon op cit and Patrick Suppes Probabilistic MetaphysicsBlackwell Oxford 1984 for my (scanty) knowledge of this problem

18 See for example OCosta de Beauregard lsquoTwo Lectures on the Directionof Timersquo Synthese vol 35 1977 pp 129ndash54

19 Bernard drsquoEspagnat lsquoThe Quantum Theory and Realityrsquo Scientific Americanvol 241 no 5 1979 pp 158ndash81

20 My indebtedness in this section and the following to Peter Achinsteinrsquos workwill be obvious to anyone who knows his writings

21 I add lsquothingrsquo because I have in mind the thesis that God is causa sui I haveno reason to dispute the thesis it falls outside the purview of my claims here

254


22 I donrsquot use the contingency claim (2) here because I want to leave it openwhether the identities are contingent or necessary

23 Peter Achinstein The Nature of Explanation pp 233ndash724 Paul Oppenheim and Hilary Putnam lsquoUnity of Science as a Working

Hypothesisrsquo Minnesota Studies in the Philosophy of Science ed HFeiglMScriven and GMaxwell University of Minnesota Press Minneapolis1958 pp 3ndash36

25 David-Hillel Ruben The Metaphysics of the Social World Routledge ampKegan Paul London 1985

26 See for example JWNWatkins Hobbesrsquo System of Ideas HutchinsonLondon 1973 chapter 3 lsquoScientific Traditionrsquo pp 28ndash42

27 UTPlace lsquoIs Consciousness a Brain Processrsquo British Journal ofPsychology vol XLVII 1956 pp 44ndash50 and reprinted in The Philosophyof Mind ed VCChappell Prentice-Hall Englewood Cliffs NJ 1962 pp101ndash9

28 John Locke An Essay Concerning the Human Understanding ed ASPringle-Pattison Oxford University Press London 1964 p 243

29 David Lewis would deny this see his lsquoEventsrsquo in Philosophical Papersvol II Oxford University Press Oxford and New York 1986 pp 241ndash69On pp 262ndash6 Lewis deals explicitly with Kimrsquos Socrates-Xantippe example

30 Jaegwon Kim lsquoSupervenience and Supervenient Causationrsquo in SpindelConference 1983 Supervenience ed Terence Horgan vol XXII Supplementto Southern Journal of Philosophy pp 45ndash61

31 See Cynthia and Graham Macdonald lsquoMental Causes and the Explanationof Actionrsquo Philosophical Quarterly vol 36 1986 pp 145ndash58 and especiallyp 157 where they argue that since the supervenience of the mental on thephysical is likely to be stipulated on a priori grounds there will not or maynot be any explanations of the mental by the physical

32 RMHare Philosophical Review vol 68 1959 pp 421ndash56 lsquoFirst let ustake the characteristic of ldquogoodrdquo which has been called its supervenienceSuppose that we say St Francis was a good man It is logically impossibleto say this and to maintain at the same time that there might have beenanother man placed exactly in the same circumstances as St Francis andwho behaved in exactly the same way but who differed from St Francis inthis respect only that he was not a good manrsquo

33 John Bacon lsquoSupervenience Necessary Coextension and ReducibilityrsquoPhilosophical Studies vol 49 1986 pp 163ndash76 Quotation from p 175

34 DHMellor lsquoIn Defense of Dispositionsrsquo Philosophical Review volLXXXIII 1974 pp 157ndash81 Quotation from p 172

35 Elizabeth Prior Dispositions Aberdeen University Press Aberdeen1985 p62 lsquohellipthe commonly accepted view that dispositional propertiescan be distinguished from categorical ones because dispositionalascription sentences possess a relationship to certain subjunctiveconditionals not possessed by categorical ascription sentences survivesunscathedrsquo

36 Pace Mellor lsquoExplanatory dispositions require some independent basis fortheir ascriptions between displays but the basis need only be anotherdispositionrsquo (op cit p 174)

255

Notes

37 David Lewis lsquoCausal Explanationrsquo in op cit pp 214ndash40 Page referencesin my text are to this article Lewisrsquos own example of small-pox immunitymisleads him because the lsquoFrsquo in his example is lsquohellipprotectshelliprsquo which canhave either a dispositional or a non-dispositional sense

38 Assuming of course that the austere theorist is wrong and that this is adistinctive metaphysical relation

39 The idea of determination can perhaps even be extended to the relationbetween a general law(s) and the less general regularities or particularoccurrences that the former explains There is a sense of determinationdescribed by Professor Anscombe in which the rules of chess mightdetermine the next move in a game The chess rules create specific movepossibilities and the current position of the pieces in conjunction with therules may reduce the possibilities to one Similarly the existence of aregularity in a system S may be determined by a set of laws governing thatsystem (GEMAnscombe lsquoCausality and Determinationrsquo reprinted inCausation and Conditionals ed Ernest Sosa Oxford University PressOxford 1975 pp 63ndash81) More general regularities determine less generalones A determinative theory of explanation can also hope to captureexplanation of laws by more general laws

256

Bibliography

Achinstein Peter 1975 lsquoThe Object of Explanationrsquo in Explanation ed StephanKoumlrner Blackwell Oxford

mdash1983 The Nature of Explanation Oxford University Press New Yorkmdash1984 lsquoA Type of Non-Causal Explanationrsquo in Midwest Studies in Philosophy

IX University of Minnesota Press MinneapolisAckermann Robert 1965 lsquoDiscussions Deductive Scientific Explanationrsquo

Philosophy of Science 32Annas Julia 1982 lsquoAristotle on Inefficient Causesrsquo Philosophical Quarterly

32Anscombe GEM 1969 lsquoCausality and Extensionalityrsquo Journal of Philosophy

LXVImdash1975 lsquoCausality and Determinationrsquo reprinted in Causation and Conditionals

ed Ernest Sosa Oxford University Press OxfordAristotle 1966 The Basic Works of Aristotle ed Richard McKeon Random

House New Yorkmdash1975 Posterior Analytics trans Jonathan Barnes Oxford University Press

OxfordArmstrong David 1987 What Is a Law of Nature Cambridge University Press

CambridgeAustin JL 1984 How To Do Things With Words second edition ed JO Urmson

and Marina Sbisagrave Oxford University Press Oxfordmdash1964 lsquoTruthrsquo reprinted in Truth ed George Pitcher Prentice-Hall Englewood

Cliffs NJBacon John 1986 lsquoSupervenience Necessary Coextension and Reducibilityrsquo

Philosophical Studies 49Beauchamp Tom and Alexander Rosenberg 1981 Hume and the Problem of

Causation Oxford University Press New YorkBell JS 1964 lsquoOn the Einstein Podolsky Rosen Paradoxrsquo Physics Imdash1966 lsquoOn the Problem of Hidden Variables in Quantum Mechanicsrsquo Review

of Modern Physics 38Bhaskar Roy 1978 A Realist Theory of Science Harvester Brighton

257

Bibliography

Braithwaite Richard 1964 Scientific Explanation Cambridge University PressCambridge

Brody Baruch 1972 lsquoTowards an Aristotelian Theory of Scientific ExplanationrsquoPhilosophy of Science 39

Bromberger Sylvain 1965 lsquoAn Approach to Explanationrsquo in AnalyticalPhilosophy second series ed RJButler Blackwell Oxford

mdash1966 lsquoWhy-Questionsrsquo in Mind and Cosmos Essays in Contemporary Scienceand Philosophy ed Robert Colodny University of Pittsburgh PressPittsburgh

Burge EL 1971 lsquoThe Ideas as Aitiai in the Phaedorsquo Phronesis 16Carnap Rudolf 1945 lsquoThe Two Concepts of Probabilityrsquo Philosophy and

Phenomenological Research VCartwright Nancy 1983 How the Laws of Physics Lie Oxford University Press

OxfordClark Romane and Paul Welsh 1962 Introduction to Logic Van Nostrand

PrincetonCoffa JAlberto 1974 lsquoHempelrsquos Ambiguityrsquo Synthese 28Collins Arthur 1966 lsquoExplanation and Causalityrsquo Mind LXXVCresswell MJ 1971 lsquoPlatorsquos Theory of Causality Phaedo 95ndash106rsquo

Australasian Journal of Philosophy 49Cummins Robert 1976 lsquoStates Causes and the Law of Inertiarsquo Philosophical

Studies 29mdashand Dale Gottlieb 1972 lsquoOn an Argument for Truth-Functionalityrsquo American

Philosophical Quarterly IXDavidson Donald 1969 lsquoTrue to the Factsrsquo Journal of Philosophy LXVImdash1975 lsquoCausal Relationsrsquo reprinted in Causation and Conditionals ed Ernest

Sosa Oxford University Press Oxfordmdash1980 lsquoMental Eventsrsquo reprinted in his Essays on Actions and Events Oxford

University Press OxfordDavies Martin 1981 Meaning Necessity and Quantification Routledge amp Kegan

Paul Londonde Beauregard OCosta 1977 lsquoTwo Lectures on the Direction of Timersquo Synthese

35drsquoEspagnat Bernard 1979 lsquoThe Quantum Theory and Realityrsquo Scientific

American 241 5Dorling John 1978 lsquoOn Explanation in Physics Sketch of an Alternative to

Hempelrsquos Account of the Explanation of Lawsrsquo Philosophy of Science 45Dretske Fred 1972 lsquoContrastive Factsrsquo Philosophical Review 81mdash1977 lsquoLaws of Naturersquo Philosophy of Science 44Duhem Pierre 1977 The Aim and Structure of Physical Theory Atheneum New

YorkEberle Rolf David Kaplan and Richard Montague 1961 lsquoHempel and

Oppenheim on Explanationrsquo Philosophy of Science 28Einstein Podolsky and Rosen 1935 lsquoCan Quantum-Mechanical Description of

Physical Reality Be Considered Completersquo Physical Review 47Fisk Milton 1970 lsquoAre There Necessary Connections in Naturersquo Philosophy

of Science 37Forge John 1982 lsquoPhysical Explanation With Reference to the Theories of

258


Scientific Explanation of Hempel and Salmonrsquo in Robert McLaughlin edWhat Where When Why Reidel Dordrecht

mdash1986 lsquoThe Instance Theory of Explanationrsquo Australasian Journal ofPhilosophy 64

Friedman Michael 1974 lsquoExplanation and Scientific Understandingrsquo Journalof Philosophy LXXI

mdash1981 lsquoTheoretical Explanationrsquo in Reduction Time and Reality ed RichardHealey Cambridge University Press Cambridge

Garfinkel Alan 1981 Forms of Explanation Yale University Press New HavenGettier Edmund 1963 lsquoIs Justified True Belief Knowledgersquo Analysis 23Glymour Clark 1978 lsquoTwo Flagpoles Are More Paradoxical Than Onersquo

Philosophy of Science 45mdash1982 lsquoCausal Inference and Causal Explanationrsquo in Robert McLaughlin ed

What Where When Why Reidel DordrechtGoodman Nelson 1961 lsquoAboutrsquo Mind LXXGoosens William 1979 lsquoCausal Chains and Counterfactualsrsquo Journal of

Philosophy vol LXXVIHaack Susan 1978 Philosophy of Logic Cambridge University Press

CambridgeHamlyn David 1976 lsquoAristotelian Epagogersquo Phronesis XXIHare RM 1959 lsquoAesthetic Conceptsrsquo Philosophical Review 68Hempel Carl 1965 Aspects of Scientific Explanation The Free Press New York

(I have in fact used the 1970 paperback with the same pagination)mdash1966 Philosophy of Natural Science Prentice-Hall Englewood Cliffs NJHocutt Max 1974 lsquoAristotlersquos Four Becausesrsquo Philosophy 49Howson Colin 1988 lsquoOn a Recent Argument for the Impossibility of a Statistical

Explanation of Single Events and a Defence of a Modified Form of HempelrsquosTheory of Statistical Explanationrsquo Erkenntnis 29

Jobe Evan 1976 lsquoA Puzzle Concerning D-N Explanationrsquo Philosophy of Science43

Kaplan David 1961 lsquoExplanation Revisitedrsquo Philosophy of Science 28Kim Jaegwon 1963 lsquoDiscussion On the Logical Conditions of Deductive

Explanationrsquo Philosophy of Science 30mdash1974 lsquoNoncausal Connectionsrsquo Nous 8mdash1983 lsquoSupervenience and Supervenient Causationrsquo in Spindel Conference

1983 Supervenience ed Terence Horgan XXII Supplement to SouthernJournal of Philosophy

Kitcher Philip 1976 lsquoExplanation Conjunction and Unificationrsquo Journal ofPhilosophy 73

mdash1985 lsquoSalmon on Explanation and Causality Two Approaches to ExplanationrsquoJournal of Philosophy LXXXII

Kneale William 1950 lsquoNatural Laws and Contrary-to-Fact ConditionalsrsquoAnalysis 10

Kyburg Henry Jr 1970 lsquoConjunctivitisrsquo in MSwain ed Induction Acceptanceand Rational Beliefs Reidel Dordrecht

Lambert Karel and Gordon GBrittan Jr 1987 An Introduction to the Philosophyof Science third edition Ridgeview Publishing Company Atascadero

Lewis David 1986 Philosophical Papers vol II Oxford University Press New

259

Bibliography

YorkLipton Peter 1987 lsquoA Real Contrastrsquo Analysis 47Lyon Ardon 1974 lsquoThe Relevance of Wisdomrsquos Work for the Philosophy of

Sciencersquo in Wisdom Twelve Essays ed Renford Bambrough BlackwellOxford

McCarthy Timothy 1977 lsquoDiscussion on an Aristotelian Model of ScientificExplanationrsquo Philosophy of Science 44

Macdonald Cynthia and Graham 1986 lsquoMental Causes and the Explanation ofActionrsquo Philosophical Quarterly 36

Mackenzie Mary lsquoPlatorsquos Analysis of Individuationrsquo unpublished manuscriptMackie John 1973 Truth Probability and Paradox Oxford University Press

Oxfordmdash1974 The Cement of the Universe Oxford University Press OxfordMeixner John 1979 lsquoHomogeneity and Explanatory Depthrsquo Philosophy of

Science 46Melling David 1987 Understanding Plato Oxford University Press OxfordMellor DH 1974 lsquoIn Defense of Dispositionsrsquo Philosophical Review LXXXIIImdash1976 lsquoProbable Explanationrsquo Australasian Journal of Philosophy 54mdash1980 lsquoNecessities and Universals in Natural Lawsrsquo in Mellor ed Science

Belief and Behaviour Cambridge University Press CambridgeMill John Stuart 1970 A System of Logic Longman LondonMiller Richard 1987 Fact and Method Princeton University Press PrincetonMoravcsik Julius 1974 lsquoAristotle on Adequate Explanationsrsquo Synthese 28Nagel Ernest 1961 The Structure of Science Harcourt Brace amp World New

YorkNickles Thomas 1977 lsquoDavidson on Explanationrsquo Philosophical Studies 31Niiniluoto Ilkka 1978 lsquoDretske on Laws of Naturersquo Philosophy of Science 45Nozick Robert 1984 Philosophical Explanations Oxford University Press

OxfordOlson Kenneth Russell 1987 An Essay on Facts Center for the Study of

Language and Information Leland Stanford Junior College StanfordCalifornia

Oppenheim Paul and Hilary Putnam 1958 lsquoUnity of Science as a WorkingHypothesisrsquo Minnesota Studies in the Philosophy of Science ed FeiglScriven and Maxwell University of Minnesota Press Minneapolis

Owen GEL 1975 lsquoTithenai ta Phainomenarsquo reprinted in Articles on Aristotle1 Science ed Jonathan Barnes Malcolm Schofield and Richard SorabjiDuckworth London

Place UT 1962 lsquoIs Consciousness a Brain Processrsquo reprinted in ThePhilosophy of Mind ed VCChappell Prentice-Hall Englewood Cliffs NJ

Plato 1955 The Phaedo trans RSBluck Bobbs-Merrill Indianapolismdash1966 Plato The Collected Dialogues ed Edith Hamilton and Huntington

Cairns Bollingen FoundationPopper Karl 1972 The Logic of Scientific Discovery Hutchinson Londonmdash1973 lsquoEpistemology Without a Knowing Subjectrsquo in Objective Knowledge

Oxford University Press OxfordPrior Elizabeth 1985 Dispositions Aberdeen University Press AberdeenPutnam Hilary 1978 Meaning and the Moral Sciences Routledge amp Kegan Paul

260


LondonRailton Peter 1978 lsquoA Deductive-Nomological Model of Probabilistic

Explanationrsquo Philosophy of Science 45mdash1981 lsquoProbability Explanation and Informationrsquo Synthese 48Redhead Michael 1989 lsquoExplanationrsquo (unpublished but delivered as a paper at

the Royal Institute of Philosophy Conference on Explanation in Glasgowand to be published in a volume of conference proceedings by CambridgeUniversity Press)

Ruben David-Hillel 1981 lsquoLewis and the Problem of Causal SufficiencyrsquoAnalysis 4

mdash1982 lsquoCausal Scepticism or Invisible Cementrsquo Ratio XXIVmdash1985 The Metaphysics of the Social World Routledge amp Kegan Paul Londonmdash1986 lsquoReview of Peter Achinsteinrsquos The Nature of Explanationrsquo British Journal

for the Philosophy of Science 37mdash1987 lsquoExplaining Contrastive Factsrsquo Analysis 47mdash1988 lsquoA Puzzle about Posthumous Predicationrsquo Philosophical Review XCVIIRussell Bertrand 1972 lsquoThe Philosophy of Logical Atomismrsquo in Russellrsquos

Logical Atomism ed David Pears FontanaCollins LondonRyan Alan 1987 The Philosophy of John Stuart Mill Macmillan London second

editionRyle Gilbert 1963 lsquoldquoIfrdquo ldquoSordquo and ldquoBecauserdquorsquo in Max Black ed Philosophical

Analysis A Collection of Essays Prentice-Hall Englewood Cliffs NJSalmon Nathan 1986 Fregersquos Puzzle MIT Press Cambridge MassSalmon Wesley 1970 lsquoStatistical Explanationrsquo in RColodny ed The Nature

and Function of Scientific Theories University of Pittsburgh PressPittsburgh

mdash1977 lsquoA Third Dogma of Empiricismrsquo in Butts and Hintikka eds BasicProblems in Methodology and Linguistics Reidel Dordrecht

mdash1984 Scientific Explanation and the Causal Structure of the World PrincetonUniversity Press Princeton

mdashRichard Jeffrey and James Greeno 1971 Statistical Explanation and StatisticalRelevance University of Pittsburgh Press Pittsburgh

Schiffer Stephen 1987 Remnants of Meaning MIT Press Cambridge MassScriven Michael 1959 lsquoTruisms as the Grounds for Historical Explanationrsquo in

Theories of History ed Patrick Gardiner The Free Press New YorkSkorupski John 1989 John Stuart Mill Routledge LondonSkyrms Brian 1975 Choice and Chance Dickinson Publishing Company Encino

and Belmont CaliforniaSorabji Richard 1980 Necessity Cause and Blame Duckworth LondonSosa Ernest 1964 lsquoThe Analysis of ldquoKnowledge that Prdquorsquo Analysis 25Stegmuumlller Wofgang 1980 lsquoTwo Successor Concepts to the Notion of Statistical

Explanationrsquo in Logic and Philosophy ed GH von Wright Nijhoff TheHague

Stemmer Nathan 1973 lsquoBrodyrsquos Defense of Essentialismrsquo Philosophy ofScience 40

Strawson Peter 1985 lsquoCausation and Explanationrsquo in Vermazen and Hintikkaeds Essays on Davidson Oxford University Press Oxford

Suppes Patrick 1984 Probabilistic Metaphysics Blackwell Oxford

261

Bibliography

Taylor Barry 1976 lsquoStates of Affairsrsquo in Truth and Meaning Essays inSemantics ed Gareth Evans and John McDowell Oxford University PressOxford

Taylor CCW 1969 lsquoForms as Causes in the Phaedorsquo Mind LXVIIITemple Denis 1988 lsquoThe Contrast Theory of Why-Questionsrsquo Philosophy of

Science 55Toulmin Stephen 1961 Foresight and Understanding Harper New YorkTuomela Raimo 1980 lsquoExplaining Explainingrsquo Erkenntnis 15Urbach Peter 1987 Francis Baconrsquos Philosophy of Science Open Court La

Sallevan Fraassen Bas 1980 lsquoA Re-examination of Aristotlersquos Philosophy of Sciencersquo

Dialogue 19mdash1980 The Scientific Image Oxford University Press OxfordVendler Zeno 1967 lsquoCausal Relationsrsquo Journal of Philosophy LXIV 21Vlastos Gregory 1954 lsquoThe Third Man Argument in the Parmenidesrsquo

Philosophical Review and reprinted in Studies in Platorsquos Metaphysics REAllen ed Routledge amp Kegan Paul London 1967

mdash1969 lsquoReasons and Causes in the Phaedorsquo Philosophical Review 78von Wright Georg Henrik 1971 Explanation and Understanding Routledge amp

Kegan Paul LondonWatkins John 1973 Hobbesrsquo System of Ideas Hutchinson Londonmdash1984 Science and Scepticism Princeton University Press PrincetonWieland W 1975 lsquoThe Problem of Teleologyrsquo reprinted in Articles on Aristotle

I Science ed Jonathan Barnes Malcolm Schofield and Richard SorabjiDuckworth London

Wilson NL 1974 lsquoFacts Events and Their Identity Conditionsrsquo PhilosophicalStudies 25

Woodward James 1984 lsquoA Theory of Singular Causal Explanationrsquo Erkenntnis21

mdash1984 lsquoExplanatory Asymmetriesrsquo Philosophy of Science 51mdash1986 lsquoAre Singular Causal Explanations Implicit Covering Law

Explanationsrsquo Canadian Journal of Philosophy 16

262

Achinstein P 8ndash9 84ndash5 165 184ndash5190ndash1 199 213 218ndash20

Anaxagoras 47Annas J 50 77Aristotle 7 15 77ndash109 111 113 125

127 147 155 169ndash70 181ndash2 192198 202 205ndash6 209 221 232

Austin JL 172ndash3 176

Bacon F 111Bacon J 225Barnes J 79 106Bell JS 216Berkeley G 13 111ndash12Bluck RS 51Brody B 107ndash8 183ndash4Bromberger S 7 147 191Burge E 45ndash6

Cartwright N 212 213Clark R 7Coffa A 153ndash4Cresswell M 51

Davidson D 165ndash7Duhem P 112ndash13Forge J 211ndash12 213 217Friedman M 4ndash5 11ndash12 207ndash8

Garfinkel A 40

Gettier E 11Glymour C 212

Haack S 157Hempel C 3ndash4 7 10 14 15ndash19 24ndash6

28 38 110ndash11 114ndash15 138ndash54 156164 181ndash98 passim 209 213

Hobbes T 13 111ndash12 221Hocutt M 77Howson C 31Hume D 13 111ndash12 204

Kim J 195 209ndash11 223ndash4 232Kitcher P 212Kyburg H 29ndash31

Lewis D 21 40 160 228ndash30Lipton P 43ndash4 187Locke J 112 221ndash2Lyon A 139 183

McCarthy T 194ndash8Mackie J 158ndash9 162 214ndash16Meixner J 189ndash90Mellor H 61ndash2 64 226Mill JS 7 15 37 110ndash38 141 145

181ndash2 192 198 201 204 206 209213

Miller R 211ndash12Moravcsik J 77 85ndash6 232

Name Index

263

Name Index

Nagel E 121Nozick R 129ndash33

Oppenheim P 138 149 220ndash1

Place UT 221Plato 7 37 45ndash76 82ndash3 96 104 113

125 127 209Popper K 12 33Putnam H 21 220ndash1

Railton P 28Ryle G 199

Salmon N 176ndash7Salmon W 24ndash5 27ndash31 38ndash9 87 153

187ndash9 193ndash4 199 211 213Schiffer S 179ndash80

Scriven M 191 199ndash200 202Skorupski J 132ndash3Sorabji R 77 93Strawson P 155ndash6 164

Taylor C 51Temple D 41ndash2Tuomela R 5

van Fraassen B 29 38 40 211 217Vendler Z 177ndash8Vlastos G 45 49ndash51

Watkins JWN 67ndash8Welsh P 7Wieland W 80ndash1 84ndash6Wilson N 168 174 177ndash8Woodward J 160ndash2 168 191

264

ambiguity of lsquoexplanationrsquo 16 28 80ndash1

Cambridge change 50 222ndash4 230ndash3causal explanation 35ndash9 45ndash6 105ndash8

140ndash1 192ndash4 209ndash33causation 50 113ndash14 185 211ndash18

230ndash3closure under conjunction 29ndash31 42ndash3closure under implication 131ndash3 187

248 fn 12contrastives 39ndash44

determinist ic v non-determinist iccausation 35ndash7 46ndash7 49 64ndash70116ndash17 149ndash54

dispositions 225ndash33

emptiness explanatory 67ndash71explaining that 15ndash16 79ndash80explanation complete and partial 16ndash21

143ndash5 149ndash51 202ndash4explanation good and bad 21ndash3 32 163ndash

4 190explanation ordinary and scientific 5ndash6

16ndash19 95ndash108 206explanation theories ofargument and non-argument theories 33ndash

5 45 97 197ndash9certainty high and low epistemic

probability theories 27ndash32 33ndash9deductivism 33ndash9 97ndash108 110ndash11 116ndash17

129ndash38determinative high and low dependency

theories 36ndash9 45 49 61ndash2 64ndash7093ndash5 116ndash17 151 230ndash3

probabilism 33ndash9 100 138 149ndash54

facts v events 23ndash5 39 51ndash2 115 139ndash40 156 160ndash80

fallacy of compositiondecomposition73

identity 157ndash8 165 174ndash6 218ndash22230ndash3

intensionality (non-extensionality) 5778 87ndash93 155ndash80 205ndash6

knowledge 6 10 11ndash12 72ndash5 96ndash7130ndash8

language usersrsquo v technical approach 11ndash15 77 84ndash7 140ndash4

laws explanation of and by 4ndash5 58 8993ndash5 97ndash8 113ndash14 115 117 118ndash23181ndash2 186ndash7 197 199ndash208

mereology 218ndash22 230ndash3

paradox of analysis 10pragmatism explanatory 21ndash3 180probability 26 63ndash72 131

Subject index

265

Subject index

processproduct ambiguity 6ndash9

quantum mechanics 216ndash17

realism explanatory 23ndash4 160 167ndash8209ndash11 230ndash3

reduction 122ndash3 218ndash22 230ndash3reflexivity of explanation relation 129

138 175 219ndash20 221regress of explanation 73ndash4 102ndash4

125ndash9relevanceirrelevance 23 162ndash4 170ndash1

183ndash90

self-explanation 129 138 175 219ndash20

221sentence v non-sentence explanation 23ndash

4 160supervenience 222ndash5symmetry of explanation relation 101

105ndash8 123 129 191ndash3 221ndash2symmetry thesis 123ndash4 145ndash9

theories in explanation 6 206ndash8 220ndash1transitivity of explanation relation 129

understanding 14ndash15unification 207ndash8

why-questions 15ndash16 79ndash80

Book Cover

Title

Contents









Notes

Bibliography

Name Index

Subject Index


David-Hillel Ruben offers a discussion of some of the main historicalattempts to explain the concept of explanation examining the worksof Plato Aristotle John Stuart Mill and Carl Hempel Building onand developing the insights of these historical figures he introducesan elaboration and defense of his own solution

In this volume Ruben relates the concept of explanation to bothepistemological and metaphysical issues Not content to confine theconcept to the realm of philosophy of science he examines it within a farmore broadly conceived theory of knowledge He concludes with his ownoriginal and challenging explanation of explanation

Explaining Explanation will be read with interest by students of generalphilosophy as well as those specializing in the philosophy of science andscholars with a more advanced level of interest



word to object




RJNelson







London and New York














160


also available







vii

Contents





viii






Notes 234

Bibliography 256

Name Index 262

Subject Index 264

ix





x







xi




1

CHAPTER I





2








3




Some explanations




4









5






6




Process and product



7












8







9







10







11








12







13






14








15







16








17







18






19







20






21








22






23




Some terminology





24










25








26






27






28







29








30








31






32







33







34






35







36






37







38








39








40








41









42







43







44



45

CHAPTER II





46







47


The Phaedo






48





(Phaedo 104c)


49






50







51







52








53








54











55








56






Some terminology




57






58






59







60









61








(Mellor 1976237)


62






63










64




Platorsquos (PP2)



65







66





Platorsquos (PP1)




67






68







69







70









71






72




The Theaetetus




73









74





75





Summary


76



77

CHAPTER III





78








79









80







81








82










83









84






85







86






87







88









89








90







91









92











93








94








95








96







(PA I 2 71b8ff)



97







98



(PA I 31)






99





(PA II 12)



(PA I 30)


100





101








102








103






(PA A3 72b5ndash20)


104







105





106











107













108



(Brody 197226)



(PA II 11)


Summary


109



110

CHAPTER IV




111







112






113






(Mill 1970310)


114





(Mill 1970213)






115









116






(Mill 1970305)



117







(Mill 1970338)

118





119






(Mill 1970227)

120






(Mill 1970386)


121









122






123







124






125






126






(Mill 1970318)



127









128












129








130







131



(Mill 1970108)




(Mill 1970120)




132






133









134






135







136







137






138








139







140







141




(Mill 1970311)




142









(Hempel 1965412)


143








144








145








146



(Hempel 1965249)





147



(Hempel 1965375)





148







149











150







151








152









153







154



Summary


155

CHAPTER V






156








157









158







159















160







161






162






of life







163







164








165








166






167








168



Explaining facts



169






170






171










172





(Wilson 1974306)




173







174






175











176







177








178








179






180





181

CHAPTER VI





182








183









184











185







186






187









188







189










190




191







192







193







194






(4) D(e)



195



(McCarthy 1977161)



(5) o turns black



196







197









198






199







200






(Scriven 1959445)


201






202







203







204






205







206






207







208



(Friedman 19817)


209

CHAPTER VII






210



(Kim 197452)


(Kim 197441)




211






212









213






214







215









216







217







218










219





220






221






222







223






224





(Kim 197442ndash3)




225









226






227







228







229






230





I said in chapter I




231







232







233



234

Notes












235

Notes


















236
















237

Notes











238












vol 78 1969 pp 291ndash325

239

Notes
















240














241

Notes














242
















243

Notes
















244











245

Notes











246

















247

Notes













248












249

Notes











250



251

Notes












252















253

Notes














254

















255

Notes




256

Bibliography
















257

Bibliography






















258























259

Bibliography


















260






















261

Bibliography














262








Garfinkel A 40








181ndash2 192 198 201 204 206 209213


Name Index

263

Name Index












264


























Subject index

265

Subject index







183ndash90








Book Cover

Title

Contents









Notes

Bibliography

Name Index

Subject Index



word to object




RJNelson







London and New York














160


also available







vii

Contents





viii






Notes 234

Bibliography 256

Name Index 262

Subject Index 264

ix





x







xi




1

CHAPTER I





2








3




Some explanations




4









5






6




Process and product



7












8







9







10







11








12







13






14








15







16








17







18






19







20






21








22






23




Some terminology





24










25








26






27






28







29








30








31






32







33







34






35







36






37







38








39








40








41









42







43







44



45

CHAPTER II





46







47


The Phaedo






48





(Phaedo 104c)


49






50







51







52








53








54











55








56






Some terminology




57






58






59







60









61








(Mellor 1976237)


62






63










64




Platorsquos (PP2)



65







66





Platorsquos (PP1)




67






68







69







70









71






72




The Theaetetus




73









74





75





Summary


76



77

CHAPTER III





78








79









80







81








82










83









84






85







86






87







88









89








90







91









92











93








94








95








96







(PA I 2 71b8ff)



97







98



(PA I 31)






99





(PA II 12)



(PA I 30)


100





101








102








103






(PA A3 72b5ndash20)


104







105





106











107













108



(Brody 197226)



(PA II 11)


Summary


109



110

CHAPTER IV




111







112






113






(Mill 1970310)


114





(Mill 1970213)






115









116






(Mill 1970305)



117







(Mill 1970338)

118





119






(Mill 1970227)

120






(Mill 1970386)


121









122






123







124






125






126






(Mill 1970318)



127









128












129








130







131



(Mill 1970108)




(Mill 1970120)




132






133









134






135







136







137






138








139







140







141




(Mill 1970311)




142









(Hempel 1965412)


143








144








145








146



(Hempel 1965249)





147



(Hempel 1965375)





148







149











150







151








152









153







154



Summary


155

CHAPTER V






156








157









158







159















160







161






162






of life







163







164








165








166






167








168



Explaining facts



169






170






171










172





(Wilson 1974306)




173







174






175











176







177








178








179






180





181

CHAPTER VI





182








183









184











185







186






187









188







189










190




191







192







193







194






(4) D(e)



195



(McCarthy 1977161)



(5) o turns black



196







197









198






199







200






(Scriven 1959445)


201






202







203







204






205







206






207







208



(Friedman 19817)


209

CHAPTER VII






210



(Kim 197452)


(Kim 197441)




211






212









213






214







215









216







217







218










219





220






221






222







223






224





(Kim 197442ndash3)




225









226






227







228







229






230





I said in chapter I




231







232







233



234

Notes












235

Notes


















236
















237

Notes











238












vol 78 1969 pp 291ndash325

239

Notes
















240














241

Notes














242
















243

Notes
















244











245

Notes











246

















247

Notes













248












249

Notes











250



251

Notes












252















253

Notes














254

















255

Notes




256

Bibliography
















257

Bibliography






















258























259

Bibliography


















260






















261

Bibliography














262








Garfinkel A 40








181ndash2 192 198 201 204 206 209213


Name Index

263

Name Index












264


























Subject index

265

Subject index







183ndash90








Book Cover

Title

Contents









Notes

Bibliography

Name Index

Subject Index

explaining explanation - philarchive

Documents