(International Library of Philosophy) Georg Henrik Von Wright-Explanation and Understanding-Kegan Paul International (1971)

E X P L A N A T IO N A N D U N D E R S T A N D IN G

International Library o f Philosophy and Scientific Method

EDITOR: TED HONDERICH A'Catalogue of books already published in the Internationai L ibrary o f Philosophy and Scientific M ethod

will be found at the end of this volume

EX PLA NA TIO N AND U N D ERSTA N D IN G by

Georg Henrik von W right

LO N D O N RO U TLED G E & K EG A N PAUL

F irst published in G reat B ritain 1 7 7 1 by Routledge & Kegan P a u l L t d Broadway House, 68-74 C arter Lane London E C / H / E L Printed in G reat B ritain by Lowe & Brydone (Printers) L td ., London

© Cornell University 17 7 1 N o p a rt o f this book may be reproduced in any form without perm ission fro m the publisher, except fo r the quotation o f b rie f passages in criticism IS B N 0 7 10 0 7 2 17 8

. . u n d tie fe r als d er T a g g e d a c h t . .

— N i e t z s c h e , Zarathustra

To Norman Malcolm

Preface

T h is book has grown out of m y interest in the theory of action, which in turn grew out of m y interest in norms and values. T o begin w ith, I was attracted by the formal-logical aspects o f action concepts. T h is was a field in which little had previously been done, but the clearing o f which seemed neces­ sary if “ deontic logic” was to stand on a firm footing. From the logic o f action m y interest shifted to the explanation o f action. M y thinking was deeply affected by Charles T aylor’s book The

Explanation of Behaviour. It m ade m e realize how deep into the traditional problem-body, not only o f philosophy of science but o f philosophy generally, questions of explanation cut. So, eventually, w hat had been a study o f action becam e a con­ tribution to the old question of how the sciences of nature are related to the study o f man. T h e questions discussed here are controversial, much de­ bated, and o f great com plexity. N ever before have I felt so strongly the dangers of misunderstanding even on the termi­ nological level. T h e nature o f an author's opinion is far from clear from the wording of a few theses which he puts forward

vii

Preface and defends. Thus, for example, I m ay be said to be an xiou s to defend a view according to which hum an actions cannot have causes. But many writers, both in past and present tim es, m a in ­ tain that actions can be caused. D o I disagree w ith th eir view s? Not necessarily. For those who think that actions h ave causes often use “ cause” in a much broader sense than I do w h en I deny this. Or they may understand “ action” d ifferen tly. It m ay very well be, then, that “ actions” in their sense h ave “ causes” in my sense, or that “ actions” in my sense have “ causes” in theirs. I am not insisting that m y use is either b etter or m ore natural as far as ordinary language goes. At the same time it would be quite wrong to suggest th a t th e difference between the views is merely term inological, i f this means that a complete clarification of the term s m u st lead to complete agreement in substance. In the elucidations o f m ean ­ ing, new concepts, not less controversial than “ cause” an d “ ac­ tion,” will be used. T h e one who affirms and th e o n e w ho disputes that actions can have causes w ill p robab ly assim ilate, link, the two ideas differently with those other concepts. T h e one will emphasize distinctions that the other w ill tend to b lu r or overlook. The “ causalist” will perhaps lin k inten tion s, m o­ tives, and reasons to causes, and actions to events. T h e “ actionist” groups the concepts differently: m otives and reasons w ith actions, and events with causes— and between the tw o groups he sees a sharp divide. The former m ay not w ish to accord experimentation a crucial role in the form ation o f the concept of cause. Or at least he will not adm it that, since exp erim en ta­ tion is a mode of action, action is conceptually fu n d am en tal to causation. The “ causalist” and the “ actionist,” in oth er words, knit differently the conceptual web against w hose backgrou n d they see the world— and they therefore see the w orld d iffer­ ently. When set in a historical perspective their w orld view s become linked with the two traditions o f thought w hich I try to describe and differentiate in Chapter I.

viii

Preface T h e first three chapters of this book were originally inde­ pendent essays. T h e y m ay still be said to be, relative to one another, self-contained. T h e fourth, however, is largely a sketch to show how patterns of explanation discussed abstractly in chapters II and I I I m ay be employed for explicative tasks in historiography and social science. Prelim inary versions of the second and third chapters were presented to various academ ic audiences from 1965 onwards. I am grateful for the challenge and stimulus which critical re­ sponses from m y audiences have given to the further develop­ m ent of m y ideas. T h e first synoptic presentation o f the material contained in this book were m y T arner Lectures at Cam bridge in autum n 1969. I wish to thank the C ouncil o f T rinity C o l­ lege, Cam bridge, for inviting m e to the lectureship. W ith ou t this external impetus, the results of m y research in the area would not have matured into the form o f a book. Extracts o f a more finished version o f the manuscript were presented in public lectures at Cornell in spring 1970. I am deeply indebted to the chairman of the Andrew D . W h ite Professors-at-Large Program and general editor o f the Contem porary Philosophy series, Professor M ax Black, for the opportunities which were essential to the finishing and publication of m y work.

G eorg H enrik von W right

ix

Contents

Preface

vii I.

T w o Traditions

i.

Two main traditions in science and in the philosophy of scientific method: the aristotelian and the galilean. Their association with man’s efforts to understand things teleologi­ cally and to explain them causally.

l

2. Characterization of positivism as a position in the phi­ losophy of science. Emphasis on the unity of scientific method, on mathematical exactitude as an ideal of perfec­ tion, and on the subjection of phenomena to general laws.

3

3. Hermeneutics as a reaction to the methodological monism of positivism. The Geisteswissenschaften. The dis­ tinction between explanation and understanding. The psy­ chological and the semantic aspects of understanding.

4

4. T he ambivalence of the positions of Hegel and Marx. Hegel and Aristotle. T he explicit “ causalism” of marxism contrasted with its implicit teleology.

7

Contents s. The revival of positivism and its immersion in the broader current of analytic philosophy. The rift in the latter. The implicit antipositivism of linguistic philosophy. The traditional positivism of analytic philosophy of science. Methodolog}’ of the behavioral and social sciences in the mid-centurv.

8

6. Hempel’s theory of scientific explanation. T he deductive-nomological and the inductive-probabilistic covering law model. The second not a model of explanation, but a de­ vice for the justification of expectations and predictions.

10

7. Division of the province of teleology into the domains of function and purpose on the one hand and of intentionalitv on the other hand. Cybernetics and “ the causalization of teleology.”

!5

8. Criticisms of the positivist view of scientific laws. C on­ ventionalism. The distinction between nomic necessities and accidental uniformities. The role of modal logic and of the problem of counterfactual conditionals in the revival of the notion of natural necessity.

18

The rise of an analytic philosophy of action. Anscombe on intentionality and practical reasoning. The critique of positivism in the analytic philosophy of history (Dray) and of social science (W inch).

22

The revival of a hermeneutic philosophy of the Geisteswissenschaften. Affinities with analytic philosophy. Rifts in marxist thinking between a “ humanist” orientation towards hermeneutics and a “ scientistic” orientation towards positivism.

29

9.

10.

II.

Causality and Causal Explanation

1. Causation not an obsolete category in the philosophy of science. The subsumption theory of explanation challenges the idea of nomic connections— and therewith the problem of causation.

34

2. Causal relations as conditionship relations. Sufficient and necessary conditions. The extensionalist-quantificational and the intensionalist-modal view of conditionship relations. 3. The asymmetry of cause and effect. This cannot be accounted for in the terms of temporal relationship alone. The possibility of “ retroactive causation.”

41

4. The formal-logical apparatus involved: propositional logic, propositional modal logic, and propositional tenselogic for a discrete time-medium. States of affairs as basic ontological category. The notion of a world and its history. A topological presentation of the possible histories of a world. The notion of a system.

43

5. Causal analysis within systems. Chains of sufficient con­ ditions can have no gaps; chains of necessary conditions may have gaps. The notion of closedness.

5°

6. Types of causal explanation. The questions “ W hy neces­ sary?” and “ How possible?” Answers to questions of the first type can be used for making predictions; answers to ques­ tions of the second type for making retrodictions. Quasi­ teleology or the causal explanation of purposefulness in nature.

55

7. T he closed character of systems established by “ putting them in motion” through an act of interference with the course of nature.

60

8. Action and causation. Distinction between doing and bringing about. Basic actions.

64

9. The experimentalist notion of causation. The distinc­ tion between cause- and effect-factors rests on the distinction between things done and things brought about through action. T he factual conditions which make action logically possible also provide a basis for distinguishing between nomic connections and accidental uniformities in nature.

69

C o n ten ts

io. The problem of asymmetry of the causal relation. T he possibility of retroactive causation reconsidered. It is sug­ gested that, by performing basic actions, an agent may bring about anterior events in his neural system. Determinism a metaphysical illusion nourished by a tendency to think that mere observation of regular sequences suffices to establish norme connections. III.

74

lntentionality and Teleological E xplanation

i . Distinction between causal and quasi-causal explana­ tions. The latter do not depend for their validity on the truth of nomic connections. Their prominence in history and social science. Distinction between teleological and quasi-teleological explanations. Dependence of the latter on nomic connections. Their prominence in the life sciences.

33

2. Behavior and action. The inner and the outer aspect of an action. Muscular activity as the immediate outer aspect of action. The result of an action distinguished from its causal antecedents and its consequences. Action and for­ bearance.

86

3. The relation between the inner and the outer aspect of an action. The view of the former as a humean cause of the latter. This view is contested by supporters of the Logi­ cal Connection Argument.

91

4. Practical inference. Is it logically conclusive? Its relation to teleological explanations. The premises of a practical in­ ference describe a volitative-cognitive complex.

96

5. Practical inference is concerned with the necessary means to a given end of action. Intention and the supposed ability of the agent to achieve the object of intention.

98

6. In the formulation of a practical inference account must be taken of the possibility that the object of intention is in the future and that the agent may become prevented from carrying his intention into effect.

10 3

7. How does one establish that an agent sets himself to do a certain thing? The burden of verification shifted to the premises of a practical inference.

107

8. How does one establish the presence of an intention and cognitive attitude in an agent? The burden of verifica­ tion shifted to the conclusion of a practical inference. Inten­ tional behavior is a meaningful gesture. The gesture has meaning only in the setting of a story about the agent.

110

9. The question of the compatibility of a causal and a teleological explanation of behavior. The two explanations have different explananda. Distinction between the intentionalist understanding of behavior as action and the teleo­ logical explanation of action as a means to an end.

1 18

10. The question of compatibility reconsidered. The intentionalist interpretation of behavior as action is contin­ gently related to the existence of a humean cause of the behavior. Belief in universal causation a dogma which can­ not be proved true on a priori grounds.

125

IV .

Explanation in History and the Social Sciences

1.

Orders of acts of intentionalist understanding of be­ havioral data. Individual and group behavior. An answer to the question “ W hat is this?” colligates facts under a new concept. “ Emergent qualities” in group behavior.

13 2

2. Genuine causal explanations in history and social science. Their role as links between explanans and explanandum in explanations which are not (genuinely) causal.

>35

3. Quasi-causal explanations in history. The shots at Sara­ jevo and the outbreak of the First W orld W ar as example. How the happening of events affects the motivation back­ ground of practical inferences ending in action.

*39

4.

External and internal changes in the motivation back-

ground of action. The importance of technological changes —a paradigm of explanation for the social process.

M4

t;. Making people do things and the notion of normative pressure. Nonnative pressure has a teleological background which is built up under the influence of reward and punish­ ment. This background can be more or less remote from the individual action. In the limiting case normative pres­ sure deteriorates to a causal mechanism of stimulus and (conditioned) response.

‘45

6. Distinction between rules which regulate conduct and rules which define various social practices and institutions. Rules of the second kind do not exert normative pressure and do not figure in the teleological explanation o f behavior. But they are of prime importance to the understanding of behavior—and therefore to the descriptive tasks o f anthro­ pologists and social scientists.

* 5*

7. Quasi-teleological explanations in history. T h e attribu­ tion of (new) significance to earlier events in the light of later happenings. Why, for conceptual reasons, there can be no such thing as a complete account of the historical past.

1 53

8. Cybernetic explanations of purposefulness. T h e work­ ing of feedback processes in history and the life o f societies is not humean causation under covering laws but m otiva­ tional necessitation through practical inferences. Feedback mechanisms and the “ negation of negation.” T h e cybernetic and systems-theoretic reinterpretation of hegelian and marxist key notions.

156

9. Distinction between two ideas of determinism in his­ tory. Determinism as predictability. Predictability on the macro- and the micro-level. The supposed role of probability and the laws of great numbers for the reconciliation o f free­ dom and necessity. The manipulability of systems from outside by an experimenter— and from inside by the subjects of the behavioral study. The fallacies of “ historicism.”

160

10. Determinism as intelligibility of individual action and of the historical process. The limits of teleological explan-

Contents ation, as that of causal explanation, a matter of experience. The claim that history has an immanent goal transcends the boundaries of a “ scientific” study of man and society.

l6 5

Notes

169

References

207

N a m e Index

223

Subject Index

226

B XV«

E X P L A N A T IO N A N D U N D E R S T A N D IN G

I

Two Traditions

l.

Scientific inquiry, seen in a very broad perspective, m ay

be said to present two main aspects. O ne is the ascertaining and discovery of facts, the other the construction o f hypotheses and theories. These two aspects of scientific activity are sometimes termed descriptive and theoretical science. Theory-building can be said to serve two main purposes. O ne is to predict the occurrence of events or outcomes of experi­ ments, and thus to anticipate new facts. T h e other is to explain, or to m ake intelligible facts which have already been recorded. T hese classifications are useful for the purposes o f a first ap­ proxim ation, bu t they m ust not be taken too rigidly. T h e dis­ covery and description o f facts cannot always be conceptually separated from a theory about them and is often an im portant step towards an understanding o f their nature.1 Prediction and explanation again are sometimes regarded as basically identical processes o f scientific thinking— differing only in the time per­ spective, so to speak.2 Prediction looks forward from what is

i

Explanation and Understanding to what will come, explanation usually looks back from w hat is to what went before. But, it is argued, the terms of the predic­ tive and the explicative relations are similar, and so is the re­ lation linking them. T he former are some facts, the latter is a law. This view of prediction and explanation, however, m ay be challenged.3 T o challenge it is to question the role o f general laws in scientific explanation and to raise the problem o f whether theory-building is intrinsically the same endeavor in the natural sciences and in the humanistic and social disciplines. Some of the problems concerning the interrelation o f the various concepts just mentioned— description, explanation, pre­ diction, and theory— m ay usefully be considered in the light o f intellectual history. Two main traditions can be distinguished in the history o f ideas, differing as to the conditions an explanation has to satisfy in order to be scientifically respectable. T h e one tradition is sometimes called aristotelian, the other g alilean .4 T h e nam es suggest that the first has very ancient roots in the intellectual history of man, while the second is of relatively recent origin. There is some truth in this, but it should be taken with a grain of salt. W hat I here call the galilean tradition has an ancestry going back beyond Aristotle to Plato.5 O ne should also bew are of thinking that the aristotelian tradition today represents merely the fading survival of obsolete elements from w hich science is gradually becoming “ liberated.” As to their views of scientific explanation, the contrast be­ tween the two traditions is usually characterized as causal versus teleological explanation.6 T h e first type of explanation is also called mechanistic,7 the second finalistic. T h e galilean tradition in science runs parallel with the advance o f the causal-m echa­ nistic point of view in man’s efforts to explain and predict phe­

2

Two Traditions nomena, the aristotelian tradition with his efforts to make facts teleologically or finalistically understandable. I shall not attem pt to survey the development of the two traditions from their beginnings. N or shall I try to assess their relative importance to the advancement of science. I shall re­ strict m y apergu in time to the era roughly from the mid-nine­ teenth century to the present day, with the emphasis on recent developments. I shall further restrict its scope to methodology, by which I mean the philosophy of scientific method. 2.

T h e great awakening or revolution in the natural sciences

during the late Renaissance and the Baroque era was to a certain extent paralleled in the nineteenth century in the systematic study of man, his history, languages, mores, and social institu­ tions. T h e work of R an ke and M om m sen in historiography, W ilh elm von H um boldt, Rasm us Rask, Jacob G rim m in lin­ guistics and philology, T ylo r in social anthropology, is compara­ ble to the achievements, two or three centuries earlier, of Coper­ nicus and Kepler in astronomy, G alileo and N ewton in physics, or V esalius and Harvey in anatom y and physiology. Since natural science was already established on the intellec­ tual stage, and the hum anistic studies with a scientific claim were newcomers, it was b u t natural that one o f the chief issues of nineteenth-century m ethodology and philosophy of science concerned the relationship between these two main branches of empirical inquiry. T h e principal stands on this' issue can be linked with the two main traditions in methodological thinking we have distinguished. O ne stand is the philosophy of science most typically repre­ sented by Auguste C om te and John Stuart M ill. It is usually called positivism. T h e nam e was coined by Com te, but used

3

Explanation and Understanding with due caution it is also appropriate for the position of M i l l 8 and for an entire intellectual tradition extending from C o m te and M ill not only down to the present day but also upw ard in the stream of time to Hume and the philosophy of the E n ­ lightenment. One of the tenets of positivism 9 is methodological monism, or the idea of the unity of scientific m ethod amidst the diversity of subject matter of scientific investigation.10 A second tenet is the view that the exact natural sciences, in particular m ath e­ matical physics, set a methodological ideal or standard w hich measures the degree of development and perfection of all the other sciences, including the hum anities.11

A

third

tenet,

finally, is a characteristic view of scientific explanation.12 Such explanation is, in a broad sense, “ causal.” 13 It consists, m ore specifically, in the subsumption of individual cases under h yp o ­ thetically assumed general laws of nature,14 including “ hum an nature.” 16 T he attitude towards finalistic explanations, i.e. towards attempts to account for facts in terms o f intentions, goals, purposes, is either to reject them as unscientific or to try to show that they can, when duly purified o f “ an im ist” or “ vitalist” remains, be transformed into causal explanations.16 Through its emphasis on unity of method, on the m athe­ matical ideal-type of a science, and on the im portance o f general laws to explanation, positivism is linked w ith th at longer and more ramified tradition in the history of ideas w hich I have here called galilean.17 3.

Another stand on the question of the relationship betw een

the sciences of nature and o f man was a reaction against positivism. The antipositivist philosophy of science w hich b e ­ came prominent towards the end of the nineteenth century is a much more diversified and heterogenous trend than posi­

4

Two Traditions tivism. T h e name “ idealism”

which is sometimes used to

characterize it is appropriate only for some facets of this trend. A better nam e for it seems to m e to be hermeneutics. (See below, pp. 29ff.) Representatives of this type of thought in­ cluded some em inent Germ an philosophers, historians, and so­ cial scientists. Perhaps the best known of them are Droysen, Dilthey, Sim m el, and M ax W eber. W indelband and Rickert of the neo-kantian Baden School are related to them. T h e Italian Croce and the em inent British philosopher of history and art Collingwood may be said to belong to the idealist wing of this anti-positivist trend in methodology. A ll these thinkers reject the methodological monism of positivism and refuse to view the pattern set by the exact natural sciences as the sole and supreme ideal for a rational understanding of reality. M any of them emphasize a contrast between those sciences which, like physics or chemistry or physiology, aim at generalizations about reproducible and pre­ dictable phenomena, and those which, like history, want to grasp the individual and unique features of their objects. W indelband coined the label “ nom othetic” for sciences which search for laws, and “ ideographic” for the descriptive study of individuality.18 T h e antipositivists also attacked the positivist view of ex­ planation. T h e G erm an historian-philosopher Droysen appears to have been the first to introduce a methodological dichotomy which has had great influence. H e coined for it the names

explanation and understanding, in G erm an Erklären and Verste­ hen .19 T h e aim of the natural sciences, he said, is to explain; the aim of history is to understand the phenomena which fall within

its domain. T hese methodological ideas were then

worked out to systematic fulness by W ilhelm D ilthey.20 For the entire dom ain of the understanding method he used the

5

Explanation and Understanding name Geisteswissenschaften. There is no good equivalent in English, but it should be mentioned that the word was origi­ nal!) coined for the purpose of translating into G erm an the English term “ moral science.” 21 Ordinary usage does not make a sharp distinction between the words “ explain” and “ understand.” Practically every ex­ planation, be it causal or teleological or of some other kind, can be said to further our understanding of things. B u t “ under­ standing” also has a psychological ring which “ explanation” has not This psychological feature was emphasized by several of the nineteenth-century antipositivist m ethodologists, perhaps most forcefully by Simmel who thought that understanding as a method characteristic of the humanities is a form o f empathy (in German Einfühlung) or re-creation in the m ind o f the scholar of the mental atmosphere, the thoughts and feelings and motivations, of the objects of his study.22 It is not only through this psychological twist, how ever, that understanding may be differentiated from explanation. Understanding is also connected with intentionality in a w ay explanation is not. One understands the aims and purposes o f an agent, the meaning of a sign or sym bol, and the significance of a social institution or religious rite. T h is intentionalistic or, as one could perhaps also call it, sem antic dim ension o f un­ derstanding has come to play a prom inent role in m ore recent methodological discussion. (C f. below, Sect. i o . ) 23 If one accepts a fundamental methodological cleavage b e­ tween the natural sciences and the historical Geisteswissen­

schaften, the question will immediately arise o f w here the social and behavioral sciences stand. T hese sciences were born largely under the influence of a cross pressure o f positivist and antipositivist tendencies in the last century. It is therefore not surprising that they should have become a battleground for

6

Two Traditions the two opposed trends in the philosophy of scientific method. T he application of m athem atical methods to political economy and other forms of social study was an inheritance of the eighteenth-century Enlightenm ent which

found

favor with

nineteenth-century positivists. C om te him self coined the name “ sociology” for the scientific study of human society.24 O f the two great sociologists of the turn of the century, E m ile D ürk­ heim was essentially a positivist as far as his methodology is concerned,25 while in M ax W eber a positivist coloring is com­ bined with emphasis on teleology (“ zweckrationales Handeln ” ) and em pathic understanding (“ verstehende Soziologie” ) .26 4.

Hegel and M arx are two great philosophers o f the past

century who have had a profound and enduring influence, not least on ways o f thinking about m ethod, but whom it is diffi­ cult to place in relation to either nineteenth-century positivism or the reactions against positivism.27 Hegelian and marxist thinking about method puts strong emphasis on laws, universal validity, and necessity.28 In this it resembles, at least super­ ficially, the positivist,

(natural-)science-oriented

trend. B u t

the idea of law which figures in the writings of Hegel and M arx when they discuss, for example, the historical process, is very different from the idea of law which underlies ( “ galilean” ) causal explanations. Similarly, the dialectic schema of developm ent through thesis, antithesis, and synthesis is not a

causal(istic) pattern of thought.29 T h e hegelian and marxist ideas of law and of developm ent are closer to what we should call patterns of conceptual or logical connections.30 In this respect they resemble the methodological ideas of an intentionalistic and teleological type expounded by such antipositivist philosophers as Croce and Collingw ood, who were influenced by hegelianism.

7

Explanation and Understanding Hegel thought of himself as a follower of A ristotle.31 U n like the Master Philosopher, Hegel had little understanding of natural science. In this the spirit of his thinking is alien to that of positivism and has close affinities with that o f the ph iloso­ phers of the Geisteswissenschaften. B u t notw ithstanding this "humanist” as against “ naturalist” emphasis, it seems to m e true to say that Hegel is the great renewer— after the M id d le Ages and therefore necessarily in opposition to the platonizin g spirit of Renaissance and Baroque science— o f an aristotelian tradition in the philosophy of method. F o r H egel, as for A ris­ totle, the idea of law is primarily that of an intrinsic con nec­ tion to be grasped through reflective understanding, n o t that of an inductive generalization established b y observation and experiment. For both philosophers, explanation

consists in

making phenomena teleologically intelligible rather than pre­ dictable from knowledge of their efficient causes.32 W h e n it is seen in its affinity and relationship to H egel, nineteenth-cen­ tury antipositivist methodology as a whole can be linked w ith an older aristotelian tradition, superseded three centuries earlier by a new spirit in the philosophy o f science whose m ost im ­ pressive champion was G alileo.33 5.

The heyday of positivism in the m iddle o f the nineteen th

century was succeeded by an antipositivist reaction tow ards th e end and round the turn of the century. B u t in the decades be­ tween the two world wars positivism returned, m ore vigorous than ever. The new movement was called neo-positivism or logical positivism, later also logical em piricism . T h e attribu te “ logical” was added to indicate the support w hich the revived positivism drew from the new developm ents in form al logic. T he revival of logic after h alf a m illennium o f d ecay and stagnation— from roughly 1350 to 1850 not counting th e b ril­

8

Two Traditions liant but isolated contributions of Leibniz in the seventeenth century— has in itself been of major importance to method­ ology and the philosophy of science. But formal logic can hardly be said to be intrinsically connected with positivism or with a positivist philosophy of science. T h e alliance between logic and positivism in this century is therefore a historical accident rather than a philosophical necessity. T h e logical positivism of the 1920’s and 19 30 ’s was a main, though by no means the sole, tributary out of which grew the broader current of philosophical thought nowadays commonly known as analytical philosophy. It would be quite wrong to label analytical philosophy as a whole a brand of positivism. Bu t it is true to say that the contributions of analytical philosophy to m ethodology and philosophy of science have, until recently, been predom inantly in the spirit of positivism, if by “ positivism ” one understands a philosophy advocating methodological monism, m athem atical ideals of perfection, and a subsumption-theoretic view of scientific explanation. T here are several reasons for this. O ne is connected with a bifurcation of analytic philosophy into two main substreams. One is the trend called linguistic philosophy or ordinary language philosophy. Its main source of inspiration was the later philosophy o f W ittgenstein and its flourishing center at Oxford in the 19 50 ’s. O f this subcurrent one m ight say that it is intrinsically disposed against positivism, though until recently this disposition remained latent rather than manifest. Ordinary language philosophy has, for understandable reasons, taken relatively little interest in the philosophy of science. T h e other subcurrent is a com pletely different case. It is the heir to the logical atomism of Russell and the early W ittgenstein and to the neopositivism of the V ien n a Circle. Its predom inant interest can rightly be said to have been the

9

Explanation and Understanding philosophy of science. But, because of its very ancestry, its intrinsic disposition has been positivistic. It also largely shares with nineteenth-century positivism an im plicit trust in progress through the advancement of science and the cultivation of a rationalist “ social engineering” attitude to human affairs.34 Analytic philosophers of science were long concerned alm ost exclusively with questions of the foundations of m athem atics and the methodology of the exact natural sciences. T h is m ust be understood partly against the background of the im pact of (mathematical) logic on that type of philosophy. G rad u ally, however, the methodology of the behavioral and social sciences and of history has begun to interest analytic philosophers, no doubt partly as a consequence of the invasion of these sciences by exact methods. W ith this shift in the field of interest, analytic philosophy of science entered the traditional battle­ ground between positivist and antipositivist m ethodology, and the old controversies flared up again in the m id-century. T h e immediate source for the revived debate was a m odern version of the old positivist theory of scientific explanation. 6.

T he discussion of problems of explanation w ithin the tradi­

tion of analytic philosophy received its decisive im petus from Carl Gustav Hempel’s classic paper “ T h e F u n ction o f G en eral Laws in History,” published in the Journal o f Philosophy in 1942. Logical positivists and other analytical philosophers had already put forward views of explanation sim ilar to those o f Hempel.35 Essentially, all these views are variants o f the ex­ planation theory espoused by the classics o f positivism , in particular by M ill. In retrospect, it seems almost an irony o f fate th at the fullest and most lucid formulation o f the positivist theory o f explanation should have been stated in connection w ith the

10

Two Traditions subject m atter for which, obviously, the theory is least suited,

viz. history. B u t it is probably chiefly for that very reason that Hem pel’s paper has provoked such an immense amount of dis­ cussion and controversy. T h e hem pelian theory of explanation has become known as the Covering L aw M odel (or T h eo ry). T h e nam e was in­ vented by one o f the theory’s critics, W illiam D ray.36 A n alternative, and perhaps better, nam e for it would be the Subsumption T heory of Explanation. In a num ber of later publications H em pel has amplified, elucidated, and in some details modified his original opinions.37 H e has also distinguished between two submodels o f the general covering law model of explanation. W e shall here call them the Deductive-Nom ological and the Inductive-Probabil­ istic models.38 T h e first can be schem atically described as follow s: Let E be an event which is known to have taken place on some occasion and which stands in need o f explanation. W h y did E happen? In order to answer the question, we point to certain other events or states o f affairs E i , . . . , E m and to one or several general propositions or laws L lt . . . , L n such that the occurrence o f E (on the occasion in question) follows logically from those laws and the fact that those other events (states) have occurred (obtain ). E in the above schematic description of H em pel’s deductivenomological m odel is called the explanandum or explicandum. I shall also call it the object of explanation. E u . . . , E,„ I shall call the explanans or explication. T h ey can also be called the basis o f the explanation. L lf . . . , L „ are the “ covering law s" under which the explanation subsumes the explanans and the explanandum.3Ö O ne m ight ask whether H em pel’s model applies only to 11

Explanation and Understanding objects of explanation which are events. O ften w e are anxious to know, not why an event has taken place, b u t w hy a certain state of affairs obtains or fails to obtain. T h is case, too, can obviously be fitted into Hem pel’s mold. It m ay even be re­ garded as the more fundamental case in th at the notion of an event can be analyzed (defined) with the aid o f the notion of a state of affairs. An event, one could say, is a pair o f suc­ cessive states.40 Another question which our description o f the m odel raises is whether the events or states E lt . . . , E „ „ w hich constitute the basis of the explanation, must have occurred or com e into being earlier than E , or whether they can be sim ultaneous w ith or perhaps even later than E . T h is is an im portan t question, some aspects of which we shall discuss later. W h e n th e events E( are earlier than the object of explanation E , w e shall refer to them as antecedents of E . Hempel’s own, now famous, exam ple o f a deductive-nom ological explanation is typically one in w hich the explanandum is an event and the explanans consists o f antecedent events and states.41 W hy did the radiator o f m y car bu rst du rin g the night? The tank was brim-full of water; the lid w as tightly screwed on; no antifreeze liquid had been added; the car had been left in the courtyard; the tem perature during the night fell unexpectedly to well below zero. T h ese were th e antece­ dents. In combination with the laws of physics— in particular the law that the volume of water expands w hen it freezes— they explain the bursting of the radiator. F ro m kn ow ledge of the antecedents and the laws we could have predicted the event with certainty. This is indeed a good exam ple o f an explanation, but not of the types of explanation for w hich his­ torians look.

12

Two Traditions Our discussion of the subsumption theory of explanation will be limited almost exclusively to the deductive-nomological model. A brief presentation, however, will be given here of the inductiveprobabilistic model, on which we shall also make a critical com­ ment.42 The object of an inductive-probabilistic explanation, too, is an individual event E . The basis is a set of other events or states Eu . . . , E m. The covering law, the “ bridge” or “ tie” connecting the basis with the object of explanation, is a probability-hypothesis to the effect that on an occasion when E ly . . . , E mare instantiated, it is highly probable that E will occur. Here it is pertinent to ask in what sense, if any, could this type of basis and covering law be said to explain the actual occurrence of an event? 43 W hat makes a deductive-nomological explanation “ explain,” is, one might say, that it tells us why E had to be (occur), why E was necessary once the basis is there and the laws are accepted. But it is part and parcel of an inductive-probabilistic explanation that it admits the possibility that E might have failed to occur. It therefore leaves room for an additional quest for explanation: why did E , on this occasion, actually occur and why did it not fail to occur? It would be the task of a deductive-nomological explanation to answer this question. Sometimes we can answer it. This happens when we are able to add to the basis of explanation some further event or state E „ +1 such that, according to accepted laws, an event of the kind E will occur on all occasions when events of the kind E,, . . . , E m+1 jointly materialize.44 One could now make a distinc­ tion and say that, failing such additional information which gives us a deductive-nomological explanation of E , we have not explained why E occurred, but only why E was to be expected. Let there be a probability law (hypothesis) to the effect that, when

Ex,

. . . , E mare instantiated, it is probable to degree p that

E will occur, where p is a mediocre or low probability. Then no

c

*3

Explanation and Understanding one would say that the probability law explains the actual occur­ rence of E. But one might be able to use the information contained in the law for deriving another probability law which says that it is highly probable that the relative frequency with which E will occur on so many occasions when E „ . . , , E „ are instantiated is near the value p. The occurrence of E with this relative frequency is another individual event. This event is here to be expected. There is a characteristic use of probability laws for predicting with high probability relative frequencies of occurrences of events, whose probabilities are any value— high, low, or mediocre. T h e case when the frequency-event is the occurrence of E itself, i.e. the occurrence of E with a relative frequency 1 on a given occasion, is a limiting case of a more general pattem for using probabilities in making predictions. The hempelian inductive-probabilistic model, I should therefore say, is just a special instance of a characteristic use of the probability-calculus for predictive purposes. The two models are much more different than is often thought. It is a primary function of the deductive-nomological model to explain why certain things happened. Therefore it also tells us, secondarily, why these things were to be expected. They could have been ex­ pected, since they had to happen. W ith the inductive-probabilistic model the roles are reversed. It explains in the first place w h y things which happened were to be expected (or not to be expected). O nly in a secondary sense does it explain why things happened, viz. “ be­ cause” they were highly probable. It seems to me better, however, not to say that the inductive-probabilistic model explains w hat hap­ pens, but to say only that it justifies certain expectations and pre­ dictions. This is not to deny the existence of (genuine) explanation pat­ terns in which probability plays a characteristic role. O ne such pattern is the following: Let there be a hypothesis to the effect that the probability o f the event E on a random instantiation of the data E lr . . .

, E mis such

and such, say p. The event is found to occur with a relative fre­

Two Traditions quency markedly different from p in a (large) set of instantiations of the data. W hy this difference? There are two principal ways of accounting for it. One is to attribute it to “ chance.” This way is always open to us, but on the whole is only a last resort. The other way is to look for and discover a further datum E , + l which was also instantiated in the set of instantiations of E u . . . , E m. A probability p', different from p, is associated with the occurrence of E on an instantiation of E lf . . . » E m, E m+1. This probability, let us assume, is such that the relative frequency with which E actually occurred in the set of instantiations was (in the sense ex­ plained before) to be expected. This is like finding a cause ( E m+l) of the observed discrepancy between frequency and probability ( p). The procedure for testing the correctness of the suggested explana­ tion is similar to the procedure which we shall later describe under the name of causal analysis. One could call it probabilistic causal analysis. It occupies an important place in the methodology of explanation, but will not be discussed in further detail in this book.45 7.

H em pel’s (deductive-nomological) explanation model does

not mention the notions of cause and effect. T h e model covers a broader range, of which causal explanations are meant to form a subrange.46 It is debatable whether all causal explana­ tions really conform to H em pel’s schema. It m ay also be asked whether the schema really carries the weight of an explanation unless the covering laws are causal. T h e answer to both questions depends upon the view one takes of the nature of causation. I shall try to show that there are important uses of “ cause” in connection with explanation which do not fit the covering law model. B u t there are also im portant uses which fit it. It seems to me, moreover, clarify­ ing to reserve the term “ causal explanation” for such uses. It is then indisputably true that causa] explanation conforms to

15

Explanation and Understanding the covering law model, though not necessarily to that sim pli­ fied version of it which we presented in the preceding section. The primary test of the claim to universal validity fo r the subsumption theory of explanation is whether the covering law model also captures teleological explanations. One could divide the domain traditionally claim ed for tele­ ology into two subprovinces. One is the domain of the notions of function, purpose (fullness)

and “ organic wholes”

("sy s­

tems” ). The other is that of aiming and intentionality .47 F u n c ­ tion and purpose figure prominently in the biological sciences, intentionality in the behavioral sciences, social study, and historiography. But the domains of biology and of behavioral science largely overlap and so, of course, do the dom ains of function, purpose, and wholes on the one hand and that of aiming and intentionality on the other. T o distinguish them may nevertheless be useful. In 1943, a year after the publication o f H em pel's paper, an important article by Rosenblueth, W iener, and B igelow ap­ peared entitled “ Behavior, Purpose, and T eleology.” 48 T h is is another milestone in the modern history of explanation theory. T he authors wrote independently o f Hem pel. B u t their contri­ bution, seen in the historical perspective, should be regarded as an effort to extend the “ causalist” 49 and with it the subsum p­ tion-theoretic view of explanation to biology and behavioral science. A key notion in the “ causalist” account o f purposefulness, proposed by the three authors of the article,50 is negative feed­

back. A system within which a cause-factor, say a heater, pro­ duces an effect, say the rise of temperature o f a room, m ay be associated with another system such that a “ failure” in the effect of the first system, say a drop in the tem perature below a certain point, produces a “ correction,” say increased heating

16

Two Traditions activity, in the operation of its cause-factor. T h e effect-factor of the second system then gives the operation of the cause-factor of the first system an “ appearance of teleology.” Y et both systems operate according to causal laws. T h e effects within both are explained on the basis of “ initial conditions” consti­ tuted by the cause-factors with the aid of covering laws, linking the causes with their effects. T he authors of the article advanced the thesis that purpose­ fulness can be explained in general by means of some such con­ catenation of causal systems.51 T h e system with the correlated feedback mechanism is called homeostatic or self-regulating. Such mechanisms are highly characteristic of living organisms. For example, the temperature control in vertebrates is a case of a “ heater” with a “ thermostat.” T h e analysis of teleology proposed by Rosenblueth, W iener, and Bigelow appears to agree with a subsumption-theoretic view of scientific explanation. It is not clear, however, whether the explanation pattern of this analysis is precisely that of the hempelian deductive-nomological schema as explained earlier. In order to see whether this is so, the analysis must be further elaborated. Im portant contributions to the logical analysis of self-regulating and other teleological processes have subse­ quently been made by several authors. Prom inent am ong them are Braithwaite and N agel.52 T h e general study of system control and of steering mecha­ nisms, of which homeostasis is only an example, is known as

cybernetics. It has had a great, not to say revolutionizing, influence on modern science, particularly on biology and en­ gineering. Some think it constitutes a scientific contribution of the mid-century comparable in im pact to the revolution in physics brought about by relativity and quantum theories in the early decades o f the century.53 A s I see it, the significance

i~

Explanation and Understanding of tins contribution to m ethodology has been a b ig advance, in the spirit of the galilean tradition, for the “ cau salist” an d mechanist" point of view. A t the sam e tim e it has rein forced some of the main tenets of the positivist p h ilosoph y o f science, in particular the unitary view o f m ethod and th e su bsu m ption theory of explanation. In quarters hostile to p ositivism this is sometimes denied by pointing to the vast differences betw een cybernetic systems and m echanistic system s o f a m ore tradi­ tional and simpler type. Such differences u n d o u b ted ly exist.54 They are reflected in the difference betw een the schem a w hich explains the working of the cybernetic steering and con trol mechanisms and the more “ sim ple-m inded”

schem atism

of

the hempelian covering law model. B u t the differences here, I should say, are essentially in degree o f com p lexity an d logical sophistication of the models, not in the b asic prin ciples o f explanation or in the view o f the nature o f scientific law s. 8.

The notion of a law o f nature, and o f law like u n iform ities

generally, holds a prominent place in the positivist p h ilosop h y of science.55 Hempel’s explanation m odels are ty p ic a lly “ posi­ tivist” in this respect. A more or less dearly articulated view o f th e nature o f natural and other scientific laws has also been ch aracteristic o f positivism. According to this view , rou gh ly sp eakin g, law s enunciate the regular or uniform con com itance (co rrelatio n ) of phenomena, i.e. features appearing in ob jects, states w hich obtain, or events which happen. T h e p rototyp e in stan ce o f a law is either a universal im plication ( “ all A are B ” ) o r a p rob ­ abilistic correlation. Ideally, the phenom ena con n ected b y law should be logically independent. T h is req u irem en t is ro u gh ly equivalent with the idea that the truth-value o f law s is n o t a matter of logical necessity, but contingent u pon th e testim on y

18

Two Traditions o f experience.56 A nd since any law ’s claim to truth always transcends the experience which has actually been recorded, laws are in principle never com pletely verifiable. Consider the following search for an explanation: W h y is this bird black? Answer: It is a raven and all ravens are black. The

answer accords

with

H em pel’s deductive-nomological

schema. B u t does it really explain why the bird is b lack ?57 Unless we are, as philosophers, wedded to the view that any subsumption of an individual case under a general proposition is an explanation, I think we instinctively doubt whether the answer is positive. W e should like to know why ravens are black, what it is about them that “ is responsible for” the color which, so we are told, is characteristic of them all. W h at is required, if our search for an explanation is to be satisfied, is that the basis of the explanation be somehow more strongly related to the object of explanation than simply by the law stating the universal concomitance of the two characteristics ravenhood and blackness. T here are two ways, it seems, in which this requirement could be satisfied. One is to look for a “ cause” o f the blackness o f ravens, i.e. for some other characteristic o f birds of this species which accounts for their color. T h e other is to give the suggested answer explicative force by declaring that blackness is, in fact, a defining characteristic of the species raven. T o accept either answer is to regard the concomitance of the two features as not merely universal, but also, in some way, necessary. T h e second approach confronts us with a view o f natural laws which m ay be regarded as an alternative to the classic positivist view. According to this alternative conception, a scientific law can be immune to refutation by experience since its truth is analytical, logical. Agreem ent with the law is then a standard whereby individual cases are classified as falling, or

*9

Explanation and Understanding not falling, under the generic phenom ena connected b y the law. All A are B, so if a thing purporting to be an A is found not to be B, then it is not really an A after all. Such standards for judging things are man-made agreements adopted in the process of concept formation. T h is view is therefore called

conventionalism68 When carried to their extremes, positivism an d con ven tion al­ ism are opposed views, in roughly the sam e sense in w hich extreme empiricism and extreme rationalism are opposed. B u t it is relatively easy to find a compromise betw een the extrem es. A sensible positivist will agree that some scientific principles have the character of analytic truths w hile others are clearly empirical generalizations. He will note, m oreover, th at the border between the two categories has often fluctuated in the course of the historical development o f a science.59 The conventionalist view of scientific law s, one m igh t say, contains no conceptual elements which are alien to a positivist philosophy of science. Although

conventionalism

o ften

at­

tacked positivism, and vice versa, it was also felt th at the two positions have much in com mon.60 A com m on ten et o f both views of scientific laws is that they deny the existence o f a “ middle” thing, sometimes called natural necessity, distin ct from empirical generality on the one hand and logical neces­ sity on the other. For the same reason both positivism an d con ven tionalism would deny that causal laws acquire their “ explicative fo rce” from the alleged fact that they state necessary connections b e­ tween events in nature. T o question the idea th at universal truth must be either accidental (contingent, em pirical) o r logi­

cally necessary is a much more serious challenge to positivism than that presented by conventionalism. This challenge, however, is traditional and 20

is associated

Two Traditions with the opposition between the "aristotelian” and “ galilean” trends in the philosophy of science. It is of particular interest here to note that it has emerged, acquiring a new urgency, within analytical philosophy itself, where it springs from two sources. One is the rebirth, in the mid-century, of interest in modal

logic and in the philosophy of the modal concepts. Philosophi­ cal logicians have become fam iliar with the view that the

logically necessary and possible is only a species of a more comprehensive genus within which a variety of forms of neces­ sity and possibility m ay be distinguished. T h e revival of modal logic has not in itself rehabilitated the notion of natural necessity as distinct from logical necessity and “ mere” acci­ dental generality. T h e idea remains controversial and is re­ garded by many analytical philosophers as suspect or definitely unsound. B u t modal logic has paved the way for a change in the positivist view of natural laws that was long accepted by analytical philosophers.61 A more im mediate challenge to the accepted view o f natural laws in the positivist tradition has come from the problem of

counterfactual conditionals. T his problem was posed by the classic papers of Chisholm

(19 4 6 )

and Goodm an

(19 4 7 ).

Since then it has been discussed in innum erable articles and books. Sim plifying somewhat, its relevance to the problem of the nature of laws is as follows: Sometimes our conviction, if we have one, that if it had not been the case that p then it would have been the case that q, is grounded on our belief in a n o in ic62 or lawlike connection between the (generic) propositions p and q. N ot just any valid universal im plication connecting the two propositions would suffice as a ground. T h e question then arises of how to charac­ terize lawlikeness or how to distinguish (nonlogical) nomic

21

Explanation and Understanding connection from ‘'accidental” universal concom itance.63 In an essay some fifteen years ago I argued that the notion o f a counterfactual conditional is itself involved in this distinction — and therefore cannot be elucidated by m eans of it.64 T h e ■‘moral'’ to be drawn from the discussion o f the problem of counterfactual conditionals is that the hallm ark o f nom ic con­ nection, lawlikeness, is necessity and not universality .65 I f this is correct, it shatters the positivist conception of law , though not necessarily the validity of the subsum ption theory of explanation. I shall not discuss counterfactual conditionals here, but I hope to be able to throw a little m ore ligh t on the nature of the “ necessity” which makes some universal regu­ larities nomic. 9.

How deep into the province of teleology do cybernetic ex­

planations penetrate? D o they extend beyond the frontiers o f biology into the sciences of man? O ne could answ er this last question by pointing to the great im pact o f cybernetic think­ ing on economics, social psychology, and even on the theory of law.66 But this answer is not very illum inating. It does not tell us whether the use in these fields o f ideas borrow ed from cybernetics gives us explanations of the subsum ption-theoretic type. M y own surmise is that on the whole they do no t. I f I am right in this and also in thinking that the cybernetic ex­ planations of homeostatic systems etc. in biological science do conform to subsumption-theoretic patterns, then the “ cyber­ netics” of social science differs much m ore

from

th at o f

biology than the assimilation o f various research activities u n ­ der this common heading might suggest. The aspects of teleology captured by cybernetic explanations conforming to a covering law model are, I think, prim arily the aspects void of intentionality. A m ong things to w hich inten-

22

Two Traditions tionality is attributed actions occupy a prominent place. T he final test of universal validity of the subsumption theory of ex­ planation is whether it can successfully account for the ex­ planation o f actions. M an y analytic philosophers, perhaps the majority, think the theory passes the test. Actions are prompted by motives; the force of motives lies in the fact that agents are disposed to follow characteristic patterns of behavior; such patterns (dis­ positions) provide the “ laws” linking the motives to the action in the individual case. T his is a consciously oversimplified ac­ count of an idea which in more or less sophisticated variations continues to exercise a strong fascination over the philosophic im agination.67 It is related to the idea that actions have causes, and thus also related to a determinist position in the old question of “ the freedom of the w ill.” There is, however, also opposition among analytic philos­ ophers to this idea of the validity of a subsumption-theoretic model of explanation of action. One line o f opposition has come from (analytic) philos­ ophers engaged in the m ethodology o f history. H ere the cri­ tique has focused on the role of general laws in history— the them e which provided the very title o f H em pel’s paper, in which the covering law model was first clearly articulated. W h y is it that the explanations which historians actually give seldom, if ever, make reference to general laws? Supporters of the subsumption theory of historical explanation have, o f course, been well aware o f this. T h eir reactions to it, however, have varied a great deal. In H em pel’s view the reason why the full form ulation of general laws is missing from historical explanations is primarily that the laws are too com plex and our knowledge of them not sufficiently precise. Explanations given by historians are in a

Explanation and Understanding characteristic sense elliptical or incom plete. T h e y are, strictly speaking, only explanation sketches. “ A n explanatory accou n t,” Hentpel says, “ may suggest, perhaps quite vividly and per­ suasively. the general outline of what, it is hoped, can eventu­ ally be supplemented so as to yield a m ore closely reasoned argument based on explanatory hypotheses which are indicated more fully.” 68 According to Popper, another prom inent representative o f the subsumption theory of explanation, the reason w h y general laws are not formulated in historical explanations is that the laws are too trivial to deserve explicit m ention. W e know them and take them for granted im plicitly.69 A radically different view of the role o f laws in historical explanations is expounded by W illiam D ray in his im portan t book Laws and Explanation in History, published in 19 5 7 . T h e reason why historical explanations do not norm ally m ake refer­ ence to laws is not that the laws are so com plex and obscure that we must be content with a mere sketch, nor that they are too trivial to deserve mention. T h e reason, according to D ray, is simply that historical explanations do not rely on general laws at all. Consider, e.g., the statement that Louis X I V died unpopular, because he pursued policies detrimental to F ren ch national interests.70 How could a covering law theorist defend his claim that there is a law implicit in the explanation? A general law which tells us that all rulers who . . . becom e unpopular, would provide a covering model of the case under discussion only if so many limiting and qualifying conditions w ere added to it that, in the end, it would be equivalent to saying th at all rulers who pursue exactly similar policies to those o f Lou is X IV under exactly similar conditions to those w hich prevailed in France and the other countries affected by L ou is's policies,

Two Traditions become unpopular. If the exact similarity of policies and pre­ vailing conditions is not specified in generic terms, this state­ ment is no “ law ” at all, since of necessity it has only one in­ stance, viz. that of Louis X I V . If the conditions of similarity are specified— which would hardly be possible in practice— we should have a genuine law, but the only instance of this law would be the one it is supposed to “ explain.” T o maintain the law would therefore in either case amount only to a reaffirmation of what has already been stated, viz. that the cause of the aging Lou is’s unpopularity were his unlucky foreign policies. D ray’s criticism of the role o f general laws in historical ex­ planations thus leads to a com plete rejection o f the covering law model. It is illum inating to compare D ray’s Laws and Ex­

planation in History with G ardiner’s The Nature of Historical Explanation, which was published five years earlier ( 19 5 2 ). As I understand the two authors, their “ methodological inten­ tions” are to a large extent similar. B u t whereas Gardiner’s in­ tentions are badly thwarted by the dominant, though perhaps implicit, influence of a positivist philosophy of science, Dray succeeds adm irably in breaking the fetters of positivism in con­ temporary “ analytic” philosophy of history. H e achieves this “ negatively” through his critique o f the covering law model as a tool for historical explanation, and “ positively” through his insistence on the sui generis character of explanation models for hum an action. T h e critique is by far the strongest aspect o f D ray’s work. T h e positive contributions reflect the groping nature o f an “ analytic” philosophy o f action, then still in its first beginnings. T o explain an action is, in D ray’s view, to show that the action was the appropriate or rational thing to do on the occasion under consideration.71 D ray calls this rational explana­

25

Explanation and Understanding tion. He does not succeed in making its nature very clear. He thinks, quite rightly it seems to me, that this type o f ex­ planation has logical peculiarities of its own. B u t he obscures his own point unnecessarily by trying to find these peculiarities in an element of valuation rather than in a type o f teleology.72 Dray’s explanation model resembles traditional ideas about the methodological role of empathy and understanding. H is book does not make contact with more recent continental p h i­ losophy of the Geisteswissenschaften. Instead there is an inter­ esting link with the hegelian trend in Collingw ood (and Oakeshott) ,73 Elizabeth Anscom be’s Intention appeared the same year as D ray’s book. It made the notion of intentionality central to subsequent discussion of the philosophy o f action am on g analytic philosophers.74 Though not directly concerned with the theory o f explana­ tion, Miss Anscombe’s book has made two im portant contri­ butions to this field too. T h e first is her observation that b e ­ havior which is intentional under one description o f it, need not be intentional under another. It thus makes a difference to the explanation of a given item of behavior how it is described,

i.e. understood as being an action. T h is is a case when the distinction between explanation and understanding becom es conceptually significant. (C f. below, C h . I l l , Sect. 2, and C h . IV , Sect. 1.) Miss Anscombe has also drawn attention to the peculiar logical character of the reasoning called in traditional term i­ nology practical syllogisms. T h e idea goes back to A ristotle and was, Miss Anscombe says, one of his best discoveries, though one which later philosophy has lost through m isinterpreta­ tion.75 T he clue to a correct interpretation, however, is not easy to find. Aristotle’s own treatment of the topic is very u n ­

26

Two Traditions systematic and his examples are often confusing. One way of reconstructing the main idea here is the following: T h e starting point or major premise of the syllogism mentions some wanted thing or end of action; the minor premise relates some action to this thing, roughly as a means to the end; the conclusion, finally, consists in use of this means to secure that end. Thus, as in a theoretical inference the affirmation of the premises leads of necessity to the affirmation o f the conclusion, in a practical inference assent to the premises entails action in ac­ cordance with them.76 I think Miss Anscom be is right in saying that the practical syllogism is not a form of demonstration and that it is a reasoning different in kind from the proof syllogism.77 B u t its peculiarities and relation to theoretical reasoning are complex and remain obscure. Practical reasoning is of great importance to the explanation and understanding of action. It is a tenet of the present work that the practical syllogism provides the sciences of man with something long missing from their m ethodology: an explana­ tion model in its own right which is a definite alternative to the subsumption-theoretic covering law m odel.78 Broadly speak­ ing, what the subsumption-theoretic model is to causal explana­ tion and explanation in the natural sciences, the practical syl­ logism is to teleological explanation and explanation in history and the social sciences. T h e works of Anscom be and D ray reflect the increasing interest within analytic philosophy in the concept of action and in forms of practical discourse. T hese pioneer works have been followed by a number of further contributions.79 But it was not until the appearance of Charles Taylor's important book The Explanation of Behaviour in 1964 that this new line in analytic philosophy becam e linked with explanation theory

27

Explanation and Understanding in psychology and the other behavioral sciences. L ik e the con­ tributions of cybernetics, but in a very different spirit, T aylo r’s work has revived the debate concerning teleology in

the

philosophy of science. One can characterize their difference in spirit as one between a galilean and an aristotelian view o f pur­ posive behavior. The efforts and ideas of analytic philosophers o f action have not failed to provoke reaction on the part o f philosophers o f a more positivist turn of mind. T h e applicability o f causal cate­ gories to the explanation of action and behavior generally has been forcefully defended also by a number of recent authors.80 A position somewhat similar to D ray’s in the “ analytic” philosophy of history is held by Peter W in ch in the “ analytic” philosophy of the social sciences. His work The Idea of a Social

Science, published in 1958, is, like D ray’s book, an attack on positivism and a defense of an understanding of social phe­ nomena by methods different in principle from those of the natural sciences. T h e tradition background o f W in ch 's work is partly the “ understanding” methodology of M a x W e b e r and partly the hegelian trend represented in England b y Collingwood and Oakeshott. T h e m ajor influence, however, comes from the later W ittgenstein. W inch’s book can be said to center on the question o f the criteria of social behavior (action). T h e social scientist m ust understand the “ meaning” of the behavioral data w hich he registers in order to turn them into social facts. H e achieves this understanding by describing (interpreting) the data in terms of the concepts and rules which determine the “ social reality” of the agents whom he studies. T h e description, and explanation, of social behavior must em ploy the same concep­ tual framework as the social agents themselves. F o r this reason the social scientist cannot remain an outsider in relation to his

28

Two Traditions object of study in the same sense in which a natural scientist can. This is the core of conceptual truth, one could say, in the psychologistic doctrine of “ em pathy.” Em pathie understanding is not a “ feeling” ; it is an ability to participate in a “ form of life.” 81 W in ch can be said to investigate the a priori o f the meth­ od (s) o f social science. In this sense his book is a contribution to m ethodology.82 Some of W inch's critics seem to think that he regards sociology as an a priori science, i.e. as a study which explains and understands social phenomena by a priori m eth­ ods. T his is a gross misunderstanding.83 W in ch ’s book is difficult and obscure. It is also one-sided, it seems to me, in that it puts too much emphasis on the im ­ portance of rules to the understanding of societal behavior. One misses in it the aspect of intentionality and teleology.84 10.

A challenge to positivist m ethodology and philosophy of

science has thus arisen within the mainstream of analytical philosophy, particularly after the appearance of the three works by Anscombe, Dray, and W inch. T h e years of their publica­ tion, 19 5 7 -19 5 8 , may be said to mark a turning of the tide. T h e critics of positivism among analytic philosophers have usually been writers whose thinking was inspired by the philo­ sophy of the later W ittgenstein. A n orientation to phenom e­ nology and other brands of nonanalytic philosophy on the European continent can also be seen in some of them .86 T his reorientation is to some extent paralleled in continental philosophy by developments disclosing affinities with analytical philosophy. I am thinking chiefly on the coming to prominence in the 1960’s of the current which calls itself hermeneutic or hermeneutic-dialectic philosophy.86 T here are two features o f hermeneutics which are specially D

29

Explanation and Understanding noteworthy as regards affinities with analytical philosophy. T h e first is the central place held in it by the idea o f language and by language-oriented notions such as meaning, intentionality, interpretation, and understanding.87 T h is is reflected in the very name “ hermeneutics,” which means the a it o f interpreta­ tion.38 The problems which occupy herm eneutic philosophers are largely the problems which also pervade the philosophy o f Wittgenstein, particularly in its later phases.89 It would not be surprising if this affinity, once it becomes clearly recognized, should lead to an impact of W ittgenstein on philosophy on the European continent comparable in degree, if not in kind, to his influence on the V ienna school of logical positivism in the 1930’s and on the Oxford school of linguistic analysis in the 1950’s. The second feature of hermeneutic philosophy which makes it more congenial to philosophers of the analytic tradition than some other offshoots of the phenomenological stem is its con­ cern with methodology and philosophy of science.90 In explicit opposition to positivism's idea of the unity of science, her­ meneutic philosophy defends the sui generis character o f the interpretative and understanding methods of the Geisteswissen­

schaften. In this it revives and continues the intellectual inheri­ tance of neokantian and neohegelian antipositivism from the decades around the turn of the century. “ Understanding,” in terms of hermeneutic philosophy, should be distinguished from empathy or Einfühlung. It is a sem antic rather than a psychological category. (C f. above, p. 6.) T h e charge so often made by positivist philosophers that under­ standing is only a heuristic device, which m ay be useful for find­ ing an explanation but which is not constitutive of the concep­ tual nature of the explanation model itself, m ay be valid against some earlier and outmoded versions of the m ethodology o f

3°

Two Traditions em pathy.91 Bu t it is not a fair objection against the method­ ology of understanding as such. A s noted above (Sect. 4) it is difficult to determine the position of Hegel and M arx in relation to positivist and anti­ positivist nineteenth-century philosophy of science. Something similar holds true of marxism as a mainstream of modem thought. A fter the Russian Revolution marxist philosophy was much absorbed in inner struggles for orthodoxy and gradually assumed a monolithic and stiffened appearance. B u t it is obviously rising again to intellectual prominence. It seems, moreover, that one can discern two main trends in this quarter.92 One is predominantly a philosophy of science fostered in the spirit of dialectical materialism. It has recently received fresh impetus from cybernetics and systems-theory and from the ap­ plications generally of mathematical tools to the life and be­ havioral sciences.93 It is not surprising that “ the causalization of teleology” should have a strong appeal to marxist material­ ists 94 Differences in tradition notwithstanding, this trend in marxism has affinities with the branch of philosophy of science in the W est which is the spiritual heir to logical positivism and the unity of science movement.95 It is sometimes called “ posi­ tivist,” but more often “ scientistic” marxism 96 T he second trend within contemporary marxism is the philo­ sophical anthropology which also calls itself socialist hum an­ ism.97 Its sources o f inspiration are partly in the writings of the young M arx, but even more perhaps in the philosophy of H e g e l98 Its dialectic seems closer to that of hermeneutic phi­ losophy than to the “ materialism” of orthodox marxism.99 Its anthropology and humanist political and social philosophy has affinities with existentialism, particularly with the later philos­ ophy of Sartre, which in turn is oriented towards M arx and

Explanation and Understanding H egel.100 Its p h ilosop h y o f scien ce is so m e tim e s a d e fe n s e o f the sciences o f m an again st th e m o n istic p re te n sio n s o f a t h in k ­ ing m oulded under the im p a ct o f th e p ro gress o f n a tu ra l s c ie n c e and technology.

I have tried to relate some developments in the philosophy of scientific method to two great traditions in the history of ideas. W e have seen how in the last hundred years philosophy of science has successively clung to one or the other o f two basically opposed positions. A fter H egel came positivism ; after the antipositivist and partly neohegelian reaction around the turn of the century came neopositivism; now the pendulum is again swinging towards the aristotelian them atics which H egel revived. It would surely be an illusion to think that truth itself un­ equivocally sided with one of the two opposed positions. In saying this I am not thinking of the triviality that both positions contain some truth and that a compromise can be achieved on some questions. This may be so. B u t there is also a basic op­ position, removed from the possibility both o f reconciliation and of refutation— even, in a sense, removed from truth. It is built into the choice of primitives, of basic concepts for the whole argumentation. T his choice, one could say, is “ existen­ tial.” It is a choice of a point of view which cannot be further grounded. There is nevertheless dialogue between the positions, and a kind of progress. T he temporary dom inance o f one o f the tw o trends is usually the result of a breakthrough follow ing a period of criticism of the other trend. W h a t emerges after the break­ through is never merely a restoration o f som ething which was there before, but also bears the impress of the ideas through whose criticism it has emerged. T h e process illustrates w hat

32

Two Traditions Hegel described with the words aufgehoben and auf bewert, perhaps best rendered in English as “ superseded” and “ re­ tained.” T h e position which is in process of becoming super­ seded usually wastes its polemical energies on fighting already outmoded features in the opposed view, and tends to see what is retained in the emerging position as only a deformed shadow of its own self. T his is what happens, for example, when positivist philosophers of science in our days object to Verste­

hen with arguments perhaps valid against Dilthey or Collingwood, or when they mistake W ittgenstein's philosophy of psy­ chology for just another form of behaviorism.

33

II

Causality and Causal Explanation

l.

Philosophers have long been accustomed to m aking a dis­

tinction between the relation of cause and effect on the one hand and the relation of ground and consequence on the other. The first is factual and empirical, the second conceptual and logical. Before the distinction became current, it was often ignored or blurred— particularly by the rationalist thinkers of the seventeenth century. W hen it became more clearly articulated, thanks not least to H um e,1 new problems arose. Perhaps all causal relations are factual. B u t, quite certainly, not all factual relations are causal. W h at, then, are the other distinguishing features of causal relations— in addition to their being empirical? According to Hum e the relation between cause and effect is a regular sequence in time of (instantiations of) generic phenomena. T h at the regularity will continue to hold in the future is an inductive generalization, based on past experience.2

34

Causality and Causal Explanation Since Hume, causation has been something of a problem child of epistemology and the philosophy of science. M any efforts have been made to show either that Hum e’s view of causation was mistaken or that, if one accepts his view, the problem of induction which it leaves open— also often called “ the Problem of H um e” — can be satisfactorily solved.3 These efforts have, on the whole, been unsuccessful, and the unsatisfactory state of the problem of induction has been called “ the scandal of Philosophy.” 4 These troubles are probably one of the reasons why some philosophers have insisted that the idea of causation plays only an insignificant role in science and m ay eventually be exorcised from scientific thinking altogether.5 T h e philosophical difficul­ ties about causation need not then be a burden on the phi­ losophy of science. T h is opinion was expressed forcefully by Bertrand Russell in his famous essay “ On the N otion of Cause.” W ith his characteristic wit Russell wrote: “ A ll philos­ ophers, of every school, imagine that causation is one of the fundam ental axioms or postulates of science, yet, oddly enough, in advanced sciences such as gravitational astronomy, the word ‘cause’ never occurs. . . . T h e law of causality, I believe, like much that passes muster among philosophers, is a relic of a bygone age, surviving, like the monarchy, only because it is erroneously supposed to do no harm .” 6 A nd he continues: “ N o doubt the reason why the old ‘law of causality’ has so long continued to pervade the books of philosophers is simply that the idea of a function is unfamiliar to most of them, and therefore they seek an unduly simplified statement.” 7 One can agree with Russell that “ the law of causality,” what­ ever that means, has no place in science proper but is a typical philosophers’ construction. Russell’s denunciation of the very notion of cause is more controversial. His words seem to sug­

35

Explanation and Understanding gest that this notion is a prescientific ancestor o f the scientific concept of a function. It has been argued against Russell that, even though the words “ cause” and “ effect” and other elements o f causal ter­

minology are not prominent in the advanced theoretical sciences, causal ideas and causal thinking are not as outm oded as the shift in terminology, e.g. from speaking about “ causal” speaking about “ functional”

relationship, m ay

to

suggest. As

Ernest Nagel observes, the notion of cause “ not only crops up in everyday speech, and in investigations into hum an affairs by economists, social psychologists, and historians, it is also per­ vasive in the accounts natural scientists give o f their laboratory procedures, as well as in the interpretations offered b y m any theoretical physicists of their m athem atical form alism .” 8 A n ­ other prominent contemporary philosopher o f science, Patrick Suppes, goes further still and says: “ Contrary to the days when Russell wrote this essay the words ‘ causality’ and ‘cause’ are commonly and widely used by physicists in their m ost ad­ vanced work.” 9 This last may nevertheless be an exaggeration. In trying to assess the importance of causation to science it is wise to remember that the word “ cause,” and causal term s generally, are used with a multitude of meanings. N ot only are “ causes” in human affairs very different from “ causes” o f natural events, but within the natural sciences as well causality is not a hom o­ geneous category. T he notion of cause which I will be discuss­ ing in this chapter is essentially tied to the idea o f action and therefore, as a scientific notion, to the idea of experim ent. It figures importantly, I think, in “ the accounts natural scientists give of their laboratory procedures.” B u t I am less certain whether this is what is involved “ in the interpretations offered by many theoretical physicists of their m athem atical form alism .”

36

Causality and Causal Explanation T h e reason why I nevertheless want to give a basic priority to this “ actionist” or “ experimentalist” notion of cause, is that, in addition to holding an important place in the experimental natural sciences, it seems largely to figure as a prototype for the idea of cause in the discussions of philosophers about uni­ versal causation, determinism versus freedom, interaction of body and mind, etc. I sympathize, however, with those who, like Russell or N orm an Cam pbell,10 have felt that this notion of cause is not so important in theoretically advanced sciences “ such as gravitational astronomy” and who think that in them causal talk can profitably be replaced by talk about various functional relations. But whether or not this attitude is justified, the fact remains that causal thinking, as such, has not been exorcised from science— and that therefore the philosophical problems about causation continue to be central to the philos­ ophy of science. T heir seriousness is felt particularly keenly in the theory of scientific explanation. T h e covering law model was originally thought of as a generalization of ideas associated with causal explanation.11 T h e specific problems about causation seemed to m any to have lost their urgency because of this widening of the conceptual horizon— just as Russell had thought that causation had be­ come philosophically uninteresting because it might be sub­ sumed under the broader category of functional relationship. B u t this is a mistake. A s we have already seen (C h . I, Sect. 8 ), the notion of law which is involved in the subsumptive explanation model is itself problematic. R ecent discussion of these problems has brought to the fore the modal ideas of natural necessity and nomic connection. These ideas in turn are closely associated with the ideas of cause and effect, so much so that one could con-

37

Explanation and Understanding venientlv group them all under the general head of causation. If one insists that the covering law model has explicative force onlv when the laws involved in it express (nonlogical) nom ic connections, this would be tantamount to saying that explana­ tion conforming to the covering law model and causal explana­ tion are, substantially, the same thing. And this would at once turn the problematics associated with the hem pelian m odel of explanation into a modern form of the problem of causation.12

2.

Russell suggested that the place of the notion of cause in

the philosophy of science be taken over by the notion of a function. There is another notion besides that of a function for which a similar claim could be made. T h is is the concept of condition. T h e discussion of cause and effect which I shall conduct here will be in terms of conditionship, and not of functional relationship. One is used to distinguishing between necessary and sufficient conditions. Other condition concepts may also be defined: con­ tributory conditions, substitutable requirements, etc. F o r the present purposes, however, these “ minor” condition concepts will not be needed.13 T he statement that the generic 14 phenom enon (state, event)

p is a sufficient condition of q can, for purposes o f a first ap­ proximation, be explained as follows: W henever p is, q w ill be there too; the presence (occurrence) o f p suffices to ensure the presence (occurrence) of q. T h at p i s a necessary condition of q means that whenever q is, p has to be there too, i.e. the presence (occurrence) of q requires or presupposes the presence (occurrence) of p. If p can be “ manipulated,” i.e. produced or prevented “ at will” (“ experimentally” ), then, by producing p, we can also bring about anything of which it is a sufficient condition, and

Cdusdhty dnJ Cdusdl Expldmtion by removing or preventing p we can ensure that anything of which p is a necessary condition does not occur. A phenom enon can be a necessary and sufficient condition of some other phenomenon. A phenomenon may have several sufficient or several necessary conditions. A condition can also be complex, i.e. a truth-functional compound of some generic phenomena. W ith regard to com plexity and plurality of con­ ditions the following asymmetries between the various kinds of condition must be noted. A com plex sufficient condition is a conjunction of phe­ nomena. M aybe p by itself is not sufficient to guarantee that r will be, nor is q by itself sufficient for this. But if p and q occur together, r is sure to be there too. A complex necessary condition again is a disjunction. M aybe p does not require the presence of q (unconditionally), nor the presence of r (uncon­ ditionally); but p m ay nevertheless require that at least one of the two, q or r, be present. D isjunctive sufficient conditions may be “ resolved” into a plurality of sufficient conditions. If p or q is sufficient for r, then p by itself is sufficient and so is q by itself. Conjunctive necessary conditions may be similarly “ resolved.” If p and q is necessary for r, then p by itself is necessary and also q by itself. These “ asymm etries” of the condition concepts can be in­ terestingly exploited in inductive logic.16 In terms of conditions one can distinguish a variety of causal factors which, when speaking vaguely of “ cause” and “ effect,” are difficult, or even impossible, to separate.16 Condi­ tion concepts are also helpful for making clearer the philos­ ophers’ ideas of (universal) Determ inism and of the (uni­ versal) L aw of Causation. I therefore find it surprising that the theory of condition concepts and its applications has remained

39

Explanation and Understanding relatively little developed and studied. In logic textbooks it is seldom even mentioned. Y et it seems to me em inently well suited as a propedeutics to logic and the m ethodology of science. The usefulness of condition concepts is not in conflict with the fact that they also provoke problems. T h e problems m ay be said to concern the “ place” of condition concepts in logic. Here two principal views confront each other. One places the condition concepts within quantification theory. In a logical language which employs names of individuals and predicates, the universal implication (x) ( P x —> Q x) would then be the “ groundform” of conditionship relations. In a poorer language which employs only propositional variables, conditionship rela­ tions could be formulated, e.g., as tense-logical statem ents, their “ groundform” now being “ whenever p then q,” or in sym bols:

A